Region-Oriented DCT Coding of Still Images Based on Voronoi Diagrams

Abstract

This paper proposes a new region-oriented coding scheme of still images based on Voronoi diagrams. In this scheme, first an image is partitioned into Voronoi regions so that their boundaries can fit well the principal contours in the image. Then image signals in each Voronoi region are encoded using orthogonal transform. We introduce the hierarchical renewal method of positions of generators to improve performance of the segmentation. And the adaptive transform method which conducts one-dimensional DCT twice is utilized to reduce the amount of computation required for transform coding. Furthermore, this paper describes a technique which optimizes representative values of linear quantizers for DCT coefficients. Simulation results indicate that the proposed coding scheme achieves the almost same coding performance as the KLT-based scheme, although the former can remarkably reduce computational load compared with the latter.