Frédéric Guérault

Influence of the noise model on level set active contour segmentation

We analyze level set implementation of region snakes based on the maximum likelihood method for different noise models that belong to the exponential family. We show that this approach can improve segmentation results in noisy images and we demonstrate that the regularization term can be efficiently determined using an information theory-based approach, i.e., the minimum description length principle. The criterion to be optimized has no free parameter to be tuned by the user and the obtained segmentation technique is adapted to nonsimply connected objects.

Nonparametric statistical snake based on the minimum stochastic complexity

We propose a nonparametric statistical snake technique that is based on the minimization of the stochastic complexity (minimum description length principle). The probability distributions of the gray levels in the different regions of the image are described with step functions with parameters that are estimated. The segmentation is thus obtained by minimizing a criterion that does not include any parameter to be tuned by the user. We illustrate the robustness of this technique on various types of images with level set and polygonal contour models. The efficiency of this approach is also analyzed in comparison with parametric statistical techniques

Unified statistically independent region processor for deterministic and fluctuating targets in nonoverlapping background.

Recently, new approaches for location of a target in nonoverlapping noise, which are optimal in the maximum-likelihood sense, have been proposed. In particular, different methods for deterministic or fluctuating targets have been developed. We propose a unified and optimal processor for a target with either known or unknown gray levels. We demonstrate the efficiency and robustness of this method in comparison with previously developed processors.

Location of target with random gray levels in correlated background with optimal processors and preprocessings

We address the problem of locating a target with a random correlated texture appearing on a disjoint random correlated background for tracking applications. We first review recent techniques based on SIR models and then we show how to enlarge the field of application of these approaches by preprocessing the input scenes. With numerical simulations, we analyze the probability of correct location of the different proposed algorithms for several types of correlated targets and backgrounds.

Optimal χ 2 filtering method and application to targets and backgrounds with random correlated gray levels

We describe a pattern recognition processor based on a new optimal x(2) filtering method that is designed to localize a target with unknown gray levels appearing on a random background. This processor consists of preprocessing of the analyzed image followed by correlations with binary masks.

Statistically independent region processor for target and background with random textures: whitening preprocessing approach

Abstract The problem of locating a target with a random texture appearing on a random background is addressed. We describe a whitening preprocessing approach which allows one to generalize the domain of application of the Statistically Independent Region processor. It is shown that the statistics of both the target and the background of the preprocessed image can be approximated by white Gaussian probability density functions, which correspond to the condition of optimality for the Statistically Independent Region processor. Furthermore, we analyze the relevance of introducing a boundary between the target and the background in the preprocessed image.

Location of target in correlated background with the SIR processor

In this paper, we address the problem of the localization of a target with random grey levels appearing on a random background. The optimal processor for any white Gaussian statistics of the targets and the backgrounds grey levels is the white Gaussian -statistically independent region processor. If the statistics of the grey levels of the input image are no white, it is shown that a whitening preprocessing can allow one to model with a a good approximation the statistics of the preprocessed image by three white Gaussian probability density functions, which characterize the target, the background and the boundary between the target and the background in the preprocessed image. The preprocessing can thus generalize the domain of application of the statistically independent region processors.

Stochastic Complexity based Image Segmentation with unknown Noise Model

We propose a general statistical image segmentation method which does not need any a priori knowledge of the probability density functions (PDF) of the grey levels of the image. This method is based on the minimization of the stochastic complexity (Minimum Description Length principle) which leads to optimize a criterion without parameter to be tuned by the user which is adapted to the PDF of the grey levels of the image. We apply this method to three partition descriptors: a polygonal active contour, a level set implementation and a polygonal active grid. We illustrate the technique on real images.

Nonparametric Statistical Level Set Snake Based on the Minimization of the Stochastic Complexity

In this paper, we focus on the segmentation of objects not necessarily simply connected using level set snakes and we present a nonparametric statistical approach based on the minimum stochastic complexity principle. This approach allows one to get a criterion to be optimized with no free parameter to be tuned by the user. We thus propose to estimate the probability law of the gray levels of the object and the background of the image with a step function whose order is automatically determinated. We show that coupling the probability law estimation and the segmentation steps leads to good results on various types of images. We illustrate the robustness of the proposed nonparametric statistical snake on different examples and we show on synthetic images that the segmentation results are equivalent to those obtained with a parametric statistical technique, although the technique is nonparametric and without ad hoc parameter in the optimized criterion.

Optimal location of a fluctuating intensity target in a nonhomogeneous and nonoverlapping background

Recently, optimal algorithms for locating a target on nonoverlapping background, based on maximum likelihood approach, have been designed. In particular, different ways of modeling the target have been proposed. When the gray levels of the target are known, the reference of the target can be modeled as a deterministic function. On the other hand, when the gray levels of the target in the input image are unknown or can vary from one image to another one, the reference of the target has to be considered as a pattern with random gray levels. Moreover, it is possible to unify both the deterministic and the random target approaches into a single model, where the target is modeled using a linear combination of deterministic values and random variables. Based on this model, we propose to design an algorithm that optimizes the likelihood ratio between the two hypothesis that a target is present and that it is absent within a small sub-window of the image. We show that this technique is more efficient than the maximum likelihood approach when the noise statistic of the background is strongly nonhomogeneous, which is the case in many real-world images. The presented algorithm is based on correlations and it can thus be implemented in an architecture using optical correlators.