Intracranial aneurysms (ICA) commonly occur in specific segments of the Circle of Willis (CoW), primarily, onto thirteen major arterial bifurcations. An accurate detection of these critical landmarks is necessary for a prompt and efficient diagnosis. We introduce a fully automated landmark detection approach for CoW bifurcations using a two-step neural networks process. Initially, an object detection network identifies regions of interest (ROIs) proximal to the landmark locations. Subsequently, a modified U-Net with deep supervision is exploited to accurately locate the bifurcations. This two-step method reduces various problems, such as the missed detections caused by two landmarks being close to each other and having similar visual characteristics, especially when processing the complete MRA Time-of-Flight (TOF). Additionally, it accounts for the anatomical variability of the CoW, which affects the number of detectable landmarks per scan. We assessed the effectiveness of our approach using two cerebral MRA datasets: our In-House dataset which had varying numbers of landmarks, and a public dataset with standardized landmark configuration. Our experimental results demonstrate that our method achieves the highest level of performance on a bifurcation detection task.

We hereby present a full synthetic model, able to mimic the various constituents of the cerebral vascular tree, including the cerebral arteries, bifurcations and intracranial aneurysms. This model intends to provide a substantial dataset of brain arteries which could be used by a 3D convolutional neural network to efficiently detect Intra-Cranial Aneurysms. The cerebral aneurysms most often occur on a particular structure of the vascular tree named the Circle of Willis. Various studies have been conducted to detect and monitor the aneurysms and those based on Deep Learning achieve the best performance. Specifically, in this work, we propose a full synthetic 3D model able to mimic the brain vasculature as acquired by Magnetic Resonance Angiography, Time Of Flight principle. Among the various MRI modalities, this latter allows for a good rendering of the blood vessels and is non-invasive. Our model has been designed to simultaneously mimic the arteries' geometry, the aneurysm shape, and the background noise. The vascular tree geometry is modeled thanks to an interpolation with 3D Spline functions, and the statistical properties of the background noise is collected from angiography acquisitions and reproduced within the model. In this work, we thoroughly describe the synthetic vasculature model, we build up a neural network designed for aneurysm segmentation and detection, finally, we carry out an in-depth evaluation of the performance gap gained thanks to the synthetic model data augmentation.

In literature about selective encryption of image and video content, image quality indices are usually used to gauge the degree of encryption. These methods have frequently been shown not to work well for the evaluation of encryption, mainly due to them being trained on predominantly high quality contents. The problem for creating a proper recognition index or visual encryption strength index is the lack of data to train on. In this paper we present the first database of encrypted video content, ranging from high quality to completely unrecognizable, together with human observer scores for quality and recognizability. We also provide a basic evaluation of visual quality indices on this database, directly and in different combination by fusion, to showcase that currently image and video quality indices are ill fit for the purpose of estimating video encryption strength/recognizably. This inability of quality indices to perform also showcases that this database fills the required blind spot of currently available data.

An accurate detection of intracranial aneurysms is of paramount importance for a timely diagnosis and a possible treatment. Indeed, intracranial aneurysms (ICA) need to be detected at an early stage, and their evolution must be closely monitored before any treatment becomes hazardous. Numerous methods have been proposed to detect ICA either on Digital Subtraction Angiography (DSA) on Computed Tomography Angiography (CTA), or Magnetic Resonance Angiography (MRA) Time-Of-Flight (TOF) modalities. In the present study, we are particularly interested in the saccular ICA occurring onto the vascular tree's bifurcations, and we specifically focus our research on MRA-TOF acquisitions. We propose a synthetic model of both the artery bifurcation and the aneurysm itself. We are able to very accurately model some vasculature bifurcations as they are represented on TOF acquisitions. Their geometrical disposition, the various background noises and the aneurysm's shapes and positions are rigorously reproduced. The purpose of this approach is to alleviate the burden of a ground-truth manual segmentation commonly required when using deep-learning for object detection or semantic segmentation. Our model is highly configurable and intends to produce vast datasets used to feed a Convolutional Neural Network (CNN) for the automatic detection and segmentation of the saccular ICAs. In this preliminary study we only intend to propose a model for 3D aneurysm-bearing bifurcations. Evidently, a thorough evaluation of the model's accuracy is conducted. A preliminary experiment was conducted on a reduced dataset in order to assess the applicability of our bifurcation model. In future works, we will enhance the bifurcation model and propose an in-depth evaluation via Deep Learning methods.

In order to build an intelligent dental care process that both facilitates the treatment and improves the diagnosis, an accurate tooth segmentation and recognition on panoramic X-ray images might prove helpful. Although many studies have been conducted on teeth segmentation, few methods allow to perform tooth recognition and numbering at the same time. The existing methods allowing both those processes rely on instance segmentation architectures. To fill some gaps in the area of dental image segmentation, we propose a novel approach of automatic joint teeth segmentation and numbering using the pioneer U- Net model. We are first to employ the conventional U-Net model and show its limitations to provide accurate segmentation, being affected by noisy pixels outside the teeth region and by missing teeth in the X-ray images. To overcome this problem and reduce the misclassifications, we use a bounding box prior at the level of the skip connections. Such an approach helps guiding the network to better locate the teeth, and hence improves the segmentation. To validate the effectiveness of the method, we have conducted two experiments on the DNS Panoramic Dataset: a first one using manual bounding boxes and another one relying on a preliminary step of object detection. The implemented networks were evaluated using the Dice coefficient index and our results showed that consideration of location information onto the skip connections improves the performances of the semantic segmentation by 5% to 10% in average Dice accuracy depending on the quality of the bounding box labels.

Assessment of visual encryption of video and image content requires a reliable estimation of content recognizability and low quality. As pointed out in the literature, current methods are insufficient and research into this topic, as well as into the relation between low quality and recognizability, is still lacking. This lack of research is primarily due to a lack of data. To improve on the status-quo we have taken a recognizability database and performed a subjective quality evaluation on a subset of the images. This gives us a new database with both subjective recognizability and quality information and allows to delve into the relation between low quality and recognizability. We analyze the relationship between quality and recognizability as well as the predictive quality of state of the art visual quality indices. We show that the visual quality indices are poor indicators for the estimation of recognizability. Furthermore, we show that they must be a poor fit because of the disparity between two distinct perceptual tasks: quality and recognizability.

The assessment of very low quality visual data is known to be difficult. In particular, the ability of humans to recognize encrypted visual data is currently impossible to determine computationally. The human vision research community has widely studied some particular topics, such as image quality assessment or the determination of a visibility threshold, while others are still barely researched, specifically visual content recognition. To this day, there does not exist a reliable recognition index that can be employed for such tasks. In order to enable the study of human image content recognition, and in an attempt to propose a corresponding recognizability index, we build a dataset of selectively encrypted images together with subjective ground-truth about their human intelligibility. The methods of acquisition, setup, protocol, outlier detection, are described and we suggest how to calculate a recognition score as well as a recognition threshold. The performance of traditional visual quality indices to predict human visual content recognition is assessed on these data and found to be inapt to estimate recognition of visual content. Contrasting, structure based recognition indices as proposed for this task are shown to represent a promising starting point for further research. To facilitate the creation of a recognition index and to foster further research into human visual content recognition and its relation to the human visual system we will make the database publicly available.

In an era when the protection of intellectual property rights becomes more and more important, providing robust and efficient watermarking techniques is crucial, both in terms of embedding and detection. In this paper, the authors specifically focus on the latter stage. Most often, the detection consists in the comparison of a fixed and non-adaptive decision threshold to a correlation coefficient. This threshold is usually determined either theoretically or experimentally. Here, it is proposed to apply Grubbs? test, a simple statistical test for outliers, on the correlation data in order to take a binary decision about the presence or the absence of the searched watermark. The proposed technique is applied to three algorithms from the literature: the correlation data generated by the detector is fed to Grubbs' test. The obtained results show that Grubbs' test is efficient, robust and reliable. Above all, it automatically adapts to the searched watermark and can be easily applied to most types of watermarking approaches.

More than ever, the growing amount of exchanged digital content calls for efficient and practical techniques to protect intellectual property rights. During the past two decades, watermarking techniques have been proposed to embed and detect information within these contents, with four key requirements at hand: robustness, security, capacity, and invisibility. So far, researchers mostly focused on the first three, but seldom addressed the invisibility from a perceptual perspective and instead mostly relied on objective quality metrics. In this paper, a novel DFT watermarking scheme featuring perceptually optimal visibility versus robustness is proposed. The watermark, a noise-like square patch of coefficients, is embedded by substitution within the Fourier domain; the amplitude component adjusts the watermark strength, and the phase component holds the information. A perceptual model of the human visual system (HVS) based on the contrast sensitivity function (CSF) and a local contrast pooling is used to determine the optimal strength at which the mark reaches the visibility threshold. A novel blind detection method is proposed to assess the presence of the watermark. The proposed approach exhibits high robustness to various kinds of attacks, including geometrical distortions. Experimental results show that the robustness of the proposed method is globally slightly better than state-of-the-art. A comparative study was conducted at the visibility threshold (from subjective data) and showed that the obtained performances are more stable across various kinds of content.

In this work we propose a novel non-blind H.264/CAVLC structure-preserving substitution watermarking algorithm. The proposed watermarking algorithm enables extremely efficient watermark embedding by simple bit substitutions (substitution watermarking). The bit-substitutions change the motion vector differences of non-reference frames. Furthermore our watermarking algorithm can be applied in applications scenarios which require that watermarking preserves the length of the bitstream units (structure-preserving watermarking). The watermark detection works in the image domain and thus is robust to video format changes. The quality and robustness of the approach are in depth evaluated and analyzed, the quality evaluation is backed up by subjective evaluations. Comparison to the state-of-the-art indicates a superior performance of our watermarking algorithm.

The present work explores the use of a specific method, called non linear hybrid watermarking to watermark High Dynamic Range (HDR) images. The non linear hybrid technique combines both additive and multiplicative watermark embedding and is based on a square root embedding equation. We evaluate the robustness and objective quality of HDR tone mapped watermarked images on two different HDR databases. The experimentations show that the watermark is imperceptible and successfully survives tone mapping attack.

Ensuring watermark invisibility in digital images is a challenging task. Most watermarking techniques empirically adjust a strength parameter in order to reach the best trade-off between invisibility and robustness. A target PSNR value is typically set in order to reach a defined quality level. Some watermarking techniques exploit local activity to increase the watermark strength in some specific images areas (edges, textures). In this work we study the visibility thresholds for wavelet domain multiplicative embedding using a watermarking oriented subjective experiment protocol. Thirty four observers were enrolled for a subjective experiment, and had to adjust the watermark strength in order to best define the visibility threshold for watermarking distortions. Three watermark embedding equations were tested in various wavelet sub-bands, and the optimal equation maximizing robustness was derived for every sub-band.

Knowing that the watermarking community use simple statistical quality metrics in order to evaluate the watermarked image quality, the authors have recently proposed a simplified objective quality metric (OQM), called “CPA”, for watermarking applications. The metric used the contrast sensitivity function, along with an adapted error pooling, and proved to perform better than state-of-the-art OQMs. In this work, we intend to improve the performance of the CPA metric. The new metric includes the most important steps of Human Visual System (HVS) based quality metric, namely spatial frequency consideration and masking effects. Besides, this work goes further than classical image quality assessment, and several objective quality metrics will be tested in a watermarking algorithm comparison scenario. We will show that the proposed metric is both able to accurately predict the observers score in a quality assessment task, and is also able to compare watermarking algorithms altogether on a perceptual quality viewpoint.

Quality assessment is of major importance when designing and testing an image/video coding technique. Compression performances are usually evaluated by means of rate-distortion curves. However, the PSNR is commonly employed as the distortion measure. We hereby present a full quality assessment benchmark for the LAR (locally adaptive resolution) coder. We conducted a subjective experiment, where nineteen observers were asked to assess the perceptual quality of LAR coded images under normalized viewing conditions. Furthermore, five objective quality assessment metrics were used in order to determine the most suitable metric for the LAR coder. Finally, both JPEG and JPEG200 images were generated and assessed during the subjective experiment in order to define the optimal quality metric which should be used when comparing the codecs' output images quality.

A novel ordering of digital Radon projections coefficients is presented here that enables progressive image reconstruction from low resolution to full resolution. The digital Radon transform applied here is the Mojette transform first defined by Guedon et al. (1995). The Mojette transform is a natural way to generate redundancy to any specified degree and has been demonstrated to be useful for redundant representation for robust data storage and transmission. Combining this with the wavelet transform facilitates compression, i.e., joint source-channel coding, along with the additional property of scalability.

The interest in objective quality assessment have significantly increased over the past decades. Several objective quality metrics have been proposed and made publicly available, moreover, several subjective quality assessment databases are distributed in order to evaluate and compare the metrics. However, several question arises: are the objective metrics behaviours constant across databases, contents and distortions? how significantly the subjective scores might fluctuate on different displays (i.e. CRT or LCD)? which objective quality metric might best evaluate a given distortion? In this article, we analyse the behaviour of four objective quality metrics (including PSNR) tested on three image databases. We demonstrate that the performances of the quality metrics can strongly fluctuate depending on the database used for testing. We also show the consistency of all metrics for two distinct displays.

One of the most challenging issues for watermarkers is to tune the strength of their watermark embedding. The strength is usually an alpha parameter which is increased until a reasonable trade-off between invisibility and robustness is achieved. The watermarking community needs efficient just noticeable difference (JND) masks to optimally embed the watermarks. The Fourier transform is particularly adapted to the human visual system (HVS) modeling. In this work, we evaluate the usability of the JND mask in the wavelet domain. The use of the mask in the DWT domain involves some approximations. We will see here that the HVS decomposition and the wavelet decomposition do not perfectly fit altogether. The efficiency of the so obtained mask is tested both in terms of invisibility and robustness.

This paper presents a new approach to measure texture directions and estimate illumination tilt angle of 3D surface textures by using mojette transform. Feature vectors are generated from variances of 72 mojette transform projections with different projection angles. The measured texture directions are compared with human perceptual judgement. Furthermore, we estimate illumination tilt angles by minimizing the Euclidean distance of the feature vector between the test image and the training sets. Experimental results show the effectiveness and accuracy of our proposed approach.

In this paper we propose a new joint encryption and lossless compression technique designed for large images. The proposed technique takes advantage of the Mojette transform properties, and can easily be included in a distributed storage architecture. The basic crypto-compression scheme presented is based on a cascade of Radon projection which enables fast encryption of a large amount of digital data. Standard encryption techniques, such as AES, DES, 3DES, or IDEA can be applied to encrypt very small percentages of high resolution images. As the proposed scheme uses standard encryption, and only transmits uncorrelated data along with the encrypted part, this technique takes benefit of the security related to the chosen encryption standard, here, we assess its performances in terms of processing time and compression ratio.

This article presents a new lossless compression algorithm based on a finite Radon transform. The baseline of this work consist in encoding the difference between several similar finite Radon projections. Two differential encoding techniques are needed to efficiently encode the Radon transformed data. An intra-projection encoding technique is first performed, this differential encoding process takes benefit of the redundancy present in such projections. The second step consists in encoding the similarities between Radon projections, this is the inter-projection encoding procedure. The used finite Radon transform (called Mojette transform), allows to ensure both compression and data redundancy. The main novelty of this work is the use of a secure distributed storage or transmission tool in a lossless compression context

This paper describes a new kind of use for secure image transmission. A stream encryption method is presented, in which a key allows the authorized users to recover the original image. Our algorithm exploits the redundancy properties of the Mojette transform. Anyone whom knows the appropriate key will be able to decode the original image whereas only a scrambled image can be decoded without this key. This data access scheme is moreover perfectly suited for a multiple description context. This method takes into account the regions of interest (ROI), according to the received data, the image decoding is first performed in order to recover these ROIs.