Multiscale Image Segmentation Using Wavelet-Domain Hidden Markov Models

A new image segmentation algorithm based wavelet-domain, referred to as joint adaptive context and multiscale segmentation (JACMS) is developed. Towards achieving lower computational complexity, we propose a fast training algorithm, when applied to image segmentation, this technique provides a reliable initial segmentation. The contextual labeling tree which is used for the context-based Bayesian interscale fusion is studied. In order to achieve higher accuracies of both texture classification and boundary localization during the interscale fusion, we develop adaptive context structures which apply to homogeneous regions or/and texture boundaries, respectively. Experiments demonstrate that the proposed algorithms yield excellent segmentation results on both synthetic and real world data examples.