Color Satellite Image Segmentation Using Markov Random Field and Multiresolutional Wavelet Transform

Image segmentation plays an important role in human vision, computer vision, and pattern recognition fields. Segmentation based on texture can improve the accuracy of interpretation. Satellite images are used in order to detect the distribution of classes such as soil, vegetation, built-up areas, roads, rivers, lakes etc. A problem that arises when segmenting an image is that the number of feature variables or dimensionality is often quite large. In this paper we used random field theory for identification of those classes and used multi resolution Haar wavelet transformation to put each pixel in desired class with great probability. Experiments are conducted on a set of 30 natural satellite texture images. A specific attention is paid to the use of Haar transform as a tool for image compression and image pixels feature extraction. Proposed algorithm is verified for simulated images and applied for a selected satellite image processing in the MATLAB environment. Keywordsrandom field theory, multi resolution analysis, texture, wavelet transformation