A WAVELET DOMAIN HIERARCHICAL HIDDEN MARKOV MODEL (WP-P8)

Author(s) : Zhen Ye (Dept. of Computer Science, Kent State University, USA) Cheng-Chang Lu (Dept. of Computer Science, Kent State University, USA)

Abstract : This paper proposes a wavelet-domain hierarchical hidden Markov model for an unsupervised texture segmentation. Based on a hybrid graph structure, the global dependencies can be captured by a quadtree structure across all scales, and local dependencies at higher resolution scales can be captured by a pyramidal graph structure. Novel contexts that include different positions, orientations, and scales are introduced. Parameters of this composite dependency model are estimated by an expectation maximization (EM) algorithm. Applications of an unsupervised texture segmentation are presented. Compared with other alternative approaches for several test images, this method can achieve a significant improvement in segmentation, especially at higher resolution scales.

Menu