Showing papers by "Yannick Berthoumieu published in 2010"

Multivariate statistical modeling for texture analysis using wavelet transforms

Abstract: In the framework of wavelet-based analysis, this paper deals with texture modeling for classification or retrieval systems using non-Gaussian multivariate statistical features. We propose a stochastic model based on Spherically Invariant Random Vectors (SIRVs) joint density function with Weibull assumption to characterize the dependences between wavelet coefficients. For measuring similarity between two texture images, the Kullback-Leibler divergence (KLD) between the corresponding joint distributions is provided. The evaluation of model performance is carried out in the framework of retrieval system in terms of recognition rate. A comparative study between the proposed model and conventional models such as univariate Generalized Gaussian distribution and Multivariate Bessel K forms (MBKF) is conducted.

Structure tensor field regularization based on geometric features

Abstract: This paper investigates structure tensor field regularization applied to directional textured image analysis. From previous works on tensor filtering, we demonstrate that, knowing that the structure tensor is a specific tool coding the local geometry of the image, the tensor field filtering process must be driven by a geometric dissimilarity measure to define the adaptability of the smoothing process. We propose a new dissimilarity measure combining two terms devoted respectively to the orientation and to the shape component of the tensor. This intelligible encoding exhibiting the geometric structure of the image enables us to overcome major drawbacks of conventional Euclidean and Riemannian approaches for which the dissimilarity measure emphasizes only the local manifold geometry. Finally, for seismic imaging application, our method compared to existing ones shows that relevant information can be extracted by enhancing the seismic structures identification.

Bidirectional Composition on Lie Groups for Gradient-Based Image Alignment

Abstract: In this paper, a new formulation based on bidirectional composition on Lie groups (BCL) for parametric gradient-based image alignment is presented. Contrary to the conventional approaches, the BCL method takes advantage of the gradients of both template and current image without combining them a priori. Based on this bidirectional formulation, two methods are proposed and their relationship with state-of-the-art gradient based approaches is fully discussed. The first one, i.e., the BCL method, relies on the compositional framework to provide the minimization of the compensated error with respect to an augmented parameter vector. The second one, the projected BCL (PBCL), corresponds to a close approximation of the BCL approach. A comparative study is carried out dealing with computational complexity, convergence rate and frequence of convergence. Numerical experiments using a conventional benchmark show the performance improvement especially for asymmetric levels of noise, which is also discussed from a theoretical point of view.

Texture classification based on the Generalized Gamma distribution and the Dual Tree Complex Wavelet Transform

Abstract: This paper deals with stochastic texture modeling for classification issue. A generic stochastic model based on three-parameter Generalized Gamma (GG) distribution function is proposed. The GG modeling offers more flexibility parameterization than other kinds of heavy-tailed density devoted to wavelet empirical histograms characterization. Moreover, Kullback-leibler divergence is chosen as similarity measure between textures. Experiments carried out on Vistex texture database show that the proposed approach achieves good classification rates.

Semi-supervised learning for location recognition from wearable video

Abstract: This paper tackles the problem of image-based indoor location recognition. The context of the present work is activity monitoring using a wearable video camera data. Because application constraints necessitate weak supervision, a semi-supervised approach has been adopted which leverages the large amount of unlabeled images. The proposed method is based on the Bag of Features approach for image description followed by spectral dimensionality reduction in a transductive setup. Additional information from geometrical verification constraints are also considered which allowed to reach higher performance levels. The considered algorithms are compared experimentally on the data acquired in the wearable camera setup.

Stochasticity: A Feature for Analyzing and Understanding Textures, with Applications to Classification and Content-Based Image Retrieval

Abstract: Stochasticity is proposed as a feature for texture characterization and analysis. Measuring stochasticity requires finding suitable representations that can significantly reduce statistical dependencies of any order. Wavelet packet representations provide such a framework for a large class of stochastic processes. The paper first addresses the selection of the best wavelet packet basis with respect to the stochasticity criterion and by using the Kolmogorov stochasticity parameter. A best basis under stochasticity consideration makes possible accurate texture description trough a dictionary of parametric models, especially for non regular textures. Among the properties of such a representation, the paper shows that texture classification is possible through stochasticity consideration. The relevance of the analysis also makes possible content-based stochasticity retrieval with semantics and with respect to the order structure of the wavelet packet bases.