基于显著性信息的压缩感知图像可分级编码  

Saliency-based scalable image compressive sensing coding

作　　者：陈守宁[1] 郑宝玉[1] 赵玉娟[2] 

机构地区：[1]南京邮电大学宽带无线通信与传感网技术教育部重点实验室,江苏南京210003 [2]江苏第二师范学院数学与信息技术学院,江苏南京210013

出　　处：《南京邮电大学学报（自然科学版）》2016年第1期42-49,共8页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(61271240);江苏省普通高校研究生科研创新计划(CXLX12_0475)资助项目

摘　　要：多媒体信号在相对不稳定和带宽有限的无线网络环境下,易出现丢包等传输问题,同时码率受限,影响接收端多媒体信号感官质量。图像压缩感知作为一种结合采样和压缩的信号处理理论,具有编码简单,解码复杂的结构特性,适用于能耗受限的传感器无线网络中的多媒体信号采集与传输问题。但是,现有图像压缩感知技术尚存在压缩效率有限,恢复质量不高等问题,同样易受无线传输环境丢包影响。文中针对图像压缩感知压缩效率和无线传输环境下传输质量受限问题,通过运用图像视觉显著性信息判决技术和路径分集技术,对图像关键区域少量的可分级信息冗余增加,进行非对称的信道保护策略,来保障图像中视觉关注较高区域恢复质量。文中提出的基于显著性信息的压缩感知图像可分级编码方法,在无差错情况下,率失真性能优于传统无显著性信息方法,在丢包网络环境下,优于CS-MDC算法。In the relatively unstable and bandwidth limited wireless network environment,multimedia signal is susceptible to transmission problems like packet loss and limited rate,affecting perceptual quality of received signal. As a signal processing theory combining sampling and compression,compressive sensing for image is suitable for multimedia signal acquisition and transmission in energy constrained wireless sensor network,due to its structure of simple encoding and complicated decoding. However,the existing image compressive sensing technology is still limited in compression efficiency and recovery quality,and also vulnerable to the packet loss in wireless transmission environment. Aiming at compressive sensing problems of compression efficiency and transmission quality in wireless environment,increasing a little scalable redundancy information in the key region of image can guarantee the recovery quality in the region with high visual intension,which is an asymmetric channel protection strategy. Saliency-based scalable compressive sensing for image is proposed and its rate distortion performance is better than that of the traditional method without considering saliency information in error-free case,and better than that of the CSMDC algorithm in packet loss case.

关 键 词：压缩感知 可分级编码 显著性信息判决 路径分集 

分 类 号：TN911.73[电子电信—通信与信息系统]