Texture Segmentation and Classification in Biomedical Image Processing

s: Methods of image analysis belong to a general interdisciplinary area of multidimensional signal processing. The paper is devoted to selected intelligent techniques of biomedical image processing and namely to mathematical methods of image features extraction and image components classification invariant to their rotation. The first method under study presents an algorithm for the given image segmentation using watershed transform allowing the estimation of image segments boundaries and image components classification. This problem is studied in connection with the application of the Radon transform used to change texture rotation to its translation followed by the shift invariant wavelet transform to estimate image components features. The second method presents basic principle of feature based image segmentation using pattern vectors assigned to all image pixels with vector values estimated from each root pixel neighbourhood properties. Proposed methods are verified for simulated images formed by a mixture of different textures and then applied to selected biomedical images.