The whisky industry heavily relies on peat as a key ingredient to impart distinctive smoky flavors to the final product. However, traditional methods for analyzing peat composition and quality are time-consuming and destructive, requiring extensive sample preparation. To address these challenges, we propose a novel nondestructive system for rapid and accurate peat analysis combining push-broom hyperspectral imaging (HSI), singular spectrum analysis (SSA), and machine learning. We introduce a faster SSA variant (SSA++) to overcome the high computational complexity of traditional SSA, enabling real-time processing of HSI data when captured in the push-broom manner. Comprehensive experiments have demonstrated the effectiveness of the proposed system, achieving a total phenol estimation of up to 99.31%R2. SSA++ maintains similar accuracy to SSA while significantly reducing computational time, enabling real-time performance. Our system offers a powerful tool for automated peat analysis, facilitating smart manufacturing and enhanced quality monitoring in the whisky industry.

Segmentation masks of pathological areas are useful in many medical applications, such as brain tumour and stroke management. Moreover, healthy counterfactuals of diseased images can be used to enhance radiologists' training files and to improve the interpretability of segmentation models. In this work, we present a weakly supervised method to generate a healthy version of a diseased image and then use it to obtain a pixel-wise anomaly map. To do so, we start by considering a saliency map that approximately covers the pathological areas, obtained with ACAT. Then, we propose a technique that allows to perform targeted modifications to these regions, while preserving the rest of the image. In particular, we employ a diffusion model trained on healthy samples and combine Denoising Diffusion Probabilistic Model (DDPM) and Denoising Diffusion Implicit Model (DDIM) at each step of the sampling process. DDPM is used to modify the areas affected by a lesion within the saliency map, while DDIM guarantees reconstruction of the normal anatomy outside of it. The two parts are also fused at each timestep, to guarantee the generation of a sample with a coherent appearance and a seamless transition between edited and unedited parts. We verify that when our method is applied to healthy samples, the input images are reconstructed without significant modifications. We compare our approach with alternative weakly supervised methods on the task of brain lesion segmentation, achieving the highest mean Dice and IoU scores among the models considered.

Functional Magnetic Resonance Imaging (fMRI) is used for extracting blood oxygen signals from brain regions to map brain functional connectivity for brain disease prediction. Despite its effectiveness, fMRI has not been widely used: on the one hand, collecting and labeling the data is time-consuming and costly, which limits the amount of valid data collected at a single healthcare site; on the other hand, integrating data from multiple sites is challenging due to data privacy restrictions. To address these issues, we propose a novel, integrated Federated learning and Split learning Spatio-temporal Graph framework (F G). Specifically, we introduce federated learning and split learning techniques to split a spatio-temporal model into a client temporal model and a server spatial model. In the client temporal model, we propose a time-aware mechanism to focus on changes in brain functional states and use an InceptionTime model to extract information about changes in the brain states of each subject. In the server spatial model, we propose a united graph convolutional network to integrate multiple graph convolutional networks. Integrating federated learning and split learning, F G can utilize multi-site fMRI data without violating data privacy protection and reduce the risk of overfitting as it is capable of learning from limited training data sets. Moreover, it boosts the extraction of spatio-temporal features of fMRI using spatio-temporal graph networks. Experiments on ABIDE and ADHD200 datasets demonstrate that our proposed method outperforms state-of-the-art methods. In addition, we explore biomarkers associated with brain disease prediction using community discovery algorithms using intermediate results of F G. The source code is available at https://github.com/yutian0315/FS2G.

An ideal artificial intelligence (AI) system should have the capability to continually learn like humans. However, when learning new knowledge, AI systems often suffer from catastrophic forgetting of old knowledge. Although many continual learning methods have been proposed, they often ignore the issue of misclassifying similar classes and make insufficient use of textual priors of visual classes to improve continual learning performance. In this study, we propose a continual learning framework based on a pre-trained vision-language model (VLM) that does not require storing old class data. This framework utilizes parameter-efficient fine-tuning of the VLM's text encoder for constructing a shared and consistent semantic textual space throughout the continual learning process. The textual priors of visual classes are encoded by the adapted VLM's text encoder to generate discriminative semantic representations, which are then used to guide the learning of visual classes. Additionally, fake out-of-distribution (OOD) images constructed from each training image further assist in the learning of visual classes. Extensive empirical evaluations on three natural datasets and one medical dataset demonstrate the superiority of the proposed framework.

In some medical imaging tasks and other settings where only small parts of the image are informative for the classification task, traditional CNNs can sometimes struggle to generalise. Manually annotated Regions of Interest (ROI) are often used to isolate the most informative parts of the image. However, these are expensive to collect and may vary significantly across annotators. To overcome these issues, we propose a framework that employs saliency maps to obtain soft spatial attention masks that modulate the image features at different scales. We refer to our method as Adversarial Counterfactual Attention (ACAT). ACAT increases the baseline classification accuracy of lesions in brain CT scans from $71.39 %$ to $72.55 %$ and of COVID-19 related findings in lung CT scans from $67.71 %$ to $70.84 %$ and exceeds the performance of competing methods. We investigate the best way to generate the saliency maps employed in our architecture and propose a way to obtain them from adversarially generated counterfactual images. They are able to isolate the area of interest in brain and lung CT scans without using any manual annotations. In the task of localising the lesion location out of 6 possible regions, they obtain a score of $65.05 %$ on brain CT scans, improving the score of $61.29 %$ obtained with the best competing method.

We present a novel method to segment instances of glandular structures from colon histopathology images. We use a structure learning approach which represents local spatial configurations of class labels, capturing structural information normally ignored by sliding-window methods. This allows us to reveal different spatial structures of pixel labels (e.g., locations between adjacent glands, or far from glands), and to identify correctly neighboring glandular structures as separate instances. Exemplars of label structures are obtained via clustering and used to train support vector machine classifiers. The label structures predicted are then combined and post-processed to obtain segmentation maps. We combine hand-crafted, multi-scale image features with features computed by a deep convolutional network trained to map images to segmentation maps. We evaluate the proposed method on the public domain GlaS data set, which allows extensive comparisons with recent, alternative methods. Using the GlaS contest protocol, our method achieves the overall best performance.

The classification of blood vessels into arterioles and venules is a fundamental step in the automatic investigation of retinal biomarkers for systemic diseases. In this paper, we present a novel technique for vessel classification on ultra-wide-field-of-view images of the retinal fundus acquired with a scanning laser ophthalmoscope. To the best of our knowledge, this is the first time that a fully automated artery/vein classification technique for this type of retinal imaging with no manual intervention has been presented. The proposed method exploits hand-crafted features based on local vessel intensity and vascular morphology to formulate a graph representation from which a globally optimal separation between the arterial and venular networks is computed by graph cut approach. The technique was tested on three different data sets (one publicly available and two local) and achieved an average classification accuracy of 0.883 in the largest data set.

We propose a novel multiple instance learning method to assess the visibility (visible/not visible) of the retinal nerve fiber layer (RNFL) in fundus camera images. Using only image-level labels, our approach learns to classify the images as well as to localize the RNFL visible regions. We transform the original feature space to a discriminative subspace, and learn a region-level classifier in that subspace. We propose a margin-based loss function to jointly learn this subspace and the region-level classifier. Experiments with a RNFL dataset containing 884 images annotated by two ophthalmologists give a system-annotator agreement (kappa values) of 0:73 and 0:72 respectively, with an inter-annotator agreement of 0:73. Our system agrees better with the more experienced annotator. Comparative tests with three public datasets (MESSIDOR and DR for diabetic retinopathy, UCSB for breast cancer) show that our novel MIL approach improves performance over the state-of-the-art. Our Matlab code is publicly available at https://github.com/ManiShiyam/Sub-category-classifiersfor- Multiple-Instance-Learning/wiki.

Accurate vessel detection in retinal images is an important and difficult task. Detection is made more challenging in pathological images with the presence of exudates and other abnormalities. In this paper, we present a new unsupervised vessel segmentation approach to address this problem. A novel inpainting filter, called neighborhood estimator before filling, is proposed to inpaint exudates in a way that nearby false positives are significantly reduced during vessel enhancement. Retinal vascular enhancement is achieved with a multiple-scale Hessian approach. Experimental results show that the proposed vessel segmentation method outperforms state-of-the-art algorithms reported in the recent literature, both visually and in terms of quantitative measurements, with overall mean accuracy of 95.62% on the STARE dataset and 95.81% on the HRF dataset.

Deep learning has shown great potential for curvilinear structure (e.g., retinal blood vessels and neurites) segmentation as demonstrated by a recent auto-context regression architecture based on filter banks learned by convolutional sparse coding. However, learning such filter banks is very time-consuming, thus limiting the amount of filters employed and the adaptation to other data sets (i.e., slow re-training). We address this limitation by proposing a novel acceleration strategy to speed-up convolutional sparse coding filter learning for curvilinear structure segmentation. Our approach is based on a novel initialisation strategy (warm start), and therefore it is different from recent methods improving the optimisation itself. Our warm-start strategy is based on carefully designed hand-crafted filters (SCIRD-TS), modelling appearance properties of curvilinear structures which are then refined by convolutional sparse coding. Experiments on four diverse data sets, including retinal blood vessels and neurites, suggest that the proposed method reduces significantly the time taken to learn convolutional filter banks (i.e., up to -82%) compared to conventional initialisation strategies. Remarkably, this speed-up does not worsen performance; in fact, filters learned with the proposed strategy often achieve a much lower reconstruction error and match or exceed the segmentation performance of random and DCT-based initialisation, when used as input to a random forest classifier.

This paper introduces a low-rank prior in small oriented noise-free image patches. Considering an oriented patch as a matrix, a low-rank matrix approximation is enough to preserve the texture details in the optimally oriented patch. Based on this prior, we propose a single-patch method within a generalized joint low-rank and sparse matrix recovery framework to simultaneously detect and remove nonpointwise random-valued impulse noise (e.g., very small blobs). A weighting matrix is incorporated in the framework to encode an initial estimate of the spatial noise distribution. An accelerated proximal gradient method is adapted to estimate the optimal noise-free image patches. Experiments show the effectiveness of our framework in detecting and removing nonpointwise random-valued impulse noise.

We explore post-processing of scanning laser ophthalmoscope (SLO) images for the automatic detection of retinal blood vessels. The retinal vasculature is first enhanced using morphological and Gaussian matched filters before a thresholding technique produces a binary vessel map. Such permutations of post-processing techniques are commonly used to achieve unsupervised classification of the vasculature in fundus images, and it is the purpose of this study to investigate their applicability to SLO imaging. We compare the results of vascular detection as performed on SLO and fundus images.

This paper introduces a `low-rank prior' for small oriented noise-free image patches: considering an oriented patch as a matrix, a low-rank matrix approximation is enough to preserve the texture details in the properly oriented patch. Based on this prior, we propose a single-patch method within a generalized joint low-rank and sparse matrix recovery framework to simultaneously detect and remove non-point wise random-valued impulse noise (e.g., very small blobs). A weighting matrix is incorporated in the framework to encode an initial estimate of the spatial noise distribution. An accelerated proximal gradient method is adapted to estimate the optimal noise-free image patches. Experiments show the effectiveness of our framework in removing non-point wise random-valued impulse noise.

This papers describes GroBa, a new lumen calibre measurement technique based on growing balloons. GroBa presents the advantages of cross-sectional based methods, as it is able to cope with irregular, non-tubular vessel structures, such as stenosis or aneurysms, but at the same time it is able to obtain precise calibre measurements even when the estimated centrelines are not accurate. Experimental results using phantoms and real subtracted full-body magnetic resonance angiograms show the potential of this work. GroBa is integrated into a fully automatic system that segments the vasculature, obtains its centrelines, measures the lumen calibre at each detected artery, and presents the calibre information in false colours in the maximum intensity projection exploiting the HSV colour-space; all without any human intervention.

In this paper we present an analysis of image features used to discriminate arteries and veins in digital fundus images. Methods proposed in the literature to analyze the vasculature of the retina and compute diagnostic indicators like the Arteriolar to Venular ratio (AVR), use, in fact, different approaches for this classification task, extracting different color features and exploiting different additional information. We concentrate our analysis on finding optimal features for the vessel classification, considering not only simple color features, but also spatial location and vessel size and testing different supervised labeling approaches. The results obtained show that best results are obtained mixing features related with color values and contrast inside and outside the vessels and positional information. Furthermore, the discriminative power of the features changes with the image resolution and best results are not obtained at the finest one. Our experiments demonstrate that using a good set of descriptors it is possible to achieve very good classification performances even without using vascular connectivity information.

We present RERBEE (robust efficient registration via bifurcations and elongated elements), a novel feature-based registration algorithm able to correct local deformations in high-resolution ultra-wide field-of-view (UWFV) fluorescein angiogram (FA) sequences of the retina. The algorithm is able to cope with peripheral blurring, severe occlusions, presence of retinal pathologies and the change of image content due to the perfusion of the fluorescein dye in time. We have used the computational power of a graphics processor to increase the performance of the most computationally expensive parts of the algorithm by a factor of over × 1300, enabling the algorithm to register a pair of 3900 × 3072 UWFV FA images in 5-10 min instead of the 5-7 h required using only the CPU. We demonstrate accurate results on real data with 267 image pairs from a total of 277 (96.4%) graded as correctly registered by a clinician and 10 (3.6%) graded as correctly registered with minor errors but usable for clinical purposes. Quantitative comparison with state-of-the-art intensity-based and feature-based registration methods using synthetic data is also reported. We also show some potential usage of a correctly aligned sequence for vein/artery discrimination and automatic lesion detection.

In this paper we present a novel method for the automatic location and segmentation of the optical disk in fundus images. It is based on the decoupling of vessel and background information obtained with morphological segmentation and inpainting. A multiresolution optimization scheme finding elliptic contours optimally adapted to a brightness model is then applied on the inpainted brightness image, under the constraint that a reasonable amount of vasculature must be present inside the disc. An effective objective function and a multiresolution scheme allow a deterministic optimizer to converge on the OD contour with good accuracy, while the vessel-based constraint limits wrong contour detections in anomalous cases. Our initial experiments (30 DRIVE images, ground truth from two doctors) suggest that the method can provide accurate results both in term of optic disc location and contour segmentation accuracy.

Image descriptors are widely adopted structures to match image features. SIFT-based descriptors are collections of gradient orientation histograms computed on different feature regions, commonly divided by using a regular Cartesian grid or a log-polar grid. In order to achieve rotation invariance, feature patches have to be generally rotated in the direction of the dominant gradient orientation. In this paper we present a modification of the GLOH descriptor, a SIFT-based descriptor based on a log-polar grid, which avoids to rotate the feature patch before computing the descriptor since predefined discrete orientations can be easily derived by shifting the descriptor vector. The proposed descriptors, called sGLOH and sGLOH+, have been compared with the SIFT descriptor on the Oxford image dataset, with good results which point out its robustness and stability.

Research on 3D video processing has gained a tremendous amount of momentum due to advances in video communications, broadcasting and entertainment technology (e.g., animation blockbusters like Avatar and Up). There is an increasing need for reliable technologies capable of visualizing 3-D content from viewpoints decided by the user; the 2010 football World Cup in South Africa has made very evident the need to replay crucial football footage from new viewpoints to decide whether the ball has or has not crossed the goal line. Remote videoconferencing prototypes are introducing a sense of presence into large- and small-scale (PC-based) systems alike by manipulating single and multiple video sequences to improve eye contact and place participants in convincing virtual spaces. All this, and more, is pushing the introduction of 3D services and the development of high-quality 3D displays to be available in a future which is drawing nearer and nearer.

This paper presents a method for automated vessel segmentation in retinal images. For each pixel in the field of view of the image, a 41-D feature vector is constructed, encoding information on the local intensity structure, spatial properties, and geometry at multiple scales. An AdaBoost classifier is trained on 789 914 gold standard examples of vessel and nonvessel pixels, then used for classifying previously unseen images. The algorithm was tested on the public digital retinal images for vessel extraction (DRIVE) set, frequently used in the literature and consisting of 40 manually labeled images with gold standard. Results were compared experimentally with those of eight algorithms as well as the additional manual segmentation provided by DRIVE. Training was conducted confined to the dedicated training set from the DRIVE database, and feature-based AdaBoost classifier (FABC) was tested on the 20 images from the test set. FABC achieved an area under the receiver operating characteristic (ROC) curve of 0.9561, in line with state-of-the-art approaches, but outperforming their accuracy ( 0.9597 versus 0.9473 for the nearest performer).

In this paper we address the problem of human body pose estimation from still images. A multi-view set of images of a person sitting at a table is acquired and the pose estimated. Reliable and efficient pose estimation from still images represents an important part of more complex algorithms, such as tracking human body pose in a video sequence, where it can be used to automatically initialise the tracker on the first frame. The quality of the initialisation influences the performance of the tracker in the subsequent frames. We formulate the body pose estimation as an analysis-by-synthesis optimisation algorithm, where a generic 3D human body model is used to illustrate the pose and the silhouettes extracted from the images are used as constraints. A simple test with gradient descent optimisation run from randomly selected initial positions in the search space shows that a more powerful optimisation method is required. We investigate the suitability of the Particle Swarm Optimisation (PSO) for solving this problem and compare its performance with an equivalent algorithm using Simulated Annealing (SA). Our tests show that the PSO outperforms the SA in terms of accuracy and consistency of the results, as well as speed of convergence.

Current assessment of progress in construction projects is a manual task that is often infrequent and error prone. Images of sites are extremely cluttered and rife with shadows, occlusions, equipment, and people- making them extremely hard to analyse. We present a first prototype system capable of detecting changes on a building site observed by a fixed camera, and classifying such changes as either actual structural events, or as unrelated. We exploit a prior building model to align camera and scene, thus identifying image regions where building components are expected to appear. This then enables us to home in on significant change events and verify the actual presence of a particular type of component. We place our approach within an emerging paradigm for integration in the construction industry, and highlight the benefits of automated image based feedback.

This paper presents an evolutionary approach to estimating upper-body posture from multi-view markerless sequences. We fit a 24-dof skeleton model to sparse 3-D stereo data from an array of cameras. We use a particle swarm optimization algorithm which is intrinsically parallel, can incorporate constraints and does not require motion models. We subdivide the high-dimensional search space based on limb dynamics from application sequences and perform hierarchical fitting from the least to the most uncertain body parts. We show experimentally the advantages of this scheme against non-hierarchical optimization in terms of sharper error decrease. We report results with 3-D scanner data of a model human and noisy, calibrated stereo disparity maps of a real videoconferencing scene. 1 Introduction and

In this paper we describe the application of Particle Swarm Optimisation to the problem of human body pose estimation from multiple view video sequences. We use a subdivision body model with an underlying skeleton layer to estimate and illustrate the body pose. The optimisation looks for the best match between the silhouettes extracted from the original video sequence and the silhouettes generated by the projection of the model in a pose suggested by the PSO. The original PSO algorithm is applied hierarchically and combined with the full overall optimisation to decrease the effects of error propagation. Results demonstrate the ability of PSO to reliably recover the correct body pose from 4-viewpoint video sequences.

This paper describes a new framework for the automated tracking of the central retinal vein in retinal images. The procedure first computes a binary image of the retinal vasculature, then obtains the skeleton (medial axis) of the vascular network. Terminal and branching points of the network are then located, and the network converted into a graph representation including length and thickness information for all vessels. Finally, a maxmin approach is used to locate the central vein: the candidates central vein are the minimal paths from the optic disk to all terminal nodes found using Dijkstra algorithm. The actual central vein is selected among the all candidates by maximizing a merit function estimating the total vessel area in the image. Results are presented and compared with those provided by a manual classification on 20 images of the DRIVE set. An overall performance ratio of 92% is achieved.

This paper introduces the first ever appearance-based simulator of burst illumination laser sequences from a single, conventional video image of a vehicle. The appearance-based approach allows us to dispose of the very complex physical models needed to achieve realism. The system uses a dictionary of 3-D, geometric object models and a dictionary of intensity-time profile examples. The latter were obtained from real images acquired for a number of different materials and surface orientations. To generate a synthetic time-gated sequence the user provides simply a single, conventional photograph of a vehicle in a desired orientation. The photograph is matched interactively to a database of 3-D geometric models of vehicles, estimating 3-D pose and approximate relative depths at all points. Depths are then used to simulate time-gating, and consequently to decide which object parts are imaged for every depth. Model surface orientation and material are used to index the example dictionary and assign an intensity-time profile to imaged pixels. Results indicate very promising performance.

This correspondence derives a formal link between temporally weighted frame differences, or disturbance fields, which carry limited information suitable for motion detection, and the optic flow (OF), which carries richer information on local image motion. We use this link to derive a novel, simple, near-recursive optic flow algorithm based on a recursive-filter formulation. Most quantities involved are computed recursively, using only data from the current and previous frame. We can limit expensive OF calculations to pixels where motion magnitude is sufficiently high using image differences which the algorithm computes anyway. Experimental results with well-known synthetic, ground-truthed test sequences and standard performance metrics indicate good quantitative performance. Tests with real sequences suggest similar or better performance than a well-known, similar algorithm due to Lucas and Kanade (LK).

This paper present a novel, robust approach to the automatic location of the optic disk in retinal (fundus) images. Instead of generating a single, high-confidence optic disk candidate, we generate sets of plausible candidates for optic disk, macula and main vessels, then search the space of all possible triplets (optic disk, macula, vessels) to identify the one satisfying a-priori anatomical constraints. Our first implementation achieved 100% success with 40 wide-field-of-view retinal images acquired by an OPTOS Panorama ophtalmoscope. It also matched the performance of a visible, recently reported algorithm (A. Hoover et al, IEEE Trans. Medical Imaging, vol.22, no.8, p.951-8, 2003) on the STARE test set, and succeeded in some cases where the aforementioned algorithm failed.

This survey paper discusses the three-dimensional image processing challenges posed by present and future immersive telecommunications, especially immersive video conferencing and television. We introduce the concepts of presence, immersion, and co-presence and discuss their relation to virtual collaborative environments in the context of communications. Several examples are used to illustrate the current state of the art. We highlight the crucial need of real-time, highly realistic video with adaptive viewpoint for future immersive communications and identify calibration, multiple-view analysis, tracking, and view synthesis as the fundamental image-processing modules addressing such a need. For each topic, we sketch the basic problem and representative solutions from the image processing literature.

We present a system detecting the presence of unconstrained man-made objects in unconstrained subsea videos. Classification is based on contours, which are reasonably stable features in underwater imagery. First, the system determines automatically an optimal scale for contour extraction by optimising a quality metric. Second, a two-feature Bayesian classifier determines whether the image contains man-made objects. The features used capture general properties of man-made structures using measures inspired by perceptual organisation. The system classified correctly approximately 85% of 1390 test images from five different underwater videos, in spite of the varying image contents, poor quality and generality of the classification task.

We derive a formal link between temporally weighted frame differences, or disturbance fields, which carry limited information commonly used for motion detection, and the optic flow, which carries rich information on local image motion. We use this to formulate a novel, simple, near-recursive optic flow algorithm based on a recursive-filter formulation. Most quantities involved are computed recursively, using only data from the current and previous frame. Experimental results with well-known synthetic, ground-truthed test sequences and standard performance metrics indicate good quantitative performance. Test with real sequences suggest similar or better performance than a similar algorithm.

This paper introduces a 2D visual servoing technique for the station keeping of an unmanned underwater vehicle (UUV) with respect to planar targets on the sea bed. The underwater vehicle is subject to sea current disturbances which make it drift from its desired position. Feature points from unmarked objects are extracted and tracked with a sparse feature tracker developed in-house. The scene depth is estimated from a planar homography. To validate our approach, we emulate the dynamics of the surge and the sway degrees-of-freedom of an UUV model with a planar Cartesian robot in our water test tank. Successful station keeping experiments obtained with a range of sea current disturbances are presented.

All the methods for estimating the fundamental matrix do not naturally exploit the rank-2 constraint. For this reason few rank-2 parameterizations of the fundamental matrix have been proposed over the years. In general they can be an over parameterization (12 parameters) being generally valid, or use a minimal set of parameters (eight) but do not cover all the rank-2 matrices. We propose a rank-2 parameterization which uses only 9 parameters, one more of the minimal parameterizations, and covers all the rank-2 matrices.

We present a new, efficient stereo algorithm addressing robust disparity estimation in the presence of occlusions. The algorithm is an adaptive, multiwindow scheme using left–right consistency to compute disparity and its associated uncertainty. We demonstrate and discuss performances with both synthetic and real stereo pairs, and show how our results improve on those of closely related techniques for both accuracy and efficiency.

We present a novel algorithm performing projective rectification which does not require explicit computation of the epipolar geometry and specifically of the fundamental matrix. Instead of finding the epipoles and computing two homographies mapping the epipoles to infinity, as done in recent work on projective rectification, we exploit the fact that the fundamental matrix of a pair of rectified images has a particular, known form. This allows us to set up a minimization that yields the rectifying, homographies directly from image correspondences. Experimental results show that our method works quite robustly even in the presence of noise, and with inaccurate point correspondences. The code of our implementation will be made available at the author's web site.

We investigate the automatic estimation of fish weight from sets of morphometric measurements. Our solution combines a vision system with a robust regression method, the Support Vector Machine (SVM). Measurements are taken automatically from two binarised views of each fish in a training sample, then fed to a quadratic SVM along with approximate weight estimates. The SVM learns the law linking weight to shape directly (without computing volume) and compensates for several inaccuracies in the training measurements. We suggest a methodology identifying optimal shape measurements for the task, and report results obtained with a sample of 99 trouts between 300 and 600g, showing good accuracy and reliability, and better performance with respect to length-weight relations adopted commonly in fisheries science. 1 Introduction This work explores a new way of estimating fish weight from shape using computer vision. The relation between weight and shape is important both for fish biology [4, 5...

This paper addresses robust feature tracking. We extend the well-known Shi-Tomasi-Kanade tracker by introducing an automatic scheme for rejecting spurious features. We employ a simple and efficient outlier rejection rule, called X84, and prove that its theoretical assumptions are satisfied in the feature tracking scenario. Experiments with real and synthetic images confirm that our algorithm makes good features track better; we show a quantitative example of the benefits introduced by the algorithm for the case of fundamental matrix estimation. The complete code of the robust tracker is available via ftp.

This paper presents a novel method for determining the location of the instantaneous epipole in a sequence of images acquired by an uncalibrated camera and containing a single, rigid motion (e.g., the camera moves in a static environment). The method uses the full perspective camera model and requires the estimation of the optical flow at a minimum of six image locations. The key observation is that the optical flow equations can be written in terms of the epipole in a strikingly simple form if the translational and rotational flow components are not separated as done usually. The epipole location can then be obtained as the minimum of a least-square residual function associated to the computed optical flow. We report and discuss initial experiments on both synthetic and real data and illustrate possible developments of this method towards the use of uncalibrated optical flow for 3-D motion and structure reconstruction.

We present a new, efficient stereo algorithm addressing robust disparity estimation in the presence of occlusions. The algorithm is an adaptive, multi-window scheme using left-right consistency to compute disparity and its associated uncertainty. We demonstrate and discuss performances with both synthetic and real stereo pairs, and show how our results improve on those of closely related techniques for both robustness and efficiency.

We report a system for sensor planning, GASP, which is used to compute the optimal positions for inspection tasks using known imaging sensors and feature-based object models. GASP (general automatic sensor planning) uses a feature inspection representation (the FIR), which contains the explicit solution for the simplest sensor positioning problem. The FIR is generated off-line, and is exploited by GASP to compute on-line plans for more complex tasks, called inspection scripts. Viewpoint optimality is defined as a function of feature visibility and measurement reliability. Visibility is computed using an approximate model. Reliability of inspection depends on both the physical sensors acquiring the images and on the processing software; therefore we include both these components in a generalized sensor model. These predictions are based on experimental, quantitative assessment. We show how these are computed for a real generalized sensor, which includes a 3-D range imaging system, and software performing robust outlier removal, surface segmentation, object location and surface fitting. Finally, we demonstrate a complete inspection session involving 3-D object positioning, planning optimal position inspection, and feature measurement from the optimal viewpoint.

This paper presents an analytical study and several practical results for computing stable reconstructions with uncalibrated, motion-based inspection and conveyor-belt installations. We achieve metric reconstruction with a simple, efficient algorithm and two nonlinear constraints expressing knowledge readily accessible in a real setup. We analyse the stability and accuracy of reconstruction with respect to the system’s mathematical structure, pixelisation, image noise, and constraint values. Extensive experiments with simulated and real data have confirmed our analysis in full, and one example is illustrated here. 1 Introduction and related work Motion-based reconstruction is a much-studied problem in computer vision (see [1, 8] for extensive reviews) which has not yet come into its own in inspection applications. However, motion-based algorithms capable of reasonably accurate 3-D measurements, requiring no calibration, and using a single off-the-shelf camera

We analyse experimentally some subpixel-accuracy disparity and uncertainty estimators based on the SSD method, frequently used in stereo and motion analysis. We identify key inadequacies, and introduce new, practical algorithms. We discuss results and performance tests, and demonstrate the effective use of the new estimators in a complete, working system reconstructing dense depth maps using dynamic stereo. The system achieves good accuracy (average percentage errors smaller than few percents) and reliable uncertainty discrimination without any expensive smoothing or regularisation of disparity maps, adopted commonly. 1 Introduction This paper analyses the estimation of subpixel-accuracy disparity and its uncertainty in the context of dynamic stereo for computing dense depth maps [3, 4, 7, 9, 10]. Careful solutions to this problem, supported by adequate solutions to resampling and drop ins/outs, allow us to achieve good reconstructions without expensive smoothing or regulari...

We use polarization analysis in triangulation-based laser scanners to disambiguate the true laser stripe from spurious inter-reflections caused by holes and concavities on metal surfaces. Stripe candidates are discriminated by projecting linearly polarised laser light and measuring the polarization state of the linearly polarized component of the observed stripe candidates. Initial experimental results are reported and discussed.

We report a novel application of polarization based vision addressing the robustness of laser triangulation range sensors. Such sensors are based on the accurate detection of a pattern of laser light projected onto a scene, usually a point or line. Typical problems arise with highly specularly reflective surfaces, which can generate visible reflections of the light in various parts of the image. This can confuse the detection algorithms and lead to wrong range measurements. This paper demonstrates experimentally the feasibility of polarization based vision for disambiguating multiple specular inter reflections of the laser light. We concentrate on metal components as they have high interest for inspection in manufacturing, and show positive results with situations of various complexities.>

This paper focuses on the experimental evaluation of a range image segmentation system which partitions range data into homogeneous surface patches using estimates of the sign of the mean and Gaussian curvatures. The authors report the results of an extensive testing program aimed at investigating the behavior of important experimental parameters such as the probability of correct classification and the accuracy of curvature estimates, measured over variations of significant segmentation variables. Evaluation methods in computer vision are often unstructured and subjective: this paper contributes a useful example of extensive experimental assessment of surface-based range segmentation.>

This paper addresses two aspects of triangulation-based range sensors using structured laser light: calibration and measurements consistency. We present a direct calibration technique which does not require modelling any specific sensor component or phenomena, therefore is not limited in accuracy by the inability to model error sources. We also sketch some consistency tests based on two-camera geometry which make it possible to acquire satisfactory range images of highly reflective surfaces with holes. Experimental results indicating the validity of the methods are reported.

We examine the use of viewer-centred representation s (VCRs) for object recognition, location and inspection, employed in the framework of advanced inspection systems. We introduce a recognitionoriented VCR of 3D objects and show how to compute it using cluster analysis of approximate visibility spaces. Using the same representational framework, we illustrate the automatic generation of VCRs for feature-based inspection of 3D objects.

Addresses two aspects of triangulation-based range sensors using structured laser light: calibration and measurement consistency. We present a direct calibration technique which does not require the modelling of any specific sensor component or phenomena, so is not limited in accuracy by the inability to model error sources. We also introduce some consistency tests based on two-camera geometry which make it possible to acquire satisfactory range images of highly reflective surfaces with holes. Experimental results indicating the validity of the methods are reported.>

Several boundary treatments for attenuating shape deformation introduced by Gaussian smoothing are discussed. The author models Gaussian smoothing by a diffusion equation, a general mathematical framework particularly useful for scale-space analysis. An adaptive, shape-reserving boundary condition for diffusion smoothing range images is introduced. It is shown that this condition is more general than similar techniques found in the literature.>

By functional reasoning we mean the ability of integrating shape, function, and plans in reasoning. the shape of many man-made objects, such as tools, is expressly designed to provide precise functionalities. Moreover, humans know how to use the same objects for different functions. In vision and reasoning we make use of complex information which is not exclusively based on geometric and spatial knowledge, but also on functional elements. They seem to play a role in object recognition and representation. This article is an insight in functional reasoning from the computational point of view. It introduces its concepts and its apparent ubiquity in human behavior. Some relevant computational literature is reported and discussed. the rest of the article is an outline of the FUR project, an attempt to develop a computational model for functional reasoning. the development state of the project is presented along with the implementation of a first prototype. Some experimental results are finally given.