Segmentation based compression for graylevel images

This paper presents a segmentation based lossy image compression (SLIC) algorithm. The segmentation scheme (Biswas and Pal, Pattern Recog. Lett. 21 (2000)), entropy based and hierarchical in nature, provides sub-images of homogeneous regions. The compression algorithm encodes a graylevel image through global approximations of sub-images by 2-d Bezier-Bernstein polynomial along with corrections, if needed, over regions in sub-images by local approximation; contours by 1-d Bezier-Bernstein polynomial and texture, if present, by Huffman coding scheme using Hilbert scan on texture blocks. Order of the 2-d polynomials has been computed with the help of an image quality index (IQI). The proposed compression algorithm also examines the compression result by encoding contours through their approximation based on stretching of discrete circular arcs. Stretching is done by affine transformation. Compression results in both the cases have been compared with JPEG results. Attempts have been made to evaluate the quality of reconstructed images through a fidelity vector whose components are different objective measures. (C) 2003 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.