An unsupervised multiresolution textured image segmentation using wavelet-domain classification

The wavelet-domain Hidden Markov Tree (HMT) is a statistic model very suitable for textured images and is used to supervise segmentation successfully. Based on this model, an unsupervised multiresolution segmentation algorithm in Wavelet-Domain is proposed. We use a Gaussian mixture model to characterize wavelet coefficients at each scale and each subband. We also use the Markov Random Field (MRF) to model the class label of each wavelet coefficient. The Expectation Maximization (EM) algorithm is used to estimate the parameters of this model directly from images. Ultimately the multiresolution classifications are fused by the Hybrid Contextual Labeling Tree (HCLT) to obtain a reliable and accurate segmentation.