Maximum likelihood texture analysis and classification using wavelet-domain hidden Markov models

Abstract

Wavelet-domain hidden Markov models (HMMs), in particular the hidden Markov tree (HMT), have been proposed and applied to image processing, e.g. denoising and segmentation. In this paper texture analysis and classification using wavelet-domain HMMs are studied. In order to achieve more accurate texture characterization, we propose a new tree-structured HMM, called the 2-D HMT-3, where the wavelet coefficients from three subbands are grouped together. Besides the interscale dependencies, the proposed 2-D HMT-3 can also capture the dependencies across the wavelet subbands that are found useful for texture analysis. The experimental results show that the 2-D HMT-3 provides a nearly 20% improvement over the method using wavelet energy signatures, and the overall percentage of correct classification is over 95% upon a set of 55 Brodatz (1966) textures.