均等色空間における画像の領域分割符号化 : 画像応用

抄録

This paper presents a new region segmentation coding based on HVC color space. To segment an image into several characteristic regions, we adopt the segmentation by K-mean clustering algorithm in HVC color space and region merging process based on the measure of Godlove's color difference. The region segmentation coding scheme are as follows; 1) We encode the contour information of the region by using the chain-code. 2) Hue signals within one region are replaced by one representative Hue signal. 3) To encode V signals, we adopt the adaptive DCT coding scheme. 4) The Chroma signal within one region is approximated by the two-dimensional polynomial function. We encode the polynomial coefficients for each region. A satisfactory quality images are obtained by the following coding rate : 0.6 - 2.0[bit/pixel].