Multiscale image segmentation using wavelet-domain hidden Markov models

We introduce a new image texture segmentation algorithm, HMTseg, based on w avelets and the hidden Markov tree (HMT) model. The HMT is a tree-structure d probabilistic graph that captures the statistical properties of the coeff icients of the wavelet transform. Since the HMT is particularly well suited to images containing singularities (edges and ridges), it provides a good classifier for distinguishing between textures. Utilizing the inherent tree structure of the wavelet EMT and its fast training and likelihood computat ion algorithms, we perform texture classification at a range of different s cales. We then fuse these multiscale classifications using a Bayesian proba bilistic graph to obtain reliable final segmentations. Since HMTseg works o n the wavelet transform of the image, it can directly segment wavelet-compr essed images without the need for decompression into the space domain. We d emonstrate the performance of HMTseg with synthetic, aerial photo, and docu ment image segmentations.