MULTIRESOLUTION IMAGE PROCESSING TECHNIQUES WITH APPLICATIONS IN TEXTURE SEGMENTATION AND NONLINEAR FILTERING A Thesis

Comer, Mary L. Ph.D., Purdue University, December 1995. Multiresolution Image Processing Techniques with Applications in Texture Segmentation and Nonlinear Filtering. Major Professor: Edward J. Delp. We present a new algorithm for segmentation of textured images using a multiresolution Bayesian approach. The algorithm uses a multiresolution Gaussian autoregressive (MGAR) model for the pyramid representation of the observed image, and assumes a multiscale Markov random eld model for the class label pyramid. Unlike other approaches, which have either used a single-resolution representation of the observed image or implicitly assumed independence between di erent levels of a multiresolution representation of the observed image, the models used in this thesis incorporate correlations between di erent levels of both the observed image pyramid and the class label pyramid. The criterion used for segmentation is the minimization of the expected value of the number of misclassi ed nodes in the multiresolution lattice. The estimate which satis es this criterion is referred to as the \multiresolution maximization of the posterior marginals" (MMPM) estimate, and is a natural extension of the single-resolution maximization of the posterior marginals (MPM) estimate. The parameters of the MGAR model | the means, prediction coe cients, and prediction error variances of the di erent textures | are unknown. The expectation-maximization (EM) algorithm is used to estimate these parameters while simultaneously performing the segmentation. Analysis and experimental results demonstrating the performance of the algorithm are presented. We also propose new approaches for the extension of binary and grayscale morphological operations to color imagery. We investigate two approaches for \color