peer reviewed

We present a scalable framework designed to craft efficient lightweight models for video object detection utilizing self-training and knowledge distillation techniques. We scrutinize methodologies for the ideal selection of training images from video streams and the efficacy of model sharing across numerous cameras. By advocating for a camera clustering methodology, we aim to diminish the requisite number of models for training while augmenting the distillation dataset. The findings affirm that proper camera clustering notably amplifies the accuracy of distilled models, outperforming the methodologies that employ distinct models for each camera or a universal model trained on the aggregate camera data.

This paper considers the deployment of flexible and high-performance surveillance systems. These systems must continuously integrate new sensors and sensing algorithms, which are autonomous (e.g., capable of making decisions independently of a central system) and possess interaction skills (e.g., capable of exchanging observations). For this purpose, our work proposes adopting an agent-based architecture derived from an organizational and holonic (i.e., system of systems) multi-agent model. It leverages autonomous processing methods, resulting in a scalable and modular multisensor and multimethod surveillance systems. A vehicle tracking case study demonstrates the relevance of our approach in terms of effectiveness and runtime.

Emergency response missions depend on the fast relay of visual information, a task to which unmanned aerial vehicles are well adapted. However, the effective use of unmanned aerial vehicles is often compromised by bandwidth limitations that impede fast data transmission, thereby delaying the quick decision-making necessary in emergency situations. To address these challenges, this paper presents a streamlined hybrid annotation framework that utilizes the JPEG 2000 compression algorithm to facilitate object detection under limited bandwidth. The proposed framework employs a fine-tuned deep learning network for initial image annotation at lower resolutions and uses JPEG 2000’s scalable codestream to selectively enhance the image resolution in critical areas that require human expert annotation. We show that our proposed hybrid framework reduces the response time by a factor of 34 in emergency situations compared to a baseline approach.

Forecasting trajectory is a complex task relying on the accuracy of past positions, a correct model of the agent’s motion and an understanding of the social context, which are often challenging to acquire. Deep Neural Networks (DNNs), especially Transformer networks (TFs), have recently evolved as state-of-the-art tools in tackling these challenges. This paper presents TrajViViT (Trajectory Video Vision Transformer), a novel multimodal Transformer Network combining images of the scene and positional information. We show that such approach enhances the accuracy of trajectory forecasting and improves the network’s robustness against inconsistencies and noise in positional data. Our contributions are the design and comprehensive implementation of TrajViViT. A public Github repository will be provided.

JPEG-XS offers low complexity image compression for applications with constrained but reasonable bit-rate, and low latency. Our paper explores the deployment of JPEG-XS on lossy packet networks. To preserve low latency, Forward Error Correction (FEC) is envisioned as the protection mechanism of interest. Although the JPEG-XS codestream is not scalable in essence, we observe that the loss of a codestream fraction impacts the decoded image quality differently, depending on whether this codestream fraction corresponds to codestream headers, to coefficient significance information, or to low/high frequency data. Hence, we propose a rate-distortion optimal unequal error protection scheme that adapts the redundancy level of Reed-Solomon codes according to the rate of channel losses and the type of information protected by the code. Our experiments demonstrate that, at 5% loss rates, it reduces the Mean Squared Error by up to 92% and 65%, compared to a transmission without and with optimal but equal protection, respectively.

The treatment of mobile tumors remains complex in radiotherapy due to breathing-related organ movements. A solution to ensure target coverage during the process is to plan the treatment taking into account safety margins. One way to significantly reduce these safety margins would be to adapt the treatment in real time using image-guided radiation therapy. The acquisition of x-ray projections during treatment is commonly used to localise the tumor in 2D but doesn’t provide 3D information. Hence, the aim of this work is to reconstruct a high resolution 3D image based on a single radiograph in order to know the 3D position of the tumor and the organs. This is done using a convolutional neural network. The results show that the proposed method is able to reconstruct a 3DCT based on a 2D projection x-ray only. The normalized root mean square error is computed between the ground truth 3DCT and the predicted 3DCT, and the mean of this metric is between 0.02713 and 0.02776 depending on the patient.

Particle filtering is a powerful tool for target tracking. When the budget for observations is restricted, it is necessary to reduce the measurements to a limited amount of samples carefully selected. A discrete stochastic nonlinear dynamical system is studied over a finite time horizon. The problem of selecting the optimal measurement times for particle filtering is formalized as a combinatorial optimization problem. We propose an approximated solution based on the nesting of a genetic algorithm, a Monte Carlo algorithm and a particle filter. Firstly, an example demonstrates that the genetic algorithm outperforms a random trial optimization. Then, the interest of non-regular measurements versus measurements performed at regular time intervals is illustrated and the efficiency of our proposed solution is quantified: better filtering performances are obtained in 87.5% of the cases and on average, the relative improvement is 27.7%.

Proton radiography using a multilayer ionization chamber can potentially be used for assessing the quality of the stopping power computation in proton therapy. However, the finite proton beam profile leads to a degradation of the depth-dose curves (`blurring') measured by the range probe, which makes the estimation of the integrated proton stopping power a complex task. Existing methods aiming at determining a map of the integrated proton stopping power currently involve the use of the planning x-ray computed tomography (CT) as a priori knowledge. Consequently, such methods are very sensitive to small misalignment between the planning CT and the proton radiography acquisitions, to errors in the stopping power computation and to changes in the anatomy of the patient. In this paper, we develop an algorithm based on a sparsity assumption that estimates the integrated proton stopping power map of an anthropomorphic phantom from proton radiography data without using any prior information from the CT.

With the emergence of UHD video, reference frame buffers (FBs) inside HEVC-like encoder and decoder chips have to sustain huge bandwidth. The power consumption due to the accesses to these off-chip memories accounts for a significant share of the codec's total consumption. This paper describes a JPEG XS-based frame buffer compression solution intended to decrease the FB's bandwidth, making HEVC more suitable for use in power-aware applications. As opposed to previous works, our solution compresses large picture areas (ranging from a CTU to a frame stripe) independently and is suitable for use in several data reuse strategies. According to experimental results, this work provides significant off-chip bandwidth and memory size reduction (DRR) at the cost of a slight rate increase. In particular, at low rates and if FBC is applied at HEVC encoder only, the DRR is as high as 83.3% with a rate increase and quality drop on average.

Up to now, many existing video transmission and storage infrastructures are not able to handle UHD uncompressed video in real-time. For instance, the transmission of 4K UHD 4:2:2 10 bits 60p requires approximatively 4 times the bandwidth available on a 3-G SDI cable. To reduce the required bitrates, a low-latency lightweight compression scheme is needed. To this end, several standardization efforts such as Display Stream Compression (DSC), Advanced DSC and JPEG XS are currently undertaken. Focusing on the broadcast industry, this paper provides a comparison of existing codecs suited for this field of application. In particular, the performances of DSC, intoPIX TICO and JPEG 2000 are evaluated. Regarding JPEG2000, two flavours are used. First, the JP2K Ultra-Low Latency (ULL) used in the industry reaches a total latency of 0.5 frame by splitting each frame into 9 stripes (each of size 3840×240) that are independently encoded. To reach even lower latency and complexity, the paper defines the JP2K Ultra-Low Latency and Complexity (ULLC) which uses smaller stripes (each of size 3840×8) to have only 32 lines of latency. Regarding the experiment details, the objective metric used to assess the quality preservation is the Peak-Signal-to-Noise ratio. First, a quality assessment is realized in single and multiple generations. Then, the error robustness is evaluated by averaging over 1000 experiments the effect of inserting a one-bit error at a random position in the compressed bitstream. This experiment is meaningful since such bit-flip error occurs on SDI cable with a probability ranging from 10-9to 10-12. From the results (displayed below), it appears that the coding schemes based on Discrete Wavelet Transform (namely TICO and JPEG 2000) give better results than DSC which is mainly based on intra-frame prediction methods.

This paper studies the impact of secure watermark embedding in digital images by proposing a practical implementation of secure spread-spectrum watermarking using distortion optimization. Because strong security properties (key-security and subspace-security) can be achieved using natural watermarking (NW) since this particular embedding lets the distribution of the host and watermarked signals unchanged, we use elements of transportation theory to minimize the global distortion. Next, we apply this new modulation, called transportation NW (TNW), to design a secure watermarking scheme for grayscale images. The TNW uses a multiresolution image decomposition combined with a multiplicative embedding which is taken into account at the distribution level. We show that the distortion solely relies on the variance of the wavelet subbands used during the embedding. In order to maximize a target robustness after JPEG compression, we select different combinations of subbands offering the lowest Bit Error Rates for a target PSNR ranging from 35 to 55 dB and we propose an algorithm to select them. The use of transportation theory also provides an average PSNR gain of 3.6 dB on PSNR with respect to the previous embedding for a set of 2000 images.

Diffusion tensor imaging (DTI) is unable to represent the diffusion signal arising from multiple crossing fascicles and freely diffusing water molecules. Generative models of the diffusion signal, such as multi-fascicle models, overcome this limitation by providing a parametric representation for the signal contribution of each population of water molecules. These models are of great interest in population studies to characterize and compare the brain microstructural properties. Central to population studies is the construction of an atlas and the registration of all subjects to it. However, the appropriate definition of registration and atlasing methods for multi-fascicle models have proven challenging. This paper proposes a mathematical framework to register and analyze multi-fascicle models. Specifically, we define novel operators to achieve interpolation, smoothing and averaging of multi-fascicle models. We also define a novel similarity metric to spatially align multi-fascicle models. Our framework enables simultaneous comparisons of different microstructural properties that are confounded in conventional DTI. The framework is validated on multi-fascicle models from 24 healthy subjects and 38 patients with tuberous sclerosis complex, 10 of whom have autism. We demonstrate the use of the multi-fascicle models registration and analysis framework in a population study of autism spectrum disorder.

This paper presents a study of the embedding of Tardos binary fingerprinting codes with watermarking techniques. By taking into account the security of the embedding scheme, we present a new approach for colluding strategies which relies on the possible estimation error rate of the code symbols (denoted$\epsilon$). We derive a new attack strategy called “$\epsilon$-Worst Case Attack” and show its efficiency using the computation of achievable rates for simple decoding. Then we consider the interplay between security and robustness regarding the accusation performances of the fingerprinting scheme and show that 1) for the same accusation rate secure schemes can afford to be less robust than insecure ones, and 2) that secure schemes enable to cast the Worst Case Attack into an interleaving attack. Additionally, we use the security analysis of the watermarking scheme to derive from$\epsilon$a security attack for a fingerprinting scheme based on Tardos codes and a new scheme called stochastic spread-spectrum watermarking. We compare a removal attack against an AWGN robustness attack and we show that for the same distortion, the combination of a fingerprinting attack and a security attack easily outperform classical attacks even with a small number of observations.

In this paper we propose a new graph formulation for solving Discrete Tomography problems in the case where only a very few number of projections are available. Graph formulations are efficient to solve many different pixel labeling problems in Image Processing. However, applying graph models for Discrete Tomography problems is a very challenging task due to the high dimensionality of the data and the complexity of the constraints formulations. Due to the NP-hardness of the problem, even the computation of a local minima is still a challenge. In this paper, we propose a graph model with a polynomial number of additional edges formulating the projection consistency and an iterative algorithm to efficiently minimize the energy function. Our formulation aims to provide 3D results based on a restricted number of projections.

Image descriptors are widely used in applications such as object recognition, pattern classification and image registration. The descriptors encode the local visual content of the image to provide a compact, robust and distinctive representation of objects. If images differ in orientation, descriptors must be rotation invariant. This paper introduces a compact rotation invariant descriptor. The approach is based on the representation of the local visual content by a graph. A function living on the graph vertices is evaluated and transformed through spectral trimming. This transform is rotation invariant and reduces the dimensionality of the descriptor. The performance of the introduced descriptor is as good as the SIFT descriptor performance, while being about ten times more compact, as shown by experiments on transmission electron microscope images.

BACKGROUND: The standard approach for preprocessing spotted microarray data is to subtract the local background intensity from the spot foreground intensity, to perform a log2 transformation and to normalize the data with a global median or a lowess normalization. Although well motivated, standard approaches for background correction and for transformation have been widely criticized because they produce high variance at low intensities. Whereas various alternatives to the standard background correction methods and to log2 transformation were proposed, impacts of both successive preprocessing steps were not compared in an objective way. RESULTS: In this study, we assessed the impact of eight preprocessing methods combining four background correction methods and two transformations (the log2 and the glog), by using data from the MAQC study. The current results indicate that most preprocessing methods produce fold-change compression at low intensities. Fold-change compression was minimized using the Standard and the Edwards background correction methods coupled with a log2 transformation. The drawback of both methods is a high variance at low intensities which consequently produced poor estimations of the p-values. On the other hand, effective stabilization of the variance as well as better estimations of the p-values were observed after the glog transformation. CONCLUSION: As both fold-change magnitudes and p-values are important in the context of microarray class comparison studies, we therefore recommend to combine the Edwards correction with a hybrid transformation method that uses the log2 transformation to estimate fold-change magnitudes and the glog transformation to estimate p-values.

A B-cell epitope is a part of an antigen that is recognized by a specific antibody or B-cell receptor. Detecting the immunogenic region of the antigen is useful in numerous immunodetection and immunotherapeutics applications. The aim of this paper is to find relevant properties to discriminate the location of potential epitopes from the rest of the protein surface. The most relevant properties, identified using two evaluation approaches, are the geometric properties, followed by the conservation score and some chemical properties, such as the proportion of glycine. The selected properties are used in a patch-based epitope localization method including a Single-Layer Perceptron for regression. The output of this Single-Layer Perceptron is used to construct a probability map on the antigen surface. The predictive performances of the method are assessed by computing the AUC using cross validation on two benchmark data sets and by computing the AUC and the precision for a third independent test set.

Travel Depth, introduced by Coleman and Sharp in 2006, is a physical interpretation of molecular depth, a term frequently used to describe the shape of a molecular active site or binding site. Travel Depth can be seen as the physical distance a solvent molecule would have to travel from a point of the surface, i.e., the Solvent-Excluded Surface (SES), to its convex hull. Existing algorithms providing an estimation of the Travel Depth are based on a regular sampling of the molecule volume and the use of the Dijkstra's shortest path algorithm. Since Travel Depth is only defined on the molecular surface, this volume-based approach is characterized by a large computational complexity due to the processing of unnecessary samples lying inside or outside the molecule. In this paper, we propose a surface-based approach that restricts the processing to data defined on the SES. This algorithm significantly reduces the complexity of Travel Depth estimation and makes possible the analysis of large macromolecule surface shape description with high resolution. Experimental results show that compared to existing methods, the proposed algorithm achieves accurate estimations with considerably reduced processing times.

In this paper, we address the issues of analyzing and classifying JPEG 2000 code-streams. An original representation, called integral volume, is first proposed to compute local image features progressively from the compressed code-stream, on any spatial image area, regardless of the code-blocks borders. Then, a JPEG 2000 classifier is presented that uses integral volumes to learn an ensemble of randomized trees. Several classification tasks are performed on various JPEG 2000 image databases and results are in the same range as the ones obtained in the literature with noncompressed versions of these databases. Finally, a cascade of such classifiers is considered, in order to specifically address the image retrieval issue, i.e., bi-class problems characterized by a highly skewed distribution. An efficient way to learn and optimize such cascade is proposed. We show that staying in a JPEG 2000 framework, initially seen as a constraint to avoid heavy decoding operations, is actually an advantage as it can benefit from the multiresolution and multilayer paradigms inherently present in this compression standard. In particular, unlike other existing cascaded retrieval systems, the features used along our cascade are increasingly discriminant and lead therefore to a better tradeoff of complexity versus performance.

A 3D-mesh watermarking method is proposed in this paper. The watermarking occurs in the spherical coordinates system, where only the vertex norm of each point is modified. The embedding method is based on Spread Transform Dither Modulation (STDM). A perceptual model based on the curvature and the roughness of the 3D mesh is defined and is used to modulate the STDM method. Our results show that the proposed method has high tolerance to noise and is robust against mesh smoothing.

In this paper, we develop an adjustable Fuzzy Extractor using the Physical Unclonable Functions (PUF) obtained by a common laser engraving method to sign physical objects. In particular, a string (or helper data) is generated by XORing a binary reduction of the PUF observation with the encoding of a randomly generated key, or identifier. Since the binary reduction (or hash) relies on keeping the sign of few random projections of the observation, a measure concentration property bounds, with a controlled accuracy, the distance between two different hashes in function of this of the original images. The error correcting code used to encode the identifier stabilizes therefore both the observation noise and the hashing distortion. In a verification stage, reobserving the PUF with the helper data in hand allows one to authenticate the object if the identifier can be exactly recovered. We conclude this work by calibrating and challenging the system on a database of laser-written PUFs, balancing helper data size, that is, hashing dimensions, and system security.

This article is a theoretical study on binary Tardos' fingerprinting codes embedded using watermarking schemes. Our approach is derived from and encompasses both security and robustness constraints. We assume here that the coalition has estimated the symbols of the fingerprinting code by the way of a security attack, the quality of the estimation relying on the security of the watermarking scheme. Taking into account the fact that the coalition can perform estimation errors, we update the Worst Case Attack, which minimises the mutual information between the sequence of one colluder and the pirated sequence forged by the coalition. After comparing the achievable rates of the previous and proposed Worst Case Attack according to the estimation error, we conclude this analysis by comparing the robustness of no-secure embedding schemes versus secure ones. We show that, for low probabilities of error during the decoding stage (e.g. highly robust watermarking schemes), security enables to increase the achievable rate of the fingerprinting scheme.

This paper proposes a new adaptive scheme for stereo image coding . The original left and right images are jointly coded in order to exploit the high correlation between the two images for a lossless coding. Our goal is to design an efficient transform that reduces the existing redundancy in the stereo pair. This approach is inspired by the lifting scheme (LS). First, illumination change compensation has been made. Secondly, the prediction filtering is locally adapted and is based on local horizontal and vertical gradient information. In the sense that small gradient evokes predictability in the same direction. The coefficients of the predictor are then optimized. Experimental results show improvement in term of performance and complexity compared to other proposed methods.

In this paper, we propose a blind and robust watermarking method for 3D polygonal meshes by minimising the mean square error between the original mesh and the watermarked mesh under several constraints. We have formulated the problem of assigning distortions to points in a 3D mesh to a quadratic programming problem, so it can be solved reliably and efficiently. Comparing with similar approaches in [1], experiments indicate the advantages of our method in resisting Gaussian noise.

We present a multi-view three dimensional intelligent surveillance system. We use a multi-agent framework to identify the behaviors of individuals in the scene. Detection and interpretation are performed completely in 3D space. A moving train coach is monitored by eight fish-eye cameras. Segmentation masks extracted from the undistorted images are fed to a distributed 3D reconstruction algorithm producing an octree-based description of the volume at each frame. Voxel-based algorithms extract connected-regions and their descriptions from consecutive models. The set of regions is mapped to a set of agents. We achieve dynamically consistent high-level interpretations by combining probabilistic models of human behaviors and intelligent reasoning.

This research aims to propose a multi-modal fusion framework for high-level data fusion between two or more modalities. It takes as input low level features extracted from different system devices, analyses and identifies intrinsic meanings in these data. Extracted meanings are mutually compared to identify complementarities, ambiguities and inconsistencies to better understand the user intention when interacting with the system. The whole fusion life cycle will be described and evaluated in an office environment scenario, where two co-workers interact by voice and movements, which might show their intentions. The fusion in this case is focusing on combining modalities for capturing a context to enhance the user experience.

BACKGROUND: The structural genomics centers provide hundreds of protein structures of unknown function. Therefore, developing methods enabling the determination of a protein function automatically is imperative. The determination of a protein function can be achieved by studying the network of its physical interactions. In this context, identifying a potential binding site between proteins is of primary interest. In the literature, methods for predicting a potential binding site location generally are based on classification tools. The aim of this paper is to show that regression tools are more efficient than classification tools for patches based binding site predictors. For this purpose, we developed a patches based binding site localization method usable with either regression or classification tools. RESULTS: We compared predictive performances of regression tools with performances of machine learning classifiers. Using leave-one-out cross-validation, we showed that regression tools provide better predictions than classification ones. Among regression tools, Multilayer Perceptron ranked highest in the quality of predictions. We compared also the predictive performance of our patches based method using Multilayer Perceptron with the performance of three other methods usable through a web server. Our method performed similarly to the other methods. CONCLUSION: Regression is more efficient than classification when applied to our binding site localization method. When it is possible, using regression instead of classification for other existing binding site predictors will probably improve results. Furthermore, the method presented in this work is flexible because the size of the predicted binding site is adjustable. This adaptability is useful when either false positive or negative rates have to be limited.

In video surveillance applications, pre-stored images are likely to be accessed remotely and interactively upon user request. In such a context, the JPEG 2000 still image compression format is attractive because it supports flexible and progressive access to each individual image of the pre-stored content, in terms of spatial location, quality level, as well as resolution. However, when the client wants to play consecutive frames of the video sequence, the purely INTRA nature of JPEG 2000 dramatically penalizes the transmission efficiency. To mitigate this drawback, conditional replenishment mechanisms are envisioned. They convey arbitrary spatio-temporal segments of the initial video sequence directly through sporadic and rate-distortion (RD) optimized refresh of JPEG 2000 packets. Hence, they preserve JPEG 2000 compliance, while saving transmission resources. The replenishment algorithms proposed in this paper are original in two main aspects. First, they exploit the specificities of the JPEG 2000 codestream structure to balance the accuracy (in terms of bit-planes) of the replenishment across image subbands in a (RD)-optimal way. Second, they take into account the still background nature of video surveillance content by maintaining two reference images at the receiver. One reference is the last reconstructed frame, as proposed in the original replenishment framework. The other is a dynamically computed estimate of the scene background, which helps to recover the background after a moving object has left the scene. As an additional contribution, we demonstrate that the embedded nature of the JPEG 2000 codestream easily supports prioritization of semantically relevant regions of interest while browsing video content. An interesting aspect of this JPEG 2000-based prioritization is that it can be regulatedaposteriori, after the codestream generation, based on the interest expressed by the user at browsing time. Simulation results demonstrate the efficiency and flexibility of the approach compared to INTER-based solutions.

In order to achieve both fast tracking and accurate object extraction, we present in this paper an original real-time active contour method, incorporating different feature maps into a common and homogeneous framework, defined by the multi-feature vector flow (MFVF). The MFVF active contour approach does not require any target prior model, and enables precise tracking of mobile deformable objects. The use of the MFVF, resulting from multiple selected features, brings robustness into the system towards complex situations, while our computationally efficient implementation of the MFVF scheme reaches the required speed range for tracking process. The proposed method has been successfully tested on real-world video sequences.

During the past years, many different registration methods have emerged. These usually lead to different results and one faces the problem of choosing the most appropriate. To mitigate this issue, several algorithms have been developed to include prior knowledge in order to force the registration to have a physical behavior. The use of most of these methods in daily practice is cumbersome because of the need to create data-specific meshes. This paper introduces a local convolutive filter in order to iteratively make the result of any registration method converge towards a linear elastic solution. This method was tested and validated on virtual data and on a real silicone rubber cube.

We demonstrate a multimodal high-level data fusion tooling integrated on a platform aimed at the rapid development and prototyping of multimodal applications through user-centered design. The platform embeds a set of pure and combined modalities as reusable components and generic mechanisms for combining modalities with a rich support for multimodal fusion in order to improve the human-computer interaction.

This paper analyzes the movements of the human body limbs (hands, feet and head) and center of gravity in order to detect and analyze simple actions such as walking and running. We propose a novel vision of the human body, by considering the limbs as cooperative agents that form a hierarchy of cooperative teams: the whole body. The movements are analyzed at individual level and at team level using a modular hierarchical structure. Knowledge of the high-level team actions (such as ldquowalkingrdquo) improves the pertinence of our predictions on the low-level individual actions (foot is moving back and forth) and allows us to compensate for missing or noisy data produced by the feature extraction system. In terms of group behavior recognition, we propose a novel framework for online probabilistic plan recognition in cooperative multiagent systems: the Multiagent Hidden Markov mEmory Model (M-AHMEM), which is a dynamic Bayesian network. Experiments on an existing video database using different models of the human body show the feasibility of the approach.

This paper tackles the challenge of interactively retrieving visual scenes within surveillance sequences acquired with fixed camera. Contrarily to today's solutions, we assume that no a-priori knowledge is available so that the system must progressively learn the target scenes thanks to interactive labelling of a few frames by the user. The proposed method is based on very low-cost features extraction and integrates relevance feedback, multiple-instance SVM classification and active learning. Each of these 3 steps runs iteratively over the session, and takes advantage of the progressively increasing training set. Repeatable experiments on both simulated and real data demonstrate the efficiency of the approach and show how it allows reaching high retrieval performances.

The image compression standard JPEG 2000 offers a high compression efficiency as well as a great flexibility in the way it accesses the content in terms of spatial location, quality level, and resolution. This paper explores how transmission systems conveying video surveillance sequences can benefit from this flexibility. Rather than transmitting each frame independently as it is generally done in the literature for JPEG 2000 based systems, we adopt a conditional replenishment scheme to exploit the temporal correlation of the video sequence. As a first contribution, we propose a rate-distortion optimal strategy to select the most profitable packets to transmit. As a second contribution, we provide the client with two references, the previous reconstructed frame and an estimation of the current scene background, which improves the transmission system performances.

Segmentation and tracking methods have been widely explore. However, they are often computationally heavy or require constraining assumptions. We present in this paper a new system for real-time simultaneous segmentation and tracking, without any hypothesis on target appearance, image background or camera properties. The proposed approach (SUGVPB) is an active contour modeled with B-splines and which evolution process is using a speeded up gradient vector flow, characterized by a faster computation of the edge diffusion process. The synergy of these two powerful components enables precise, robust and real-time tracking of complete non-rigid mobile objects. Our method has been validated on synthetic as well as natural video sequences.

In this paper we compare three non-rigid registration methods for atlas-based segmentation: B-splines, morphons and a combination of morphons and demons. To assess the quality of each method, we use a data set of four patients, containing for each patient the computed tomography (CT) image and a manual segmentation of the organs at risk performed by an expert of the head and neck anatomy. Non-rigid registration algorithms have been used to match the patient and atlas images. Each deformation field, resulting from the non-rigid deformation, have been applied on the masks corresponding to segmented regions in the atlas. The atlas based segmented masks have been compared to manual segmentations performed by the expert. The results show that the combined method (morphons + demons) achieves the best performances on this dataset resulting in an average improvement of 6% with respect to morphons and 18% with respect to B-spline.

We address the problem of distributed source coding of binary sources with side information at the decoder. We propose a compression scheme using irregular turbo codes. We optimize them by maximizing the distance between the two curves of their EXIT charts for given statistics and a compression rate close to the Slepian-Wolf limit in order to increase the robustness of the scheme. This optimized irregular code enables us to achieve compression rates superior to the ones achieved by previously proposed turbo schemes without increasing the decoding complexity.

State-of-the-art blind and robust 3D watermarking schemes already withstand combinations of a wide variety of attacks (e.g. noise addition, simplification, smoothing, ...) except cropping. This attack is however very common and should be dealt with in a copyright protection framework. In this paper, we propose a technique which enables to extend the robustness of such schemes to cropping. Our algorithm proceeds by the automatic detection of robust shape feature points which are then used for the embedding of a watermark in a local neighborhood. We show that robustness against cropping and other common attacks is achieved provided that at least one feature point as well as its corresponding local neighborhood are retrieved.

This study concerns the modeling and the design of a monostatic heterodyne pulsed LIDAR. The heart of the system is constituted of a 1.55 mum able to produce high pulse repetition frequency. The aim of this work is to assess its efficiency to perform accurate wind speed measurements in the low atmospheric boundary layer, from a 2-D scanning pattern, in the presence of refractive turbulence. A complete LIDAR numerical simulation technique has been developed. Its main originality is the integration of both optical and fluid dynamics numerical methods to take into account the signal coherence loss due to refractive turbulence and speckle effect as well as the fine structures of the wind field. The wind speed profiles along each line-of-sight are retrieved from the return signal using a low-order autoregressive model. An adequate averaging model is then used estimate horizontal components of the wind speed for altitudes up to 150 m.

A new spectral model of the return signal from a LIDAR Doppler wake vortex detector is proposed. It has been experimentally discovered during ground-based and flight test campaigns but suffered a lack of theoretical evidence. Using high resolution fluid simulations of wake vortices, we highlight the physical meaning of this model. Comparisons with the traditional single Gaussian model show the superiority of this new approach is consistent with previous experimental results.

MERIT is a research-oriented Joint European Master on Information and Communication Technologies funded by the Erasmus Mundus program of the European Community. Its well funded scholarship scheme offers an outstanding opportunity for students, professors and researchers from non-European countries to participate in education programs (e.g remote sensing) in Europe.

This paper presents a knowledge-based method of early-stage\nhigh level multimodal fusion of data obtained from speech\ninput and visual scene. The ultimate goal is to develop a\nhuman-computer multimodal interface to assist elderly people\nliving alone at home to perform their daily activities, and to\nsupport their active ageing and social cohesion. Crucial for\nmultimodal high level fusion and successful communication\nis the provision of extensive semantics and contextual\ninformation from spoken language understanding. To address\nthis we propose to extract natural language semantic\nrepresentations and map them onto the restricted domain\nontology. This information is then processed for multimodal\nreference resolution together with visual scene input. To\nmake our approach flexible and widely applicable, a priori\nsituational knowledge, modalities and the fusion process are\nmodelled in the ontology expressing the domain constraints.\nHere, we illustrate ontology-based multimodal fusion on an\nexample scenario combining speech and visual scene\nanalysis.

Our study considers the development of a reliable tracker for non-rigid objects evolving on cluttered background in crowded scenes captured by moving cameras. For this purpose, we propose an original method that combines two approaches, respectively based on parametric active contours (PAC) and on point distribution model (PDM). The PAC tracker relies on an effective and effcient implementation of contour convergence mechanism to bring a smooth contour to the edges of the target in real-time. The PDM approach collects feature points in the region delineated by the PAC tracker to build and update a model of the target in term of a feature point distribution. Formally, when a novel frame is considered, its feature points are matched with the PDM model. The matching information is used to initialize the novel PAC, whose convergence identify the points that are relevant to update the PDM for the next frame. Hence, the two approaches complement each others. The a priori information provided by the PDM makes the system robust towards occlusions, while the deformation of the PAC increases its robustness towards target appearance changes. Simulations on real-word video sequences demonstrate the performance of our approach.

This paper considers the issues of scheduling and caching JPEG2000 data in client/server interactive browsing applications, under memory and channel bandwidth constraints. It analyzes how the conveyed data have to be selected at the server and managed within the client cache so as to maximize the reactivity of the browsing application. Formally, to render the dynamic nature of the browsing session, we assume the existence of a reaction model that defines when the user launches a novel command as a function of the image quality displayed at the client. As a main outcome, our work demonstrates that, due to the latency inherent to client/server exchanges, a priori expectation about future navigation commands may help to improve the overall reactivity of the system. In our study, the browsing session is defined by the evolution of a rectangular window of interest (WoI) along the time. At any given time, the WoI defines the position and the resolution of the image data to display at the client. The expectation about future navigation commands is then formalized based on a stochastic navigation model, which defines the probability that a given WoI is requested next, knowing previous WoI requests. Based on that knowledge, several scheduling scenarios are considered. The first scenario is conventional and transmits all the data corresponding to the current WoI before prefetching the most promising data outside the current WoI. Alternative scenarios are then proposed to anticipate prefetching, by scheduling data expected to be requested in the future before all the current WoI data have been sent out. Our results demonstrate that, for predictable navigation commands, anticipated prefetching improves the overall reactivity of the system by up to 30% compared to the conventional scheduling approach. They also reveal that an accurate knowledge of the reaction model is not required to get these significant improvements.

This paper presents a novel technique for 2D human motion capture using a single non calibrated camera. The user's five extremities (head, hands and feet) are extracted, labeled and tracked after silhouette segmentation. As they are the minimal number of points that can be used in order to enable whole body gestural interaction, we henceforth refer to these features as crucial points. The crucial point candidates are defined as the local maxima of the geodesic distance with respect to the center of gravity of the actor region that lie on the silhouette boundary. In order to disambiguate the selected crucial points into head, left and right foot and left and right hand classes, we propose a Bayesian framework that combines a MAP approach weighted by a prior model and the intensities of the tracked crucial points. Due to its low computational complexity, the system can run at real-time paces on standard personal computers, with an average error rate range between 2% and 7% in realistic situations, depending on the context and segmentation quality.

Lack of temporal coherence in video segmentation algorithms often leads to flickering or to discrepancies on the segmentation mask boundaries. Active contour video segmentation algorithms can lead to very smooth segmentation masks when they are defined in three dimensions (two spatial and one temporal). In this paper, we use an active contour method to attract an active surface toward a non-smoothed segmentation mask boundary. This active surface produces a new segmentation mask which is smoother than the first one. To achieve this task, we introduce a new region-based term for active contour segmentation in the variational framework. This term attracts the evolving curve to a reference contour. The energy criterion and the evolution equation are defined in n-dimension and we investigate the particular case of two regions. The use of active surfaces to smooth video segmentation masks appears to be a very powerful tool for video postprocessing. The resulting masks are smoother than the original ones, the discrepancies of the segmentation masks are removed and the flickering on the boundary of the segmentation masks are considerably reduced

This paper considers the remote interactive browsing of large JPEG2000 images. In contrast with previous contributions, we focus on the dynamic nature of the system. Practically, we study the conditions under which a priori knowledge about the user behavior may help to improve the browsing system reactivity, when combined with appropriate pre-fetching mechanisms. In particular, our simulations show that, due to the latency inherent to client/server exchanges, a benefit can be drawn by scheduling future window of interest (WoI) data before all current WoI data have been sent to the client. They also reveal that an accurate knowledge of the user behavior is not necessary to get important improvements over conventional scheduling approaches.

This paper analyzes the movements of the human body limbs (hands, feet and head) and center of gravity in order to detect simple actions such as walking, jumping and displacing an object. Correlation between these points is guaranteed by considering them as cooperative agents forming a cooperative team : the whole body. The movements are analyzed at individual level and at team level using a hierarchical structure. We make use of a novel framework for online probabilistic plan recognition in cooperative multiagent systems (the multiagent abstract hidden Markov memory model, M-AHMEM) to modelize the human body and detect the actions performed. Knowledge of the high-level team actions (such as "walking") improves the pertinence of our predictions on the low-level individual actions (hand is moving back and forth) and allows us to compensate for missing or erroneous data produced by the feature extraction system. Experiments on real video sequences show the feasibility of the approach.

This paper describes a simple and "pre-cortical" visual attention model, which does not take image directions into account. We compute rarity-based saliency maps and then we describe the relation between texture and visual attention. Finally we decompose the image into several textures with different regularities. Our purpose is to compress textures into images using small repeating patterns.

<?Pub _touchup WordspNominal="1.6" WordspMinimum="1.3" WordspMaximum="2.9"?>The image compression standard JPEG 2000 proposes a large set of features that is useful for today's multimedia applications. Unfortunately, it is much more complex than older standards. Real-time applications, such as digital cinema, require a specific, secure, and scalable hardware implementation. In this paper, a decoding scheme is proposed with two main characteristics. First, the complete scheme takes place in a field-programmable gate array without accessing any external memory, allowing integration in a <?Pub /_touchup?>secured system. Second, a customizable level of parallelization allows to satisfy a broad range of constraints, depending on the signal resolution. The resulting architecture is therefore ready to meet upcoming digital cinema specifications.

In watermarking, the exploitation of synchronization marks embedded within the cover signal has been widely used to provide robustness against geometrical distortions. However, in most schemes, synchronization marks are easily detectable which makes them vulnerable to malicious removal or tampering. Relying on embedded signals which exhibit strong reference features to recover synchronization can therefore lead to security flaws in the watermarking system. In this paper, we present and discuss the synchronization recovery process through detection of the periodicity of the embedded signal. We then propose a scheme to hide synchronization marks through the modulation of the embedded signal by a content-based pseudo-random signal. We study the robustness of this pseudo-random masking signal and illustrate the efficiency of this approach on watermark patterns with a periodic structure.

This article presents a modular architecture for multicamera tracking in the context of sports broadcasting. For each video stream, a geometrical module continuously performs the image-to-model homography estimation. A local-feature based tracking module tracks the players in each view. A supervisor module collects, associates and fuses the data provided by the tracking modules. The originality of the proposed system is three-fold. First, it allows to localize the targets on the ground with rotating and zooming cameras; second, it does not use background modeling techniques; and third, the local tracking can cope with severe occlusions. We present experimental results on raw TV-camera footage of a soccer game.

Recent works have shown that the particle filter using color as observation feature is a powerful technique for tracking deformable objects in image sequences with complex backgrounds. This paper presents a hybrid valued sequential state estimation algorithm, and its particle filter-based solution, that extends the standard color particle filter in two ways. Firstly, track initialization is embedded in the particle filter without relying on an external target detection algorithm. Secondly, the algorithm is able to track multiple objects sharing the same color description. We evaluate the performance of the proposed filter on various real-world video sequences with appearing and disappearing targets.

Our research deals with a fully automatic and fast localization of possible tumoral areas on computed tomography scanner (CT scan) images. The aim of this method is not to segment tumors but only to highlight the areas where tumor has the greater probability to be located. To achieve this task, we use the bilateral symmetry of the human body and the asymmetry introduced by the presence of tumors. Our work was initially dedicated to the head and neck area but it should work well for any other body part and even better for more symmetric areas like the brain.

This paper presents a novel technique for 2D human motion estimation using a single non calibrated camera. The user's five crucial human features (head, hands and feet) are extracted, labeled and tracked, after silhouette segmentation. The crucial points candidates are defined as the local maxima of the geodesic distance with respect to the center of gravity of the actor region (silhouette) following the silhouette boundary. Selected crucial points are then classified as head, hands or feet using a probabilistic approach weighted by a prior human model. The system can run at 50 Hz paces on standard personal computers.

This paper introduces a new metric to gather a large collection of segmented images into a same reference system. Different positions for each subject (pose parameters) as well as high energy shape variations need to be compensated before performing statistical analysis (like principal components analysis) on the database. The atlas is obtained as the hidden variable of an expectation-maximization (EM), looking for the right signal intensity at each voxel in the collection of subjects. Each subject is aligned on the current probabilistic atlas by maximizing mutual information. A fast stochastic optimization algorithm is used for estimating pose and scale parameters and a variational approach have been designed to estimate non-rigid transformations. We illustrate the effectiveness of this method for the alignment of 31 brain segmented in 4 labels: background, white and gray matter and ventricles. Our approach has the advantage of keeping a reasonably low complexity even for large databases.

In this paper we address the problem of video segmentation. We propose a new method which combines an active contour and a watershed presegmentation. We first apply active contours to track moving objects with a basic motion criterion. Then, we eliminate the variation due to motion on the object boundaries by attracting the curve to the nearest boundaries of a watershed oversegmentation. We constrain the curve length to smooth the propagation driven by the distance to the watershed boundaries. The proposed method improves active contour segmentation and gives smoother results than watershed segmentation

We present a new blind robust watermarking scheme for 3D meshes. Feature points are used to build a partition of the mesh shape that resists to common 3D watermarking attacks. These points are automatically selected through a multi-scale estimation of the curvature tensor field. The automatic capture of robust feature points and its use for blind detection in a robust watermarking scheme are the contribution of this paper. Our watermarking scheme proceeds by first partitioning the mesh shape using a geodesic Delaunay triangulation of the detected feature points. Each of these geodesic triangle patches is then parameterized and remeshed by a subdivision strategy to obtain a robust base meshing. Then remeshed patches are watermarked in the spectral domain and original mesh points are finally projected on the corresponding watermarked patches. This strategy shows good preliminary results as it resists to affine transforms, white noise addition, smoothing, crop and sampling changes such as decimation, subdivision or remeshing.

Robust image hashing defines a feature vector that characterizes the image, independently of non-significant distortions of its content. As a consequence, the comparison between robust image hash vectors is able to indicate whether the corresponding images are equivalent or not, independently of visually non-significant distortions due for example to compression or re-sampling. We define a robust image hash based on radial projections of the image pixels. Specifically, our proposed radial hASH (RASH) considers moments of different orders to describe the luminance pdf of the pixels encountered on a set of lines articulated around the center of the image. In short, each RASH component is defined based on the moment of the pixels belonging to a specific line. Our paper provides a careful analysis of the robustness and discriminating capabilities of the RASH vectors computed based on different moment orders. As a first contribution, it demonstrates that the second order moment, i.e. the variance of the pixels on a line, allows for optimal trade-offs between the robustness and the discriminating capabilities of the resulting RASH vector. As a second contribution, extensive simulations prove that a decision engine based on RASH vectors cross-correlations is able to successfully identify pairs of equivalent or distinct images. Bottom line, the RASH assets are a low computational complexity, a strong robustness to both filtering and geometrical distortions, and a risk of collision that is estimated to less than 10 per million of images.

We propose a novel method, guided slice marching to segment opacified vessels tree in 3D image sets (CT scans). It combines a front propagation technique, slice marching, and an anatomical model to guide the propagation for solving the particular case of touching vessels. The formulation of this method, which is based on interface evolution theory, enables easy integration of an a priori model of knowledge of vessels topology to handle the case of touching vessels, where image-based method systematically fails. The a priori knowledge is expressed as parametric curves that model vessels centerline. That information is injected in the fast marching method through the speed of propagation, setting it to zero at missing vessels boundaries. The model is intended to be re-used across patients, and must therefore be registered with the image.

This paper introduces a new registration method estimating the displacement field of bodies which deformations are constrained by an articulated rigid body. We propose an articulated transformation model embedded in a general registration scheme. A fast stochastic gradient descent optimization strategy suitable for noisy cost functions has been chosen to maximize the mutual information metric. Once registered, we propose to propagate the deformation by a linear elastic model through the use of a tetrahedral mesh. We demonstrate this method on 3D CT of neck images where bony structures between different patient images, as vertebrae, may rigidly move while other tissues may deform.

Remote access to large scale images arouses a growing interest in fields such as medical imagery or remote sensing. This raises the need for algorithms guaranteeing navigation smoothness while minimizing the network resources used. In this paper, we present a model taking advantage of the JPEG 2000 scalability combined with a prefetching policy. The model uses the last user action to efficiently manage the cache and to prefetch the most probable data to be used next. Three different network configurations are considered. In each case, comparison with two more classic policies shows the improvement brought by our approach.

Compared to other hospital units, the emergency department presents some distinguishing characteristics of its own. Emergency health-care delivery is a collaborative process involving the contribution of several individuals who accomplish their tasks while working autonomously under pressure and sometimes with limited resources. Effective computerization of the emergency department information system presents a real challenge due to the complexity of the scenario. Current computerized support suffers from several problems, including inadequate data models, clumsy user interfaces, and poor integration with other clinical information systems. To tackle such complexity, we propose an approach combining three points of view, namely the transactions (in and out of the department), the (mono and multi) user interfaces and data management. Unlike current systems, we pay particular attention to the user-friendliness and versatility of our system. This means that intuitive user interfaces have been conceived and specific software modeling methodologies have been applied to provide our system with the flexibility and adaptability necessary for the individual and group coordinated tasks. Our approach has been implemented by prototyping a web-based, multiplatform, multiuser, and versatile clinical information system built upon multitier software architecture, using the Java programming language.

This paper presents a hardware implementation of a decoder for Digital Cinema images. This decoder enables us to deal with image size of 2K with 24 frames per second and 36 bits per pixels. It is the first implementation known nowadays that perfectly fits in one single Virtex-II® FPGA and includes AES decryption, JPEG 2000 decompression and fingerprinting blocks. This hardware offers therefore high-quality image processing as well as robust security.

The paper introduces a correlation-based method for the three-dimensional reconstruction of scenes from a multi-camera imaging system. Our technique is to cast the matching and reconstruction problems into a single 3D process that uses perspective distortions to retrieve directly, in a dense fashion, the 3D planes locally tangent to the scene. Avoiding image assumptions like fronto-parallelism, block shapes, perspective distortion models or camera models, our method only needs a local planarity hypothesis of the scene.

Our research deals with a fully automatic and fast visualization of possible tumoral areas on CT Scan images. To achieve this task, we use the bilateral symmetry of the human body and the asymmetry introduced by the presence of tumors. Our work was initially dedicated to the head and neck area but it should work well for any other body part and even better for more symmetric areas like the brain.

We discuss the issues related to image watermarking benchmarking and scenarios based on digital rights management requirements. We show that improvements are needed in image quality evaluation, especially related to image geometrical deformation assessments, in risk evaluation related to specific delivery scenarios and in multidimensional criteria evaluation. Efficient benchmarking is still an open issue and we suggest the use of open-source Web-based evaluation systems for collective progress in this domain.

In order to select error-correcting codes for various applications, their performances have to be determined. However, when targeting error-rate computations for block error-correcting codes, many required results are missing in the coding literature. Even in the simple case of binary codes and bounded-distance decoding, classical texts do not provide a bit-error rate (BER) expression taking into account both decoding errors and failures. In the case of nonbinary codes used to protect binary symbols, such as Reed-Solomon codes in many applications, there is no available result making realistic channel assumptions in order to derive BERs. Finally, for the more complex case of complete decoding, only some bounds are available, such as the union one. This paper presents new approximations of error rates for block error-correcting codes as a function of the channel BER (crossover probability). We extend an existing approximation in order to consider not only bounded-distance decoding, but also complete "nearest-neighbor" decoding. We also develop approximations able to deal with nonbinary codes. Combined with state-of-the-art approximations, these new results enable the computation of bit-, symbol-, and word-error rates in various decoding situations. They can consider separately errors related to erroneously decoded words and decoding failures, and they provide accurate estimates of error rates. As they do not require detailed information about the structure of codes, they are general enough to be used in simple comparisons between different codes, avoiding the need for simulations.

This paper proposes a new method for the design of lifting filters to compute a multidimensional nonseparable wavelet transform. Our approach is stated in the general case, and is illustrated for the 2-D separable and for the quincunx images. Results are shown for the JPEG2000 database and for satellite images acquired on a quincunx sampling grid. The design of efficient quincunx filters is a difficult challenge which has already been addressed for specific cases. Our approach enables the design of less expensive filters adapted to the signal statistics to enhance the compression efficiency in a more general case. It is based on a two-step lifting scheme and joins the lifting theory with Wiener's optimization. The prediction step is designed in order to minimize the variance of the signal, and the update step is designed in order to minimize a reconstruction error. Application for lossy compression shows the performances of the method.

Watermarking is largely used for copyright protection and fast search of images in databases. Another method for securely identifying images is to use hash functions. Digital signature standard, used in cryptosystem to dispute authentication documents, is based on hash functions. A digital signature is a bit stream dependent on key and content of document. For each document, the digital signature algorithm provides a unique output bit stream. In order to be efficient in images, the digital signature should be different if and only if the image content, and not the input bit stream, is different. Our new method is a one-way function for images. Using the radon transform and principal component analysis, we extract characteristics robust against geometrical transformation (rotation and scaling) and image processing attacks (compression, filtering, blurring).

Spectral decomposition of mesh geometry has first been introduced by Taubin for geometry processing purposes. It has been extended to address transmission issues by Kami & Gotsman. Such decompositions give rise to pseudofrequential information of the geometry defined over the mesh connectivity. For large meshes a piecewise decomposition has to be applied in order to restrict the complexity of the transform. In this paper, we propose to introduce overlap for its spectral representation. We show visual gains obtained in compression and progressive transmission of mesh geometry.

The need for reversible or lossless watermarking methods has been highlighted in the literature to associate subliminal management information with losslessly processed media and to enable their authentication. The paper first analyzes the specificity and the application scope of lossless watermarking methods. It explains why early attempts to achieve reversibility are not satisfactory. They are restricted to well-chosen images, strictly lossless context and/or suffer from annoying visual artifacts. Circular interpretation of bijective transformations is proposed to implement a method that fulfills all quality and functionality requirements of lossless watermarking. Results of several bench tests demonstrate the validity of the approach.

This paper deals with some detection issues of watermark signals. We propose an easy way to implement an informed watermarking embedder whatever the detection function. This method shows that a linear detection function is not suitable for side information. This is the reason why we build a family of nonlinear functions named JANIS. Used with a side-informed embedder, its performance is much better than the classical spread spectrum method.

Digital watermarking consists of hiding subliminal information into digital media content, also called host data. It can be the basis of many applications, including security and media asset management. In this paper, we focus on the imperceptibility requirement for image watermarking. We present the main features of the human visual system (HVS) to be translated into watermarking technology. This paper highlights the need for dedicated inputs from the human vision community. The human visual system (HVS) is very complex and able to deal with a huge amount of information. Roughly speaking, it is composed of a receiver with a pre-processing stage, the eye and the retina, a transmission channel, the optic nerve, and a processing engine, the visual cortex. Mainly because of our lack of knowledge about brain behavior, i.e. about the way a stimulus is processed through its huge neural network, the large effort to understand and model the HVS behavior has partly remained fruitless. The aim of this paper is not to provide a thorough description of the HVS. For complete HVS models and more specific details, the reader is referred to existing literature. Here, we only try to understand, in a synthetic way and from an engineering perspective, the HVS features on which the designer of a watermarking algorithm can rely, i.e. its sensitivity and masking capabilities.

Many new multimedia group applications rely on the selection of a rendezvous point on a network at which to hold the meeting. Each participant sends his multimedia data stream through forward paths to this central node, which then broadcasts a virtual videoconference scene through a multicast backward path towards all the participants. We model the optimization of these paths as the shortest path problem and the optimal Steiner tree problem. We propose a heuristic to select a rendezvous point minimizing a weighted sum of forward and backward paths. Computational results are presented showing the performance of our approach.

Digital watermarking consists of hiding subliminal information into digital media content, also called host data. It can be the basis of many applications, including security and media asset management. In this paper we focus on the imperceptibility requirement for image watermarking. We first provide a functional inventory of image watermarking applications and emphasize the dependency between the application purpose and its need for invisibility. Then, we present a global framework common to most existing watermarking systems. It illustrates the methodology followed to translate human vision research into watermarking technology. It suggests future prospects and highlights the need for dedicated inputs from the human vision community.

This paper presents a new approach for watermarking of digital images providing robustness to geometrical distortions. The weaknesses of classical watermarking methods to geometrical distortions are outlined first. Geometrical distortions can be decomposed into two classes: global transformations such as rotations and translations and local transformations such as the StirMark attack. An overview of existing self-synchronizing schemes is then presented. Theses schemes can use periodical properties of the mark, invariant properties of transforms, template insertion, or information provided by the original image to counter geometrical distortions. Thereafter, a new class of watermarking schemes using the image content is presented. We propose an embedding and detection scheme where the mark is bound with a content descriptor defined by salient points. Three different types of feature points are studied and their robustness to geometrical transformations is evaluated to develop an enhanced detector. The embedding of the signature is done by extracting feature points of the image and performing a Delaunay tessellation on the set of points. The mark is embedded using a classical additive scheme inside each triangle of the tessellation. The detection is done using correlation properties on the different triangles. The performance of the presented scheme is evaluated after JPEG compression, geometrical attack and transformations. Results show that the fact that the scheme is robust to these different manipulations. Finally, in our concluding remarks, we analyze the different perspectives of such content-based watermarking scheme.

In this paper, we investigate the performances of Gaussian modeling and linear prediction tools for error detection and concealment in the transmission of still images. We consider the transmission of subband encoded images through two types of channels. We model the residual correlation between subband coefficients by considering them as jointly Gaussian variables. The first transmission medium considered is a packet-oriented channel, where some packets are lost during transmission. The problem is to estimate the values of missing coefficients. In this case, particular care must be taken while evaluating correlation matrices from incomplete data. The other system considered is based on a discrete memoryless noisy channel affecting the data being transmitted. The challenge is here first to determine the locations of the errors--which is done through hypotheses tests--and then to replace them by estimates based on their neighbors. The reconstruction via linear prediction is shown to give better results than median filtering based reconstruction. Error detection through this Gaussian model also shows promising results, in particular when channel statistics are taken into account in a joint source-channel decoding framework.

This paper presents a watermarking scheme for image or video objects. The watermarking of video objects implies different constraints from raw watermarking methods. The mark has to be detected after object manipulations such as rotations, translations and VOL modifications. To achieve these requirements, the embedding scheme exploits the shape of the object: a random sequence is transformed to fit the scale and the orientation of the object. The detection of the mark is performed applying an inverse transform and calculating a correlation between the random sequence and the warped object. Our results illustrate the fact that the presented method is robust to object manipulation.

This paper presents a watermarking scheme for image or video objects. The watcamarking of video objects implies different constraints from raw watermarking methods. The mark has to be detected after object manipulations such as rotations, translations arid VOL modifications. To achieve these requirements, the embedding scheme exploits the shape of the object: a random sequence is transformed to fit the scale and the orientation of the object. The detection of the mark is performed applying an inverse transform and calculating a correlation between the random sequence and the warped object. Our results illustrate the fact that the presented method is robust to object manipulation.

This paper proposes some modifications to reference models of the human visual system for the benefit of a best performing fingerprinting scheme inside wavelet decompressors. Two new algorithms have been developed, based on more accurate modeling for luminance masking and on experimental tests for texture masking. In order to manifest the improvements, the insertion gain of the masks relative to non-perceptual watermarking schemes is calculated and compared to the one obtained by the reference technique. Also, the masks' energy distribution is checked to validate them as weighting functions of the watermark.

This paper introduces the problem of designing a quincunx lifting scheme well-adapted to lossy compression applications. The main goal is to design quincunx filters adapted to the statistical properties of the input signal in order to perform a better signal reconstruction after deterioration due to a lossy coding. To perform such a filtering, the basic idea is to combine optimization methods and a lifting scheme. The latter is an attractive tool to perform wavelet transforms.

We are present an algorithm for watermarking digital images with the aim of obtaining a very high quality of the watermarked image and a strong resistance to print and scan processes. We propose a method which combines an additive watermarking algorithm in the spatial domain providing resistance against cropping and exhaustive search and a synchronisation template in the Fourier domain providing resistance against geometrical deformations. The additive watermark in the spatial domain is based on an original generalized 2-D cyclic patterns for secret message embedding. The cyclic property and pattern redundancy facilitate detection and synchronisation against cropping and image processing basic attacks (like compression, filtering, blurring). This algorithm has to be complemented by a template insertion for providing resistance against rotation and scaling which are caused by the print and scan processes. We generate the template in the Fourier domain inserting some points locally. The watermark and the template are weighted by a human visual masking function. The global scheme, though very classical in its global concept, provides very efficient protection to a digital image which could be delivered both in a digital high quality format and in a printed form.

The need for reversible or lossless watermarking methods has been highlighted to associate information with losslessly processed media or to enable their authentication. The paper first analyzes the specificity and the application scope of lossless watermarking methods. An original circular interpretation of a bijective transformation is then proposed to implement a method that fulfill all quality and functionality requirements.

We present a new algorithm for the nonrigid registration of three-dimensional magnetic resonance (MR) intraoperative image sequences showing brain shift. The algorithm tracks key surfaces of objects (cortical surface and the lateral ventricles) in the image sequence using a deformable surface matching algorithm. The volumetric deformation field of the objects is then inferred from the displacements at the boundary surfaces using a linear elastic biomechanical finite-element model. Two experiments on synthetic image sequences are presented, as well as an initial experiment on intraoperative MR images showing brain shift. The results of the registration algorithm show a good correlation of the internal brain structures after deformation, and a good capability of measuring surface as well as subsurface shift. We measured distances between landmarks in the deformed initial image and the corresponding landmarks in the target scan. Cortical surface shifts of up to 10 mm and subsurface shifts of up to 6 mm were recovered with an accuracy of 1 mm or less and 3 mm or less respectively.

This paper describes a new method for generating 2-D watermarks with good synchronizability and secrecy properties. The aim of the work is primarily to develop a watermarking scheme that efficiently and securely deals with cropping and object segmentation. It differs from other existing schemes by its innovative pattern generation algorithm. On the basis of a 1-D sequence, we propose an original method to generate a redundant 2-D pattern with cyclic properties depending on a secret key. The cyclic property facilitates the synchronization for the detection of the watermark when the reference marks are lost. Possible applications fields of this technique are object oriented scene authentication as well as classical intellectual property rights issues.

In this paper, SNR scalable representations of video signals are studied. The investigated codecs are well suited for communications applications because they are all based on backward motion-compensated predictive coding, which provides the necessary low-delay property. In a very-low bit rate context (VLBR), the matching pursuits (MP) signal representation algorithm is used to represent the displaced frame difference (DFD) of each layer of a multilevel decomposition of the video signal. A number of conventional prediction schemes that can be generalized to any DFD representation technique are considered. They are compared with an original and MP specific DFD prediction method. Two scenari have been considered. In the first scenario, an enhancement layer is built on a base layer that has been encoded using a classical, i.e., nonscalable scheme. In that case, all methods appear to be comparable, In the second scenario, the fact that the base layer is used as a reference for an enhancement layer is taken into account to build it. In that case, the proposed MP prediction method clearly outperforms all other conventional approaches, Additional lessons can be drawn from this work. The same motion vectors can be used in both SNR layers, and the DFD prediction between layers improves coding efficiency. Moreover, the MP representation of the signal enable us to measure the predictability of the high SNR layer DFD from the low SNR layer DFD, i.e., to quantify the part of the low SNR layer information that also belongs to the high SNR layer.

We propose a new signed or unsigned Euclidean distance transformation algorithm, based on the local corrections of the well-known 4SED algorithm of Danielsson (1980). Those corrections are only applied to a small neighborhood of a small subset of pixels from the image, which keeps the cost of the operation low. In contrast with all fast algorithms previously published, our algorithm produces perfect Euclidean distance maps in a time linearly proportional to the number of pixels in the image. The computational cost is close to the cost of the 4SSED approximation.

In this paper we propose an original study of the reconstruction of subband compressed images impaired by channel transmission errors. The method proceeds in two steps:first a detection scheme is applied to determine which coefficients of the subband decomposition have been affected by transmission, and then an estimation step tries to evaluate the erroneous coefficients. In our model, subband coefficients are considered to be drawn from jointly Gaussian random processes. Based on this assumption, conditional statistics can be computed which enable to test the likelihood of a given set of received coefficients with respect to the rest of the image. The detection and estimation processes are derived from these statistics. The method is validated through simulation and visual results are provided. The drawbacks of the method are outlined and explained through the discrepancies between the Gaussian assumption and real world images, namely around image edges.

In this paper, we extend structures and factorizations for paraunitary filter banks yielding linear phase filters to the oversampled case. We derive minimal factorizations for a subset of these banks, and propose some design examples.

"Matching pursuits" is a signal expansion technique whose efficiency for video coding has already been largely demonstrated in the MPEG-4 framework. In this paper, our attention focuses on complexity issues. First, the most expensive step of the signal expansion process is significantly speeded up by exploiting results achieved in the wavelet and multiresolution theory. A subband dictionary is proposed as an alternative to the Gabor dictionary that has been used up to now. Equivalent levels of quality are achieved with both dictionaries, but the computational cost is significantly reduced when using the subband one. Then, we explain how, with any dictionary, the linearity of the inner product could be exploited to further speed up the process in return for an increased amount of memory.

This paper presents a methodology for the restoration of the visual quality of still images affected by coding noise. This quality restoration is achieved only by considering the additive coding noise and is therefore limited to an adaptive postprocessing filtering. It is based on a model of the human visual system that considers the relationship between visual stimuli and their visibility. This phenomenon known as masking is used as a criterion for the locally adaptive filtering design. An image transformation that yields visual stimuli tuned to the frequency and orientation according to the perceptual model is proposed. It allows a local measure of the masking of each perceptual stimulus considering the contrast between signal and estimated noise. This measure is obtained by analytic filtering. Processing schemes are presented with applications to the discrete cosine transform (DCT) and subband coded images. One proposed solution considers the characteristics of DCT coding noise for the estimation of the noise. Another solution is based on a "blind" neural estimation of the noise characteristics. Experimental results of the proposed approaches show significant improvements of the visual quality, which validates our perceptual model and filtering.

This paper proposes an asymmetric scheme for motion estimation/compensation. While the estimation is performed with a classical block matching algorithm, the motion information is decoded by using an active mesh in order to implement the compensation stage. A mesh is positioned by taking into account the relevant spatial information of the image to be compensated and is used afterwards to reconstruct the motion information. Two main issues have to be addressed for conducting such a motion compensation technique: (i) how to optimally design an active mesh, (ii) how to reverse and interpolate a backward motion field estimated on an a priori grid of fixed-size blocks so as to determine a forward motion field on variable size triangular patches. While proposing a solution to these two problems, particular attention is paid to the computational burden. Such a scheme opens the possibility for added manipulation functionalities because of mesh capabilities.

Iterated transformation theory (ITT) coding, also known as fractal coding, in its original form, allows fast decoding but suffers from long encoding times. During the encoding step, a large number of block best-matching searches have to be performed which leads to a computationally expensive process. We present a new method that significantly reduces the computational load of ITT based image coding. Both domain and range blocks of the image are transformed into the frequency space. Domain blocks are then used to train a two dimensional Kohonen (1982, 1989, 1990) neural network (KNN) forming a code book similar to vector quantization coding. The property of KNN (and self-organizing feature maps in general) which maintains the topology of the input space allows to perform a neighboring search so as to find the piecewise transformation between domain and range blocks.

"Matching pursuits" is a signal expansion technique whose efficiency for video coding has already been demonstrated in the MPEG-4 community. In this paper, we focus our attention on some aspects of the algorithm. Firstly, the relevance of the orthogonalization of the signal expansion is investigated. Secondly, a subband dictionary is proposed as an alternative to the Gabor dictionary, used up to now. Equivalent levels of quality are achieved with both dictionaries but the computational cost is significantly reduced when using the subband one. Eventually, the entropy coding stage of the algorithm is studied.

The European High-Performance Information Infrastructure in Medicine, n(o)B3014 (HIM3) project of the Trans-European Network--Integrated Broadband Communications (TEN-IBC) program, started on March 1996 and finished on February 1997, aimed to test the medical usability of the European asynchronous transfer mode (ATM) network in medical image transmission. The Department of Radiology, University of Pisa, Pisa, Italy, and St-Luc University Hospital, Brussels, Belgium, involved in the project as healthcare partners in the radiological domain, established several connection sessions finalized to test the usability of Digital Imaging and Communication (DICOM) image transmission and interactive telediagnosis tools in the daily radiological practice. The Pisa site was connected to the Italian ATM pilot (Sirius Network) through the Tuscany metropolitan area network (MAN), while St-Luc University Hospital was connected to Belgium ATM network through the Brussels MAN. By means of international connections provided by the European JAMES project, a link between the two sites was established, connecting both national ATM networks. Due to the large variety of hardware present in the medical centers, multiplatform software tools were used and tested: central test node (CTN) release 2.8 [3], VAT [6], NV-3.3 [7], and IDI (UCL homemade multiplatform teleradiology tool for interactive visualization and processing of DICOM images). During the telediagnosis session, lead by radiologists in both hospitals, each site submitted neuroradiological clinical cases to the other for remote consultation. The connection, available for a period of two weeks, at 2-Mbit/s bandwidth, allowed the transmission of MR images (256 x 256 x 12 bit) and simultaneous multimedia interactive discussion of the cases. Both off-line transmission and review of the images, using the CTN DICOM transfer routines, and on-line interactive image discussion, using the IDI telediagnosis software, were tested successfully from the technical and medical point of view.

A new region growing algorithm has been proposed for computing Euclidean distance maps in a time comparable to widely used chamfer distance transform. We show how this algorithm can be extended to more complex tasks such as the computation of distance maps on anisotropic grids and the generation of a new type of Euclidean skeletons.

This paper introduces a new image restoration method based on a 1-D Kalman filtering. Using the model of tuned channels, the corrupted image is decomposed into a set of perceptual components characterized by different orientations and frequencies. The restoration step is then performed on each component in one dimension following the appropriate orientation with the well-known Kalman algorithm. Since the decomposition provides perfect reconstruction, the restored image is the recomposition of all the restored components. This approach yields relevant results for 2-D blurred images, using 1-D low order models. Unlike traditional 2-D Kalman restoration techniques, its implementation has no excessive computational load.

The tool introduced in this paper allows to automatically decide in an image or in a video sequence which regions are important and which ones are not. For this purpose, fuzzy logic has been used to modelize human subjective knowledge about the way to allocate priorities to regions. The resulting classification can be used in a wide range of applications going from image coding to image understanding.

The present paper introduces an original way to perform motion estimation and compensation. The scheme is asymmetric in the sense that the compensation stage is not merely the reverse operation of the estimation process. Motion estimation is computed using a classical block matching algorithm and the set of vectors that compose the motion information are classically transmitted to the decoder. The motion compensation uses them to deform a wireframe structure whose vertices are displaced according to the motion vectors. Conditions must be set up on the number of wireframe vertices and on their location. In comparison with a block-based compensation, the main advantage is the possibility to take into account the object information, which results in perceptually more pleasant predicted pictures.

A multiresolution lossless image compression scheme based on several new tools will be presented in this paper. An improved multiresolution Haar transform will be developed, combined with pre- and post-processing and with a new entropy coder, to give the overall compression scheme. This research is mainly useful for image compression in the medical imaging area, where changes between original and decompressed images are to be avoided.

Motion estimation is crucial in domains such as robot vision or image sequence coding. Fast or real time computing is often required but is problematic due to computational complexity. In this paper, an algorithm using multiresolution multigrid Markov random fields is parallelized. Such realizations have already been studied for fine parallelizations on massively parallel machines. The goal of this paper is to achieve a coarse parallelization on a network of workstations, using the Parallel Virtual Machine (PVM) software package.

A simple interpolation method from low frequencies to high frequencies in a two-band filter bank is shown to be a very efficient way for enhancing signal zooming or for erasing blocking artifacts. This paper introduces a new and simple way for generating that spatial adaptability of the filters. It shows that a linear interpolation between the two subbands can be reduced to a modification of the filters. From this result, a local adaptation of the filters is envisaged for two applications: signal zooming and reduction of blocking artifacts in multigrid decomposition of images.

Registration of three-dimensional medical image modalities (PET and MRI) of the human brain is performed by a multi-scale morphological shape description combined with 3D Chamfer matching. Objects of Interest are first segmented. For MR images of the head, a Directional Watershed algorithm is used to obtain an accurate segmentation of the brain. PET images are segmented by thresholding combined with region growing. The shape description is then performed by generalized morphological skeletons, introduced in this article, providing a multi-scale representation of the object shape. Registration is finally operated by Chamfer matching, using a rigid transform for PET-MRI registration. Non-rigid registration is also evoked.

We present a project developed in our University in order to provide Medical Imaging Departments with efficient software tools for 3D medical image processing and transmission. In this context, the IMIS software package has been developed, combining, in a modular programming strategy, an easily upgradable graphical user interface, using the Tcl/Tk toolkit, with high performance image processing techniques, such as 3D lossless multiresolution image compression for telediagnosis on narrow-band ISDN networks. IMIS is based on a set of tools allowing the handling of 3D medical images such as magnetic resonance images (MRI), positron emission tomography (PET), and computed tomography (CT) of functional MRI (fMRI).

This paper investigates the development of a new image quality criterion based on the psychovisual model of tuned channels and more particularly the phenomenon of masking. The masking model parameters have been evaluated by psychovisual tests assuming a logarithmic relationship between the visibility threshold and the contrast value of the background for a given perceptual channel. The masking has the consequence that only a part of the noise of a noisy image is really visible and affects the visual quality of the image. The idea of the proposed criterion is to evaluate the image quality after the cancellation of the masked noise defined as the invisible noise. The criterion has been used in order to compare different coders and different post-processings of noisy images.

Digital TV broadcasting needs new cryptological tools for conditional access, copyright protection and image authentication. The aim of this paper is to overview the corresponding systems' features. The description of a conditional access system is given. It is shown that equitable systems need the use of a trusted third party. The design of efficient copyright protection by watermarking images and image authentication by signatures are also briefly discussed.>

The aim of image restoration is to find an estimate of the ideal image using a priori information about blur and/or noise and/or the ideal image. Classical criterion are minimum least square, minimum mean square error or maximum a posteriori probability. The choice of the criterion used to measure the estimation quality is crucial for the design of the restoration algorithm. The authors propose a new criterion based on a visual model: it is based on perceptual masking. Thereafter, they propose a new restoration algorithm dealing only with additive noise. The perceptual components of the image to restore which are corrupted by an additive noise above a visibility threshold are simply set to zero. Some results obtained for the post-processing of JPEG images are presented.>

The quadtree-structured decomposition of images is used in image compression in order to have a low cost region description associated with a particular transform. When this transform is reduced to the mean value on each block, high compression ratios can be obtained. However, in this case a blocking effect arises. In the paper, a reconstruction method using smooth non-symmetrical interpolation functions is introduced. This method represents the information of the tree in a far more satisfactory way for the visual system than non-overlapping blocks reconstruction.>

We present a new method for the automatic recognition of cancerous cells from a digitized picture of a microscopic section. The method is based on the analysis of four criteria of malignancy, in relation with the shape and the size of the observed cells. It provides the physician with nonsubjective numerical values for the four concerned criteria of malignancy, in order to help him to decide whether the tissue is cancerous or not. The automatic approach described uses mathematical morphology, first to remove the background noise from the image and to operate a segmentation of the nuclei of the cells, which contain most of the features of malignancy, next to analyse the shape and the size of these nuclei and lastly to evaluate their texture. From the values of the four extracted criteria, an automatic classification of the image is using a Kohonen neural network.>

This paper deals with the question of optimizing the filters in the upsampling stage of a TV/HDTV compatible pyramidal coder. From a coding gain point of view, both the decimation and upsampling filters should be optimized. In the frame of compatible coding, not only the coding efficiency influences the choice of the decimation filter but also the compatible image quality. Therefore, assuming this filter has been fixed, we analyze the question of optimizing the upsampling filter in order to obtain the highest coding gain. This question is addressed for a mean squared error (MSE) criterion. In addition, assuming the base layer (TV) signal can be quantized, the influence on the quantization noise on the optimal interpolation filter is investigated and the problem is handled for the MSE criterion. As the statistical properties of pictures are required in the optimization, a model is then developed to compute these properties when there is motion. The model takes into account the processing of progressive sources and, concerning interlaced sequences, the independent processing of fields or the processing of merged fields. Results are then derived for the three types of processing.

In this paper, we present systolic architecture based on the lattice structure of filters. The main characteristic of this architecture is the systolism: computations are pipelined in many identical locally interconnected processing elements (PEs). These PEs are simple and can reach high frequency clock while working at any time of the process. So, the speed of the circuit can be increased. The implementation of the filters, through VLSI techniques, is facilitated by the repetitive nature of the elements. In section 2, we describe the multiresolution scheme and the lattice structures. If the lattice structure appears as an efficient remedy for the finite length of the multipliers, special attention has to be kept on the computation noise which appears when the datapath is limited to some finite width. The goal of the so-related study consists in keeping this computation noise below the quantization noise (coming from the quantizers) at a reasonable cost. In section 3, we present the basic processing element and its use among the different stages of the filter. Section 4 deals with the finite representation of the data throughout the datapath.

In standard DCT coding schemes like MPEG, the sequences compression is achieved by motion compensation, transformation, quantization, and entropy coding. In this paper, we have followed the same path by adapting to the image signal the elements of the coding scheme. The motion compensation is achieved by a block-matching method, where the size of the blocks is adapted to the signal. Great attention has been paid to the relevance of the motion field. Combined with the motion compensation, the two fields of each frame are merged, taking into account the measured motion vectors, to compose a pseudo-progressive frame. The encoding is applied to this `motion-compensated progressive' frame. A wavelet decomposition is then applied on each (inter or intra) frame. Such a transform, intrinsically owning linear- phase and perfect reconstruction properties, has been optimized for maximizing a perceptually weighted coding gain. The wavelet coefficients are thereafter vector-quantized, in order to reach the maximum perceptual SNR : frequency weighting is taken into account. The relevance of the measured vector field allows a precise spatio-temporal quantization optimization. The vectors are entropy coded taking into account the remaining inter-band dependence, by an adapted entropy code. Results obtained from 1 Mbit/s to 8 Mbit/s are shown for moving sequences at the conference.

The classical optimization criterion used is the signal-to-(quantization) noise ratio (SNR). A general quantization methodology is developed which is based on the optimization of a frequency-weighted SNR: first, a quality criterion based on the weighted SNR is derived from the CCIR (Comite Consultatif International de Radiodiffusion) curves for analog TV: then, a methodology for optimizing the quantization is derived in a general form.>

The authors consider the problem of designing multiresolution transforms that are adapted to the given image signal, in the sense that they maximize the coding gain at each resolution level. A simple alternating optimization algorithm is derived for solving this problem in the framework of the lattice realization of para-unitary quadrature mirror filters. The resulting large coding scheme is discussed in some detail, and its performance is compared with that of the discrete cosine transform (JPEG) technique and with that of some nonadapted multiresolution transforms.>

Consideration is given to the problem of designing a signal-adapted two-band filtering system that has the perfect reconstruction property for application to multiresolution signal coding. A natural optimization criterion for this problem is introduced, and an iterative algorithm for computing optimal filters in lattice form is described. Some simulation results for the application of the technique to image coding are presented.>

This paper presents a human visual quality criterion for still monochrome pictures. The main idea is based on the decomposition of the signal in tuned channels with respect to the location in the visual field, the orientation and the frequency. The specific sensitivity of each channel and the self-masking are taken into account to evaluate the quality of the picture.

Subband Coding of images, which can be seen as an extension of orthogonal transformations, has been introduced by Woods and O’Neil [8]. In [8], the subbands are further encoded by DPCM. In this study, it is proposed to use subband coding as a stand alone coding technique. The emphasis is put on the question of subband quantization. A methology is proposed where the quantization step of the subbands is taken so as to minimize a quantization noise power weighted by a sensitivity function of the eye.