Philippe Tarroux

Unsupervised segmentation of Markov random field modeled textured images using selectionist relaxation

Among the existing texture segmentation methods, those relying on Markov random fields have retained substantial interest and have proved to be very efficient in supervised mode. The use of Markov random fields in unsupervised mode is, however, hampered by the parameter estimation problem. The recent solutions proposed to overcome this difficulty rely on assumptions about the shapes of the textured regions or about the number of textures in the input image that may not be satisfied in practice. In this paper, an evolutionary approach, selectionist relaxation, is proposed as a solution to the problem of segmenting Markov random field modeled textures in unsupervised mode. In selectionist relaxation, the computation is distributed among a population of units that iteratively evolves according to simple and local evolutionary rules. A unit is an association between a label and a texture parameter vector. The units whose likelihood is high are allowed to spread over the image and to replace the units that receive lower support from the data. Consequently, some labels are growing while others are eliminated. Starting with an initial random population, this evolutionary process eventually results in a stable labelization of the image, which is taken as the segmentation. In this work, the generalized Ising model is used to represent textured data. Because of the awkward nature of the partition function in this model, a high-temperature approximation is introduced to allow the evaluation of unit likelihoods. Experimental results on images containing various synthetic and natural textures are reported.

Unsupervised image segmentation using a distributed genetic algorithm

Abstract A new methodological approach to digital image processing applied to the particular case of gray-level image segmentation is introduced. The method is based on a modified and simplified version of classifier systems. The labeling function is implemented as a spatially structured set of binary-coded production rules. The labeling is iteratively modified using a distributed genetic algorithm. Results are presented which illustrate both the mechanisms underlying the functioning of the method and its performance on natural images. The relationships between this approach and other related techniques are discussed and it is shown that it compares favorably with these.

Two-dimensional electrophoresis computerized processing.

This paper describes various methods suitable for implementation of two-dimensional processing software. The different steps leading to a complete processing are described, from the digitalization of the image to the processing of the resulting data. The characteristics of a convenient digitalization system are discussed. The different software devoted to spot detection is reviewed with respect to the presence or otherwise of a spot model and its characteristics. The major techniques for gel matching are compared as are designs for database structures suitable for tabulation of measurements. Finally, the need for a sophisticated system of data processing is stressed and its main requirements are described.

Procedures for two-dimensional electrophoretic analysis of nuclear proteins

Abstract A series of mehtods designed for electrophoresis of nuclear proteins is described. They deal with the low solubility of many nuclear proteins and with the presence of large amounts of nucleic acids. These were eliminated by enzymatic digestion, centrifugation, partition or precipitation. A combination of RNAse digestion and centrifugation is the method of choice when proteolysis is low. In the opposite case, precipitationor partition methods are preferred, at the expense of precipitation of some proteins. When the nuclear RNA content is low, centrifugation in an Airfuge is the simplest and the most efficient mehtod, superseding the widely used S1 nuclease method.

A new tool to study genetic expression using 2-D electrophoresis data: the functional map concept

A method derived from general computerized data analysis techniques is used here first to tabulate the characteristics of individual peptide spots observed by two‐dimensional electrophoresis and to compare these characteristics among spots. Multivariate analysis of such data arrays then leads to the grouping of spots in an n‐dimensional space, according to their expression characteristics. Such maps display zones which are characteristic of the tissue studied. These functional maps provide a powerful analytical tool for describing such functional interactions and also for predicting the function of unidentified peptides observed on the electrophoresis gels.

Covert attention with a spiking neural network

We propose an implementation of covert attention mechanisms with spiking neurons. Spiking neural models describe the activity of a neuron with precise spike-timing rather than firing rate. We investigate the interests offered by such a temporal code for low-level vision and early attentional process. This paper describes a spiking neural network which achieves saliency extraction and stable attentional focus of a moving stimulus. Experimental results obtained using real visual scene illustrate the robustness and the quickness of this approach.

Attentional mechanisms for interactive image exploration

A lot of work has been devoted to content-based image retrieval from large image databases. The traditional approaches are based on the analysis of the whole image content both in terms of low-level and semantic characteristics. We investigate in this paper an approach based on attentional mechanisms and active vision. We describe a visual architecture that combines bottom-up and top-down approaches for identifying regions of interest according to a given goal. We show that a coarse description of the searched target combined with a bottom-up saliency map provides an efficient way to find specified targets on images. The proposed system is a first step towards the development of software agents able to search for image content in image databases.

Complete computer system for processing chromatographic data

Abstract A complete data processing system for complex chromatogram analysis is described. Although this system is essentially designed to treat high-performance liquid chromatographic results, it can be applied to several other separation techniques. Original methods are indicated for chromatogram peak integration and evaluation of peak areas, even when separation of compounds is incomplete. For gradient analyses, techniques for filtering and correction of baseline are described. The ability to treat radioactive counting data enables the system to be applied to the processing of metabolic arrays. The use of even derivatives to obtain a criterion of peak purity and the problem of information losses during sampling are also discussed.

Unsupervised Texture Segmentation using Selectionist Relaxation

We introduced an unsupervised texture segmentation method, the selectionist relaxation, relying on a Markov Random Field (MRF) texture description and a genetic algorithm based relaxation scheme. It has been shown elsewhere that this method is convenient for achieving a parallel and reliable estimation of MRF parameters and consequently a correct image segmentation. Nevertheless, these results have been obtained with an order 2 model on artificial textures. The purpose of the present work is to extend the use of this technique to higher orders and to show that it is suitable for the segmentation of natural textures, which require orders higher than 2 to be accurately described. The results reported here have been obtained using the generalized Ising model but the method can be easily transposed to other models.

Detection of compositional constraints in nucleic acid sequences using neural networks

We describe in this paper a neural network method for the detection of compositional constraints in introns and exons. The first part of the algorithm (learning phase) consisted in presenting examples of intron and exon sequences to the network and in modifying its connections using the back-propagation algorithm. Previous connectionist methods achieved the learning of exons and introns using the latter as negative examples to the former. However, we chose to learn introns and exons jointly, using junk DNA as a common counter-example. In a second part (generalization phase), we tested the neural networks in the search for exons and introns in the human globin cluster. Their performances were also checked on the classification of unknown examples. As with the previous approaches, this technique discriminates introns and exons: values of the correlation coefficients are respectively 0.50 and 0.64 for the best achieved network. Moreover, using junk DNA sequences in the learning phase allows one to detect constrained regions inside the intron and the exon sequences (i.e. sequences that differ, by their nucleic acid compositions, from junk DNA). The application of our approach could be useful in the study of the internal organization of these sequences.