一种基于人类视觉系统的去块效应算法  

A Deblocking Algorithm Based on HVS

作　　者：许志良[1] 谢胜利[1] 

机构地区：[1]华南理工大学电子与信息学院

出　　处：《电子与信息学报》2005年第11期1717-1721,共5页Journal of Electronics & Information Technology

基　　金：国家杰出青年科学基金(60325310)教育部跨世纪优秀人才基金国家自然科学基金(60274006)资助课题

摘　　要：基于块离散余弦变换的图像和视频压缩主要缺点就是在低比特率时会在块边界出现明显的方块效应。本文提出一种充分利用人类视觉特性,在图像的平滑区和纹理区分别采用一维DCT 域滤波和空间域滤波的去块效应算法。实验结果表明该算法既能有效地去除方块效应又能保护图像的边缘信息。A major drawback related to the Block-based Discrete Cosine Transform（BDCT） for image and video compression is that the decoded images, especially at very low bit rates,exhibit highly noticeable blocking artifact near the block boundaries. In this paper, a deblocking algorithm based on Human Visual System （HVS） is proposed. For smooth regions, one dimension DCT domains filter is applied to reduce the blocking artifacts, but for texture regions a spatial filter is used. The experimental results show that the proposed algorithm reduces the blocking artifacts effectively and preserves the original edges faithfully.

关 键 词：块离散余弦变换 方块效应 人类视觉系统 斜率均方差 

分 类 号：TN919.8[电子电信—通信与信息系统]