Research For Image Representation And Image Segmentation Algorithms

An image often contains a lot of redundant information, we hope to extract instinctive visual information in image representation, and so is image compression. We wipe off redundant information and reserve sensitive information in image compression.Here we study two image representation algorithms: HIMPA and epitome. HIMPA is a hybrid of ICA and MPPCA. The epitome of an image is its miniature, condensed version containing the essence of the textural and shape properties of the image. We implement two above image representation algorithm and compare their performance against DCT for ten images using PSNR measurement.Image segmentation extracts meaningful regions or objects from visual data. It decomposes images, according to specific features of interest, into distinct regions to make high-level tasks such as object tracking, recognition, and scene interpretation possible.We study three image segmentation algorithms: SMAP, QBS, HMTseg. SMAP is a Bayesian image segmentation scheme which replaces the MRP model with a novel multiscale random field (MSRF), and replaces the MAP estimator with a sequential MAP (SMAP) estimator derived from a novel estimation criterion. QBS proposes a general quadrilateral-based framework for image segmentation, in which quadrilaterals are first constructed from an edge map, where neighboring quadrilaterals with similar features of interest are then merged together to form regions. Finally, each segmented region is accurately and completely described by a set of quadrilaterals. HMTseg is a new image texture segmentation algorithm, based on wavelets and the hidden Markov tree (HMT) model. The HMT is a tree-structured probabilistic graph that captures the statistical properties of the coefficients of the wavelet transform. Utilizing the inherent tree structure of the wavelet HMT and its fast training and likelihood computation algorithms, HMTseg performs texture classification at a range of different scales and then fuses these multiscale classifications using a Bayesian probabilistic graph to obtain reliable final segmentations. We implement these algorithms, then evaluate and compare their performance with objective measurement and subjective measurement.