以輪廓及紋理做彩色影像壓縮

Abstract

輪廓式壓縮法已經被研究了好幾年。然而﹐大部分被提出來的方法都是為 黑白影像而設計的﹐並且它們所還原之影像的失真程度都很嚴重。因此﹐ 在本論文中﹐我們提出一個輪廓式的彩色影像壓縮法﹐希望能夠減少所還 原之影像的失真程度。首先﹐我們用「區塊擴大法」及Weber's ratio來 切割影像。之後,再對所切出之區塊的紋理及輪廓編碼。為了對區塊的紋 理編碼﹐我們用雙變數多項式及此區塊的平均顏色來分別表示較大及較小 的區塊之紋理。此外﹐我們也用「鍊式編碼法」來對區塊的輪廓編碼。實 驗結果顯示﹐我們的方法可適當地表現出大部分複雜的紋理﹐且失真程度 可以接受。而壓縮率大都在 40 倍到 80 倍之間。 Contour-based compression methods have been studied for several years.However, most of the reported approaches are designed for gray-level images, and the distortion in the reconstructed images is serious. Inthis thesis, we propose a contour-based compression method for color images and wish to reduce the distortion in the reconstructed images.First, we segment the image by region growing and Weber's ratio in color space. After that, the contours and textures of the segmented regions are coded. To encode the texture of a region, a polynomial function (or the average color of a region) is used to represent thetexture, if the region is large (small). In addition, the chain codingis used to encode the region contour. Experimental results show thatmost of the complex texture can be reconstructed suitably and the distortion is acceptable. The compression ratios are usually between40:1 and 80:1.