摘要
本文提出一种基于方向提升小波变换的多描述图像编码方法,以避免压缩图像在互联网和无线网络环境下因数据包传输出错或丢失向导致的图像质量严重下降．基本思想是在空域将一幅图像按梅花形下采样为两个描述,为解决传统空域下采样多描述编码技术编码效率低的问题,提出能充分利用相邻像素相关性的方向提升小波变换,并通过凹叉树的分割实现方向的组织和选择．边缘信道解码出的低分辨率图像,通过插值算法实现原始分辨率单描述的图像重建．中央信道解码出的图像,通过数据融合算法实现两描述的图像重建,进而进一步改善图像视觉质量．
In this paper, a new multiple description image coding scheme based on directional lifting wavelet transform is proposed, which can avoid the serious quality degradation of compressed images in the Internet or wireless environment due to the packet error or packet loss. The basic idea is to divide an image into two descriptions with quincunx downsampling. To tackle the problem of low coding efficiency in the traditional spatial domain multiple description image coding techniques, we propose a directional lifting technique to extensively exploit the correlations among neighboring pixels, in which the lifting directions are selected and organized based on the quad-tree partition. A novel interpolation scheme is employed in the side decoder to reconstruct one description with full resolution. The data fusion algorithm is employed in the central decoder to reconstruct the whole image, which can further improve the quality of the reconstructed image.
出处
《自动化学报》
EI
CSCD
北大核心
2007年第6期567-576,共10页
Acta Automatica Sinica
基金
国家自然科学基金(60572104
60533030
60375007)
国家863计划(2006AA01Z317)
北京市自然科学基金(40410031
4061001)
多媒体与智能软件技术北京市重点实验室开放基金项目资助~~
关键词
多描述编码
梅花下采样
提升小波变换
方向预测
图像插值
Multiple description coding, quincunx downsarnpling, lifting wavelet transform, directional prediction,image interpolation
