An unsupervised classification algorithm is derived by modeling observed data as a mixture of several mutually exclusive classes that are each described by linear combinations of independent non-Gaussian densities. The algorithm estimates the data density in each class by using parametric nonlinear functions that fit to the non-Gaussian structure of the data. This improves classification accuracy compared with standard Gaussian mixture models. When applied to textures, the algorithm can learn basis functions for images that capture the statistically significant structure intrinsic in the images. We apply this technique to the problem of unsupervised texture classification and segmentation. intrinsic structure of its image texture. In a mixture model, the observed data can be categorized into several mutually exclusive classes (3). When the data in each class are modeled as multivariate Gaussian, it is called a Gaussian mixture model. We generalize this by assuming that the data in each class are generated by a linear combination of independent, non-Gaussian sources, called as ICA mixture model. The algorithm for learning the parameters of the model uses gradient descent algorithm to maximize log likelihood function. We apply this learning algorithm to the problem of unsupervised classification and segmentation of textures. 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 site 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. Claude.I, smolarz. A (17) focus on stochastic approaches and, specifically, on texture modelling by bidimensional autoregressive models (2D-AR models). They describe the AR model and propose a method for choosing an adapted neighbourhood and evaluation. Then, the segmentation algorithm is presented with the classification criterion and the contextual information.