Object-based image segmentation using DWT/RDWT multiresolution Markov random field

This paper introduces a segmentation algorithm for object-based image coding techniques. This scheme is based on discrete wavelet transform (DWT)/redundant discrete wavelet transform (RDWT) and multiresolution Markov random field (MMRF). The DWT based MMRF works well for noise-free images. It merges details in the original image with their respective visual objects and divides the image into different segments according to their textures. The RDWT based MMRF is a generalization of the DWT based MMRF. When the noise level is high, the RDWT based MMRF reduces the influence of noise in the segmentation procedure and generates much better results at some computing costs. The proposed algorithm has been successfully integrated with our DWT based region of interest (ROI) compression coder, the generalized self-similarity trees (GST) codec, for networking applications.