A new textured image segmentation algorithm by autoregressive modelling and multiscale block classification

Usually, we begin the analysis of an image by partitioning it into several homogeneous regions; this is the image segmentation. The homogeneity can be defined by many properties such as gray level intensity, color, texture, etc. In many cases, the texture is the only available information. Texture analysis, as a result, has received considerable attention for the last two decades. A large number of approaches for texture classification and segmentation have been suggested. Commonly, two types of approaches are distinguished, adapted respectively to macro- and microtextures, namely, the structural and statistical approaches. As far as the latter is concerned, we can cite probabilistic methods based on texture modelling, statistical methods which characterize an image in terms of numerical attributes or features and new tools like neural networks, wavelets, multiresolution and multiscale approaches, and fuzzy modelling. A few methods also come from signal processing and seem to be promising: bidimensional autoregressive modelling and, time-frequency and time-scale representations. We focus on stochastic approaches and, specifically, on texture modelling by bidimensional autoregressive models (2D-AR models). We describe the AR model; we propose our method for choosing an adapted neighbourhood and evaluation. Then, our segmentation algorithm is presented with the classification criterion and the contextual information. Finally, we present experimental results showing the influence of the context and demonstrating the improvements brought by the adapted models and the multiscale approach.