Fuzzy segmented image coding using orthonormal bases and derivative chain coding

摘要

A framework for fuzzy segmentation-based image coding (FSIC) is presented. The segmentation is carried out on the basis of a region-growing algorithm which uses a fuzzy rule-based system for the evaluation of the homogeneity criterion. This approach allows to incorporate complex systems of fuzzy rules and to find an appropriate solution even for contradicting image features. The specific features and rules have been chosen such that image segments with smooth contours result and very small region are avoided. Coding the contours of these regions by a derivative chain code leads to an efficient image compression scheme that can be utilized, where medium and high compression ratios together with acceptable reconstruction quality and great flexibility are needed. Segmentation results are compared with standard segmentation-based image coding (SIC) and with JPEG-compression. Afterwards, a method is presented to improve the reconstruction quality which is based on least-squares approximation. To efficiently calculate the estimate for each segmented region an orthonormal set of basis functions is constructed from a linearly independent starting base using the Gram–Schmidt method. With this approximation the reconstruction quality is improved significantly and false contours are reduced. Besides, it offers the possibility of progressive image reconstruction. This characteristic makes the approximation with orthonormal bases interesting for many state-of-the-art audio-visual applications. Segmentation results are shown as well as their improvement using orthonormal bases.