基于JPEG2000的压缩图像纹理信息重建  

The Rebuilding of Texture Information of Compressed Image Based on JPEG2000

作　　者：林立宇[1] 杨志高[2] 袁军[3] 秦前清[2] 

机构地区：[1]武汉大学多媒体网络通信工程湖北省重点实验室,湖北武汉430072 [2]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079 [3]武汉大学电子信息学院,湖北武汉430072

出　　处：《武汉大学学报（理学版）》2006年第1期95-99,共5页Journal of Wuhan University:Natural Science Edition

基　　金：国家自然科学基金资助项目(40204008)

摘　　要：提出了基于JPEG2000的压缩图像纹理信息重建的一种方法.该方法在编码端用高斯模型来描述图像的统计纹理信息,并对这些纹理信息进行统计和编码,在解码端根据编码的信息重建图像的纹理.通过和非线性量化扩展补偿法的结果在客观指标和主观品质上的比较,说明本文提出的方法可以通过较小的比特开销(低于10%)重建丢失的纹理信息,从而获取压缩图像的品质在主观视觉效果上的提高.Some details of the image will lose in the image lossy compression. Occasionally, these details, such as the texture of the human skin, will make an image look unnaturally though the real distortion of the image is not sever very much. This paper proposes a new method of rebuilding the statistical texture in compressed image which is based on JPEG2000 in order to get better subjective compensation by less cost of bit rate. Meanwhile compared with non-linear extended quantization methods＇ in both quality parameter and subjective effect, the method make it possible to regenerate the texture for the decoded image. By encoding the statistical result of the texture information of the compressed image during the coding process, which is on the assumption that the Gauss distributing is effective here.

关 键 词：图像压缩 JPEG2000 统计纹理 主观视觉补偿 小波变换 

分 类 号：TN919.81[电子电信—通信与信息系统]