Olivier Lezoray

Université de Caen Normandie | UNICAEN · Groupe de Recherche en Informatique, Automatique, Image et Instrumentation (GREYC)

In this paper, we consider a biologically inspired spiking neural network model for motion detection. The proposed model simulates the neurons’ behavior in the cortical area MT to detect different kinds of motion in image sequences. We choose the conductance-based neuron model of the Hodgkin–Huxley to define MT cell responses. Based on the center-s...

Skin lesion is one of the most critical challenges nowadays due to the difficulty of distinguishing a benign lesion from a malignant one. Melanoma represents a malignant melanocytic type of cancer among the most dangerous ones. In contrast, basal cell carcinoma and squamous cell carcinoma represent no malignant melanocytic types of cancer that thre...

Deep CNNs have recently led to new standards in all fields of computer vision with specialized architectures for most challenges, including Video Object Segmentation and Pose Tracking. We extend Space-Time Memory Networks for the simultaneous detection of multiple object parts. This enables the detection of human body parts for multiple persons in...

With the advent of whole slide image (WSI) scanners, pathology is undergoing a digital revolution. Simultaneously, with the development of image analysis algorithms based on artificial intelligence tools, the application of computerized WSI analysis can now be expected. However, transferring such tools into clinical practice is very challenging as...

For various forms of skin lesion, many different feature extraction methods have been investigated so far. Indeed, feature extraction is a crucial step in machine learning processes. In general, we can distinct handcrafted and deep learning features. In this paper, we investigate the efficiency of using 17 commonly pre-trained convolutional neural...

In the recent era, deep learning has become a crucial technique for the detection of various forms of skin lesions. Indeed, Convolutional neural networks (CNN) have became the state-of-the-art choice for feature extraction. In this paper, we investigate the efficiency of three state-of-the-art pre-trained convolutional neural networks (CNN) archite...

Personalized medicine exploits the patient data, for example, genetic compositions, and key biomarkers. During the data mining process, the key challenges are the information loss, the data types heterogeneity and the time series representation. In this paper, a novel data representation model for personalized medicine is proposed in light of these...

Le défi de la segmentation d’instance a été exploré principalement avec des images rectilinéaires. Cependant, la segmentation d’instance dans des images fisheye implique des difficultés supplémentaires et n’a pas encore été complètement explorée. Un modèle CNN capable de fonctionner aussi bien sur des données rectilinéaires que sur des données fish...

The extension of mathematical morphology to multivariate data has been an active research topic in recent years. In this paper we propose an approach that relies on the consensus combination of several stochastic permutation orderings. The latter are obtained by searching for a smooth shortest path on a graph representing an image. This path is obt...

The emergence of personalized medicine and its exceptional advancements reveal new needs regarding the availability of adequate medical decision-making models. Considering detailed data on this medicine, the creation of a medical decision-making system may encounter many inhibitory factors, such as data representation, data reduction, data classifi...

Deep learning has recently improved the performance of Speaker Identification (SI) systems. Promising results have been obtained with Convolutional Neural Networks (CNNs).This success are mostly driven by the advent of large datasets.However in the context of commercial applications, collection of large amount of training data is not always possibl...

In this paper, we present a video-based emotion recognition neural network operating on three dimensions. We show that 3D convolutional neural networks (3D-CNN) can be very good for predicting facial emotions that are expressed over a sequence of frames. We optimize the 3D-CNN architecture through hyper-parameters search, and prove that this has a...

In this paper, we present a full-reference quality assessment metric based on the information of visual saliency. The saliency information is provided under the form of degrees associated to each vertex of the surface mesh. From these degrees, statistical attributes reflecting the structures of the reference and distorted meshes are computed. These...

The authors address the problem of editing signals such as 2D colour images or 3D coloured meshes that are represented under the general framework of graph signals. As state‐of‐the‐art editing approaches decompose an image into several layers in order to manipulate them, they propose a hierarchical multi‐layer decomposition of graph signals that re...

In this paper, we propose a new hand gesture recognition method based on skeletal data by learning SPD matrices with neural networks. We model the hand skeleton as a graph and introduce a neural network for SPD matrix learning, taking as input the 3D coordinates of hand joints. The proposed network is based on two newly designed layers that transfo...

This paper proposes a new neural network based on SPD manifold learning for skeleton-based hand gesture recognition. Given the stream of hand's joint positions, our approach combines two aggregation processes on respectively spatial and temporal domains. The pipeline of our network architecture consists in three main stages. The first stage is base...

The graph Laplacian plays an important role in describing the structure of a graph signal from weights that measure the similarity between the vertices of the graph. In the literature, three definitions of the graph Laplacian have been considered for undirected graphs: the combinatorial, the normalized and the random-walk Laplacians. Moreover, a no...

The problem of interactive contour extraction of targeted objects of interest in images is challenging and finds many applications in image editing tasks. Several methods have been proposed to address this problem with a common objective: performing an accurate contour extraction with minimum user effort. For minimal paths techniques, achieving thi...

Convolutional neural networks (CNN) have deeply impacted the field of machine learning. These networks, designed to process objects with a fixed topology, can readily be applied to images, videos and sounds but cannot be easily extended to structures with an arbitrary topology such as graphs. Examples of applications of machine learning to graphs i...

Many approaches for background subtraction and people detection have been developed so far. However, the best state-of-The-Art methods do not yet give satisfactory results in real transportation environments. Indeed, these latter configurations imply several difficulties such as fast brightness changes, noise, shadows, scrolling background, etc., a...

Many computer graphics applications use visual saliency information to guide their treatments such as adaptive compression, viewpoint-selection, segmentation, etc. However, all these applications rest on a partial estimation of visual saliency insofar that only geometric properties of the considered 3D mesh are taken into account leaving aside the...

This guest editorial introduces and summarizes the Special Section on Superpixels for Image Processing and Computer Vision.

Après le son, les images et les videos, les modèles 3D représentés par des maillages polygonaux constituent le contenu émergent actuel de part les avancées technologiques récentes dans le domaine de l’acquisition 3D. Les maillages 3D sont souvent amenés à subir plusieurs distorsions au cours de diverses étapes (de pré-traitement ou post-traitement)...

In this paper, we propose a spiking neural network model for edge detection in images. The proposed model is biologically inspired by the mechanisms employed by natural vision systems, more specifically by the biologically fulfilled function of simple cells of the human primary visual cortex that are selective for orientation. Several aspects are s...

After the sound, 2D images and videos, 3D models represented by polygonal meshes are the actual emergent content due to the technological advance in terms of 3D acquisition. 3D meshes can be subject to several degradations due to acquisition, compression, pre-treatment or transmission that distort the 3D mesh and therefore affect its visual renderi...

We present in this report a 3D colored mesh database established within the GREYC laboratory. 15 real objects with different colors, geometries and textures were acquired using the NextEngine 3D color laser scanner. On the 3D original colored meshes, we have applied several distorsions according to different strengths and situations (on rough or sm...

In many applications such as video surveillance or autonomous vehicles, people detection is a key element, often based on feature extraction and combined with supervised classification. Usually, output of these methods is in the form of a bounding-box containing an extracted people along with the background. But in specific application contexts, th...

We propose to tackle the problem of RGB-D image disocclusion inpainting when synthesizing new views of a scene by changing its viewpoint. Indeed, such a process creates holes both in depth and color images. First, we propose a novel algorithm to perform depth-map disocclusion inpainting. Our intuitive approach works particularly well for recovering...

We propose in this paper a novel perceptual viewpoint-independent metric for the quality assessment of 3D meshes. This full-reference objective metric relies on the method proposed by Wang et al. that compares the structural informations between an original signal and a distorted one. In order to extract the structural informations of a 3D mesh, we...

La détection de la saillance visuelle est une étape de pré-traitement importante pour plusieurs applications traitant des données 3D. Ce papier propose une nouvelle approche utilisant un descripteur local sous forme de patch de taille adaptative pour le calcul de la saillance visuelle des maillages 3D. Ce descripteur est utilisé pour une mesure de...

The generalization of mathematical morphology to multivariate vector spaces is addressed in this paper. The proposed approach is fully unsupervised and consists in learning a complete lattice from an image as a nonlinear bijective mapping, interpreted in the form of a learned rank transformation together with an ordering of vectors. This unsupervis...

We propose an exemplar-based video completion algorithm together with a geometry-guided space-time artifact reduction technique. The proposed completion algorithm is the video extension of an inpainting algorithm proven to be effective on still images. Then, the proposed space-time artifact reduction technique blends multiple patches, guided by a t...

In this paper we propose a depth-aided patch based inpainting method to perform the disocclusion of holes that appear when synthesizing virtual views from RGB-D scenes. Depth information is added to each key step of the classical patch-based algorithm from [Criminisi et al. 2004] to guide the synthesis of missing structures and textures. These cont...

In this paper, we present an approach for the segmentation of people silhouettes in images. Since in real-world images estimating pixel probabilities to belong to people or background is difficult, we propose to optimally combine several ones. A local window classifier based on SVMs with Histograms of Oriented Gradients features estimates probabili...

In this paper, a new formulation of patch-based adaptive mathematical morphology is addressed. In contrast to classical approaches, the shape of structuring elements is not modified but adaptivity is directly integrated into the definition of a patch-based complete lattice. The manifold of patches is learned with a nonlinear bijective mapping, inte...

Diffusion methods have proven their efficiency for tasks such as semi-supervised segmentation. The introduction of patches as a part of their speed function allows to deal with textured images. However, the computational burden with such variants stays too important for low-level tasks. In this paper, we propose a multivalued color-based potential...

In this paper we propose an approach to inpaint holes in depth maps that appear when synthesizing virtual views from a RGB-D scenes. Based on a superpixel oversegmentation of both the original and synthesized views, the proposed approach efficiently deals with many occlusion situations where most of previous approaches fail. The use of superpixels...

Mesh surface saliency detection is an important preprocessing step for many 3D applications. This paper proposes a novel saliency computation method by the use of a local vertex descriptor in the form an adaptive patch. This descriptor is used as a basis for similarity measurement and integrated into a weighted multi-scale saliency computation. Exp...

Nous proposons une méthode pour restaurer les trous d'une carte de profondeur qui apparaissent lors de la synthèse de vues virtuelles à partir de scènes RGB-D. Basée sur une sursegmentation en superpixels des vues originales et synthétisées, l'approche proposée gère efficacement de nombreuses occlusions où la plupart des approaches existantes échou...

Dans cet article nous proposons d'utiliser l'équation Eikonale sur graphes pour le partitionnement généralisé de données. Nous proposons une nouvelle fonction de potentielle qui favorise la création de partitions homogènes, ainsi qu'une méthode itérative permettant de placer astucieusement de nouveaux germes sur le graphe. L'application de ces cont...

Les méthodes de diffusion ont prouvé leur efficacité pour des tâches comme la segmentation semi-supervisée. L'intégration de patches dans le calcul de leur fonction de vitesse permet de traiter aisément des images texturées. Cependant le surcoût calculatoire reste trop important pour des tâches de bas niveau. Dans cet article, nous proposons une no...

Les méthodes d'inpainting par patch provoquent souvent des artefacts de bloc engendrés par les discontinuités visibles entre les morceaux de patchs recollés. Dans cet article nous proposons une méthode de mélange spatio-temporel de patchs réduisant ce type d'artefacts. Nous définissons un mod ele tensoriel guidant le mélange afin de mieux conserver...

Dans cet article nous proposons deux améliorations du célèbre inpainting par patch originalement formulé par Criminisi et al. [5]. Tout d'abord nous introduisons un terme de donnée basé sur les tenseurs de structures pour améliorer l'ordre de reconstruction. Ensuite, nous proposons une heuristique améliorant la recherche d'un bon patch candidat pou...

L’attention visuelle humaine est attirée par des régions spécifiques appartenant aux objets (pouvant être représentés par des maillages 3D). Celle-ci dépend fortement du degré de la saillance exposée par ces régions. Dans ce papier, nous proposons une nouvelle approche multi-échelle pour mesurer la saillance visuelle d’un objet 3D. Des descripteurs...

Despite the tremendous advances made in recent years, in the field of patch-based image inpainting algorithms, it is not uncommon to still get vis- ible artefacts in the parts of the images that have been resynthetized using this kind of methods. Mostly, these artifacts take the form of discontinuities between synthetized patches which have been co...

This paper concentrates on the use of Echo State Networks (ESNs), an effective form of reservoir computing, to improve microscopic cellular image segmentation. An ESN is a sparsely connected recurrent neural network in which most of the weights are fixed a priori to randomly chosen values. The only trainable weights are those of links connected to...

Our visual attention is attracted by specific areas into 3D objects (represented by meshes). This visual attention depends on the degree of saliency exposed by these areas. In this paper, we propose a novel multi-scale approach for detecting salient regions. To do so, we define a local surface descriptor based on patches of adaptive size and filled...

With the advance of three-dimensional (3-D) scanning technology, the cultural heritage community is increasingly interested in 3-D scans of cultural objects such as antiques, artifacts, and heritage sites. Digitization of these objects is commonly aimed at heritage preservation. Since 3-D color scanning has the potential to tackle a variety of trad...

This paper proposes a technical review of exemplar-based inpainting approaches with a particular focus on greedy methods. Several comparative and illustrative experiments are provided to deeply explore and enlighten these methods, and to have a better understanding on the state-of-the-art improvements of these approaches. From this analysis three i...

3D object shapes (represented by meshes) include both areas that attract the visual attention of human observers and others less or not attractive at all. This visual attention depends on the degree of saliency exposed by these areas. In this paper, we propose a technique for detecting salient regions in meshes. To do so, we define a local surface...

In this paper, we propose two major improvements to the exemplar-based image inpainting algorithm, initially formulated by Criminisi et al. [1]. First, we introduce a structure-tensor-based data-term for a better selection of pixel candidates to fill in based on priority. Then, we propose a new lookup heuristic in order to locate the best source pa...

Accurate skin lesion segmentation is critical for automated early skin cancer detection and diagnosis. We present a novel method to detect skin lesion borders in multispectral dermoscopy images. First, hairs are detected on infrared images and removed by inpainting visible spectrum images. Second, skin lesion is pre-segmented using a clustering of...