We propose 3D Geometric Mesh Network (3DGeoMeshNet), a method for 3D mesh reconstruction and completion. 3DGeoMeshNet is a novel Graph Convolutional Network (GCN)-based framework that leverages anisotropic convolution layers to effectively learn multi-scale global and local features directly in the spatial domain. Unlike traditional approaches that convert meshes into voxel grids or point clouds, our method operates directly on the polygonal mesh structure, preserving geometric fidelity throughout the learning and reconstruction process. In addition to 3D mesh reconstruction, we extend our framework to tackle 3D mesh completion task, where missing or incomplete regions of the mesh are first accurately recovered. The completion results are further refined through a set of pre- and post-processing steps. We extensively evaluate our approach on two benchmark datasets, COMA and DFAUST, and achieve SOTA results for 3D mesh reconstruction on both datasets. Additionally, our mesh completion experiments on the COMA dataset demonstrate the promising capability of 3DGeoMeshNet in recovering incomplete geometries. We further showcase the versatility of our method through additional applications, including mesh denoising, interpolation, and extrapolation, highlighting the robustness and generalization ability of our framework across various 3D mesh processing tasks.

With the advent of deep generative models, there has been some recent interest in the manipulation of people’s facial features. This has many potential applications in fashion and biometrics. However, it is a complex task. Indeed, a modification of a given attribute should not have any effect on the others, identity should be preserved, and image quality should not be altered. So far, the evaluation of the proposed methods has been mostly qualitative, which is insufficient to demonstrate progress and performance. We propose a comprehensive evaluation framework to estimate the quality of facial attribute editing methods with respect to several criteria: image quality, effective modification of the targeted attribute, level of entanglement between attributes and identity preservation. Three generative models are used to demonstrate the proposed evaluation framework over three datasets and three editing methods, resulting in the analysis of over 29k generated images.

3D meshes are fundamental data representations for capturing complex geometric shapes in computer vision and graphics applications. While Convolutional Neural Networks (CNNs) have excelled in structured data like images, extending them to irregular 3D meshes is challenging due to the non-Euclidean nature of the data. Graph Convolutional Networks (GCNs) offer a solution by applying convolutions to graph-structured data, but many existing methods rely on isotropic filters or spectral decomposition, limiting their ability to capture both local and global mesh features. In this paper, we introduce 3D Geometric Mesh Network (3DGeoMeshNet), a novel GCN-based framework that uses anisotropic convolution layers to effectively learn both global and local features directly in the spatial domain. Unlike previous approaches that convert meshes into intermediate representations like voxel grids or point clouds, our method preserves the original polygonal mesh format throughout the reconstruction process, enabling more accurate shape reconstruction. Our architecture features a multi-scale encoder-decoder structure, where separate global and local pathways capture both large-scale geometric structures and fine-grained local details. Extensive experiments on the COMA dataset containing human faces demonstrate the efficiency of 3DGeoMeshNet in terms of reconstruction accuracy.

We explore 3D mesh saliency detection as a nonlinear regression problem, leveraging geometric features derived from spiral patches. The saliency prediction involves a multi-scale analysis, utilizing roughness, geometric, and spectral saliencies, which are calculated through spiral patches at three levels of mesh decimation. These features are processed using a multi-layer perceptron to predict vertex-level saliency. Evaluation on the Schelling dataset demonstrates the approach’s efficacy, achieving competitive results compared to state-of-the-art methods, particularly in saliency prediction and extraction of keypoints. The method emphasizes the integration of advanced geometric descriptors with machine learning for enhanced 3D content analysis.

We propose SARMA (Saliency Analysis with a Residual Mesh-based Architecture), a graph neural network designed to predict visual saliency on 3D surface meshes. Unlike traditional methods that rely on handcrafted geometric features, SARMA learns saliency patterns in an end-to-end fashion using residual Feature-Steered Graph Convolution (FeaStConv) layers. The network takes a mesh as input, represented as a graph of vertices and edges, and outputs per-vertex saliency values. To capture perceptually relevant geometry, we enrich the input features with discrete mean curvature alongside 3D coordinates. The model consists of three FeaStConv layers, each followed by a residual connection that stabilizes training and mitigates oversmoothing. We evaluate SARMA on the Schelling dataset of 3D models annotated with human saliency points, using PLCC and AUC as evaluation metrics. Our approach outperforms prior handcrafted and deep learning-based methods in terms of correlation with ground truth saliency, demonstrating the effectiveness of residual graph-based architectures for perceptual analysis of 3D shapes.

The emergence of autonomous transportation systems marks a significant milestone in modern mobility, promising enhanced safety, efficiency, and convenience. However, ensuring the safety of passengers remains a paramount concern in the development and deployment of such systems. Monitoring the interior of autonomous vehicles has emerged as a critical aspect to guarantee passenger safety, requiring robust on-board action recognition systems. This paper provides an overview of the challenges and advancements in on-board action recognition for interior monitoring in autonomous vehicles. A comprehensive review of datasets pertinent to interior monitoring is presented, encompassing diverse scenarios and conditions to facilitate the training and evaluation of on-board action recognition models. Furthermore, we explore the methodologies employed in the development of these systems, including traditional computer vision techniques, deep learning architectures, and multimodal approaches . By synthesizing insights from existing research and highlighting key challenges and advancements, this paper aims to contribute to the ongoing discourse on enhancing safety measures in future autonomous transportation systems (bus, metro , train, car) through effective interior monitoring and action recognition technologies.

The advent of 3D data has revolutionized various industries, from architecture and engineering to healthcare and entertainment, enabling more precise simulations and realistic visualizations. However, 3D data is susceptible to noise and loss during generation and transmission, making quality assessment crucial for ensuring accuracy and usability. While existing literature addresses quality assessment for 3D point clouds and meshes separately, a gap exists in assessing the quality of 3D colored meshes due to the lack of reference datasets. This paper proposes an approach for No Reference 3D Colored Mesh Visual Quality Assessment (CMVQA), based on previous work related to quality assessment of 3D non colored meshes quality assessment. Our approach combines geometric and color features with spatial domain features extracted from mesh projections. Through extensive experiments and comparisons with full-reference metrics, including image quality metrics, our proposed approach demonstrates superior performance.

Abstract 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 are used by four comparisons functions genetically optimized that quantify the structure differences between a reference and a distorted mesh. We also present a statistical comparison study of six full-reference quality assessment metrics for 3D meshes. We compare the objective metrics results with humans subjective scores of quality considering the 3D meshes in one hand and the distorsion types in the other hand. Also, we show which metrics are statistically superior to their counterparts. For these comparisons we use the Spearman Rank Ordered Correlation Coefficient and the hypothetic test of Student (ttest). To attest the pertinence of the proposed approach, a comparison with a ground truth saliency and an application associated to the assessment of the visual rendering of smoothing algorithms are presented. Experimental results show that the proposed metric is very competitive with the state-of-the-art.

This paper proposes a new neural network based on Symmetric Positive Definite (SPD) manifold learning for real-time skeleton-based hand gesture recognition. The transformation of the input skeletal data into SPD matrices allows to encode efficiently high-order statistics such as covariances or correlations between the joints' features. These matrices are combined and transformed by our deep neural network which is thus constrained to work on the manifold of such matrices. The online recognition is performed using two sliding windows moving along the gesture's stream in order to simultaneously detect and classify the occurrence of a new gesture within the stream. The proposed network is validated on a challenging dataset and shows state-of-the-art performances both in terms of accuracy and inference time.

Deep learning has recently improved the performance of Speaker Identification (SI) systems. Promising results have been obtained with Convolutional Neural Networks (CNNs). This success is mostly driven by the advent of large datasets. However in the context of decentralized commercial applications, collection of large amount of training data is not always possible. In addition, robustness of a SI system is adversely effected by short utterances. Therefore, in this paper, we propose a novel text-independent speaker identification system able to identify speakers by learning from only few training short utterances examples. To achieve this, we combine a two-layer wavelet scattering network coupled with a CNN. The proposed architecture takes variable length speech segments. To evaluate the effectiveness of the proposed approach, Timit and Librispeech datasets are used in the experiments. Our experiments shows that our hybrid architecture provides satisfactory results under the constraints of short and limited number of utterances. These experiments also show that our hybrid architecture are competitive with the state of the art.

With the advent of data living on vertices of graphs, there is much interest in processing the so-called graph signals for partitioning tasks. As active contours have had much impact in the image processing community, their formulation on graphs is of importance to the field of graph signal processing. This paper proposes an adaptation on graphs of a model that combines the Geodesic Active Contour and the Active Contour Without Edges models. In addition, specific terms depending on graphs are introduced in the formulation. This adaptation is solved using a level set formulation with a gradient descent that can be expressed as a morphological front evolution process. Experimental results on different kinds of graphs signals show the benefit of the approach.

High-order statistics have been proven useful in the framework of Convolutional Neural Networks (CNN) for a variety of computer vision tasks. In this paper, we propose to exploit high-order statistics in the framework of Recurrent Neural Networks (RNN) for skeleton-based hand gesture recognition. Our method is based on the Statistical Recurrent Units (SRU), an un-gated architecture that has been introduced as an alternative model for Long-Short Term Memory (LSTM) and Gate Recurrent Unit (GRU). The SRU captures sequential information by generating recurrent statistics that depend on a context of previously seen data and by computing moving averages at different scales. The integration of high-order statistics in the SRU significantly improves the performance of the original one, resulting in a model that is competitive to state-of-the-art methods on the Dynamic Hand Gesture (DHG) dataset, and outperforms them on the First-Person Hand Action (FPHA) dataset.

In real world applications, the performances of speaker identification systems degrade due to the reduction of both the amount and the quality of speech utterance. For that particular purpose, we propose a speaker identification system where short utterances with few training examples are used for person identification. Therefore, only a very small amount of data involving a sentence of 2-4 seconds is used. To achieve this, we propose a novel raw waveform end-to-end convolutional neural network (CNN) for text-independent speaker identification. We use wavelet scattering transform as a fixed initialization of the first layers of a CNN network, and learn the remaining layers in a supervised manner. The conducted experiments show that our hybrid architecture combining wavelet scattering transform and CNN can successfully perform efficient feature extraction for a speaker identification, even with a small number of short duration training samples.

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 relies on morphological filtering. Since morphological filtering operators require a complete lattice, a dedicated approach for the morphological processing of vectorial data on graphs is used. By iterating the application of morphological filterings of decreasing sizes, the graph signal is decomposed into several detail layers, each capturing a given detail level. Editing applications such as abstraction, sharpness enhancement and tone mapping are shown to illustrate the benefits of the proposed approach.

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 based on a convolutional layer to increase the discriminative power of learned features. The second stage relies on different architectures for spatial and temporal Gaussian aggregation of joint features. The third stage learns a final SPD matrix from skeletal data. A new type of layer is proposed for the third stage, based on a variant of stochastic gradient descent on Stiefel manifolds. The proposed network is validated on two challenging datasets and shows state-of-the-art accuracies on both datasets.

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 transform a set of SPD matrices into a SPD matrix. For gesture recognition, we train a linear SVM classifier using features extracted from our network. Experimental results on a challenging dataset (Dynamic Hand Gesture dataset from the SHREC 2017 3D Shape Retrieval Contest) show that the proposed method outperforms state-of-the-art methods.

Editing of 3D colored meshes represents a fundamental component of nowadays computer vision and computer graphics applications. In this paper, we propose a framework based on the p-laplacian on directed graphs for structure-preserving filtering. This relies on a novel objective function composed of a fitting term, a smoothness term with a spatially-variant pTV norm, and a structure-preserving term. The last two terms can be related to formulations of the p-Laplacian on directed graphs. This enables to impose different forms of processing onto different graph areas for better smoothing quality.

We address the problem of sharpness enhancement of 3D colored meshes. The problem is modeled with graph signals and their morphological processing is considered. A hierarchical framework that decomposes the graph signal into several layers is introduced. It relies on morphological filtering of graph signal residuals at several scales. To have an efficient sharpness enhancement, the obtained layers are blended together with the use of a nonlinear sigmoid detail enhancement and tone manipulation, and of a structure mask.

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 the lost structures of the objects and to inpaint the depth-map in a geometrically plausible manner. Then, we propose a depth-guided patch-based inpainting method to fill-in the color image. Depth information coming from the reconstructed depth-map is added to each key step of the classical patch-based algorithm from Criminisi et al. in an intuitive manner. Relevant comparisons to the state-of-the-art inpainting methods for the disocclusion inpainting of both depth and color images are provided and illustrate the effectiveness of our proposed algorithms.

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. [1] 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 use a multi-scale visual saliency map on which we compute the local statistics. The experimental results attest the strong correlation between the objective scores provided by our metric and the human judgments. Also, comparisons with the state-of-the-art prove that our metric is very competitive.

Very few research works have addressed the problem of directly manipulating raw HDR vectors for general HDR image processing. In this paper a framework is proposed towards this aim and is based on a new representation of HDR images in the form of an ordering of vectors and an index image. This enables to formulate vector-preserving image processing methods dedicated to HDR images. The ordering relies on three steps: dictionary learning, manifold learning, and out of sample extension. The performance of the proposed approach is illustrated with innovative examples of HDR image filtering and enhancement.

This paper presents a framework for morphological processing of graph signals and investigates its usage for colored images and meshes editing tasks. The proposed method enables, with the help of the construction of a manifold-based ordering of color vectors, to define a new representation of graph signals in the form of an ordering of vectors and an index. The ordering relies on three steps: dictionary learning, manifold learning, and out of sample extension. This enables to formulate morphological operators for graphs signals and we demonstrate the performance of the proposed method on various colored image and mesh editing applications (simplification, abstraction, enhancement).

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 makes the algorithm more robust to inaccurate depth maps, while giving an efficient way to model the image. Extensive comparisons to relevant state-of-the-art methods show that our approach outperforms qualitatively and quantitavely these existing approaches.

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 function that partly alleviates this flaw. It allows to efficiently perform semi-supervised segmentation of natural and textured images.

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. Experimental results demonstrate that the proposed approach achieves competitive and innovative results, especially for 3d colored meshes.

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 improvements over Criminisi et al. algorithm are then presented and detailed: 1) a tensor-based data term for a better selection of pixel candidates to fill in; 2) a fast patch lookup strategy to ensure a better global coherence of the reconstruction; and 3) a novel fast anisotropic spatial blending algorithm that reduces typical block artifacts using tensor models. Relevant comparisons with the state-of-the-art inpainting methods are provided that exhibit the effectiveness of our contributions.

In this paper we propose to use the Eikonal equation on graphs for generalized data clustering. We introduce a new potential function that favors the creation of homogeneous clusters together with an iterative algorithm that place seeds vertices at smart locations. Oversegmentation application shows the effectiveness of our approach and gives results comparable to the state-of-the-art methods.

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 patches to copy/paste to these targeted points. These two contributions clearly make the inpainting algorithm reconstruct more geometrically coherent images, as well as speed up the process drastically. We illustrate the great performances of our approach compared to existing state-of-the-art methods.

In this paper we introduce a new unified framework for multi-scale detail manipulation of graph signals. The key to this unification is to model any kind of data as signals defined on appropriate weighted graphs. Graph signals are represented as the sum of successive layers, each capturing a given scale of detail. Detail layers are obtained through a series of regularization procedures based on total variation penalization over graphs. Layers are then processed separately before being recombined, thus achieving detail manipulation. The benefit of the approach is shown on images, 3D meshes and 3D colored point clouds.

We propose a new data model for graph-cut image segmentation, defined according to probabilities learned by a classification process. Unlike traditional graph-cut methods, the data model takes into account not only color but also texture and shape information. For melanoma images, we also introduce skin chromophore features and automatically derive “seed” pixels used to train the classifier from a coarse initial segmentation. On natural images, our method successfully segments objects having similar color but different texture. Its application to melanoma delineation compares favorably to manual delineation and related graph-cut segmentation methods.

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 a superpixel partition. Finally, the pre-segmentation is globally regularized at the superpixel level and locally regularized in a narrow band at the pixel level.

In this paper, we consider the adaptation of two Partial Differential Equations (PDEs) on weighted graphs, p-Laplacian and eikonal equations, for semi-supervised classification tasks. These equations are a discrete analogue of well known geometric PDEs, which are widely used in image processing. While the p-Laplacian on graphs was intensively used in data classification, few works relate to the eikonal equation for data classification. The methods are illustrated through semi-supervised classification tasks on databases, where we compare the two algorithms. The results show that these methods perform well regarding the state-of-the-art and are applicable to the task of semi-supervised classification.

Partial difference equations (PDEs) and variational methods for image processing on Euclidean domains spaces are very well established because they permit to solve a large range of real computer vision problems. With the recent advent of many 3D sensors, there is a growing interest in transposing and solving PDEs on surfaces and point clouds. In this paper, we propose a simple method to solve such PDEs using the framework of PDEs on graphs. This latter approach enables us to transcribe, for surfaces and point clouds, many models and algorithms designed for image processing. To illustrate our proposal, three problems are considered: (1) p -Laplacian restoration and inpainting; (2) PDEs mathematical morphology; and (3) active contours segmentation.

We propose a transcription on graphs of recent continuous global active contours proposed for image segmentation to address the problem of binary partitioning of data represented by graphs. To do so, using the framework of Partial difference Equations (PdEs), we propose a family of nonlocal regularization functionals that verify the co-area formula on graphs. The gradients of a sub-graph are introduced and their properties studied. Relations, for the case of a sub-graph, between the introduced nonlocal regularization functionals and nonlocal discrete perimeters are exhibited and the co-area formula on graphs is introduced. Finally, nonlocal global minimizers can be considered on graphs with the associated energies. Experiments show the benefits of the approach for nonlocal image segmentation and high dimensional data clustering.

The generalization of mathematical morphology to multivariate images is addressed in this paper. The proposed approach is fully unsupervised and consists in constructing a complete lattice from an image as a rank transformation together with a learned ordering of vectors. This unsupervised ordering of vectors relies on three steps: dictionary learning, manifold learning and out of sample extension. In addition to providing an efficient way to construct a vectorial ordering, nonlocal configurations based on color patches can be easily handled and provide much better results than with classical local morphological approaches.

In this paper we propose an adaptation of PDEs level sets over weighted graphs of arbitrary structure, based on PdEs and using a framework of discrete operators. A general PDEs level sets formulation is presented and an algorithm to solve such equation is described. Some transcriptions of well-known models under this formalism, as the mean-curvature-motion or active contours, are also provided. Then, we present several applications of our formalism, including image segmentation with active contours, using weighted graphs of arbitrary topologies.

In this paper, we introduce fast and robust digital algorithms for solving the Dirichlet problem with ∞-harmonic functions on graphs. Several PDEs and variational techniques have been proposed for a number of interpolation problems. Our motivation for this work is to extend some of these PDEs on graphs to deal with interpolation problems with a new approach in a discrete framework using the ∞-Laplacian on weighted graphs arbitrary topology. We show the experimental results for some applications of image interpolation that demonstrate the efficiency of our method and point out the interest of the novel algorithm with p = ∞ for interpolation problems.

A new hierarchical representation of general discrete data sets living on graphs is proposed. The approach takes advantage of recent works on graph regularization. The different levels of the hierarchy are discovered as the regularization process is performed. The role of the merging criterion that is common to hierarchical representations is greatly reduced due to the regularization step. This yields a robust representation of data sets. Moreover, the approach is particularly well adapted to the processing of digital images, where nonlocal processing allows to better handle repetitive patterns usually present in natural images.

Mathematical morphology (MM) offers a wide range of operators to address various image processing problems. These operators can be defined in terms of algebraic (discrete) sets or as partial differential equations (PDEs). In this paper, we introduce a nonlocal PDEs-based morphological framework defined on weighted graphs. We present and analyze a set of operators that leads to a family of discretized morphological PDEs on weighted graphs. Our formulation introduces nonlocal patch-based configurations for image processing and extends PDEs-based approach to the processing of arbitrary data such as nonuniform high dimensional data. Finally, we show the potentialities of our methodology in order to process, segment and classify images and arbitrary data.

Weighted Graph regularization provides a rich framework that allows to regularize functions defined over the vertices of a weighted graph. Until now, such a framework has been only defined for real or multivalued functions hereby restricting the regularization framework to numerical data. On the other hand, several kernels have been defined on structured objects such as strings or graphs. Using definite positive kernels, each original object is associated by the ``kernel trick'' to one element of an Hilbert space. As a consequence, this paper proposes to extend the weighted graph regularization framework to objects implicitly defined by their kernel hereby performing the regularization within the Hilbert space associated to the kernel. This work opens the door to the regularization of structured objects.

The main tools of Mathematical Morphology are a broad class of nonlinear image operators. They can be defined in terms of algebraic set operators or as Partial Differential Equations (PDEs). We propose a framework of partial difference equations on arbitrary graphs for introducing and analyzing morphological operators in local and non local configurations. The proposed framework unifies the classical local PDEs-based morphology for image processing, generalizes them for non local configurations and extends them to the processing of any discrete data living on graphs.

In this paper we present a method for impulse noise removal that makes use of spectral clustering and graph regularization. The image is modeled as a graph and local spectral analysis is performed to identify noisy and noise free pixels. On the set of noise free pixels, a topology adapted graph regularization is performed. Experimental results show the benefits of the proposed approach regarding the standard VMF when noise proportion is high.

In this paper we present a simple colorization method that relies on nonlocal graph regularization. We introduce nonlocal discrete differential operators and a family of weighted p-Laplace operators. Then, p-Laplace regularization on weighted graphs problem is presented and the associated filter family. Image colorization is then considered as a graph regularization problem for a function mapping vertices to chrominances. Several results illustrate our framework and demonstrate the benefits of nonlocal graph regularization for image colorization.

In this paper, a novel approach to mathematical morphology operations is proposed. Morphological operators based on partial differential equations (PDEs) are extended to weighted graphs of the arbitrary topologies by considering partial difference equations. We focus on a general class of morphological filters, the levelings; and propose a novel approach of such filters. Indeed, our methodology recovers classical local PDEs-based levelings in image processing, generalizes them to nonlocal configurations and extends them to process any discrete data that can be represented by a graph. Experimental results show applications and the potential of our levelings to textured image processing, region adjacency graph based multiscale leveling and unorganized data set filtering.

The process of segmenting images is one of the most critical ones in automatic image analysis whose goal can be regarded as to find what objects are presented in images. Artificial neural networks have been well developed. First two generations of neural networks have a lot of successful applications. Spiking neuron networks (SNNs) are often referred to as the 3rdgeneration of neural networks which have potential to solve problems related to biological stimuli. They derive their strength and interest from an accurate modeling of synaptic interactions between neurons, taking into account the time of spike emission. SNNs overcome the computational power of neural networks made of threshold or sigmoidal units. Moreover, SNNs add a new dimension, the temporal axis, to the representation capacity and the processing abilities of neural networks. In this paper, we present how SNN can be applied with efficacy in image segmentation.

A model selection method based on tabu search is proposed to build support vector machines (binary decision functions) of reduced complexity and efficient generalization. The aim is to build a fast and efficient support vector machines classifier. A criterion is defined to evaluate the decision function quality which blends recognition rate and the complexity of a binary decision functions together. The selection of the simplification level by vector quantization, of a feature subset and of support vector machines hyperparameters are performed by tabu search method to optimize the defined decision function quality criterion in order to find a good sub-optimal model on tractable times.

We introduce a nonlocal discrete regularization framework on weighted graphs of the arbitrary topologies for image and manifold processing. The approach considers the problem as a variational one, which consists of minimizing a weighted sum of two energy terms: a regularization one that uses a discrete weighted p-Dirichlet energy and an approximation one. This is the discrete analogue of recent continuous Euclidean nonlocal regularization functionals. The proposed formulation leads to a family of simple and fast nonlinear processing methods based on the weighted p-Laplace operator, parameterized by the degree p of regularity, the graph structure and the graph weight function. These discrete processing methods provide a graph-based version of recently proposed semi-local or nonlocal processing methods used in image and mesh processing, such as the bilateral filter, the TV digital filter or the nonlocal means filter. It works with equal ease on regular 2-D and 3-D images, manifolds or any data. We illustrate the abilities of the approach by applying it to various types of images, meshes, manifolds, and data represented as graphs.

A tool for diagnosis assistance by automatic segmentation of microscopic cellular images is introduced. This method is based on an automatic segmentation technique combining (with the Dempster-Shafer rule) the results obtained by support vector machines (SVM) applied within different color spaces. This combination is performed by integrating uncertainties and redundancies for each color space. Those uncertainties are computed as a posteriori probabilities according to the SVM obtained results. An improvement of the final segmentation quality is performed by taking into account the inconsistencies of several pixel classifications

In this paper, we study the ability of the cooperation of two-color pixel classification schemes (Bayesian and K-means classification) with color watershed. Using color pixel classification alone does not sufficiently accurately extract color regions so we suggest to use a strategy based on three steps: simplification, classification, and color watershed. Color watershed is based on a new aggregation function using local and global criteria. The strategy is performed on microscopic images. Quantitative measures are used to evaluate the resulting segmentations according to a learning set of reference images.

This article aims at showing an architecture of neural networks designed for the classification of data distributed among a high number of classes. A significant gain in the global classification rate can be obtained by using our architecture. This latter is based on a set of several little neural networks, each one discriminating only two classes. The specialization of each neural network simplifies their structure and improves the classification. Moreover, the learning step automatically determines the number of hidden neurons. The discussion is illustrated by tests on databases from the UCI machine learning database repository. The experimental results show that this architecture can achieve a faster learning, simpler neural networks and an improved performance in classification.