利用纹理信息的图像分块自适应压缩感知  

Block-Based Adaptive Compressed Sensing of Image Using Texture Information

作　　者：王蓉芳[1] 焦李成[1] 刘芳[1] 杨淑媛[1] 

机构地区：[1]西安电子科技大学智能感知与图像理解教育部重点实验室,陕西西安710071

出　　处：《电子学报》2013年第8期1506-1514,共9页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61173090;No.61072106;No.61072108);高等学校学科创新引智计划(111计划)(No.B07048);教育部"长江学者和创新团队发展计划"(No.IRT1170)

摘　　要：与一维信号不同,二维图像有明显的纹理信息.本文分析了不同图像之间,以及同一图像不同子块之间,不同纹理引起的信息量差异,在分块压缩感知算法的基础上,提出了利用纹理信息的图像分块自适应压缩感知算法.自适应性体现在自适应采样和自适应收缩阈值两个方面.引入两种滤波器,分别形成了两种分块自适应压缩感知算法.采用自然和医学两类测试图像,验证了两种新算法的性能.实验结果表明,利用了纹理信息的分块自适应压缩感知算法,在重构图像的质量和视觉效果上,都有明显的优势.Different from 1D signals,2D images have obvious texture information.Based on the basic framework of block-based compressed sensing,we propose a block-based adaptive compressed sensing of image using texture information,which exploits the property that different texture would produce varied information content.The adaptability lies in two aspects：one is the adaptive sampling and another is the adaptive threshold.Two different filters are introduced,which forms two kinds of new algorithms,respectively.The proposed algorithms are applied to nature and medical test images.The results show that our improved algorithms provide superior performance on both the reconstructed image quality and the visual effect.

关 键 词：分块压缩感知 纹理信息 自适应采样 自适应阈值 滤波器 

分 类 号：TN911[电子电信—通信与信息系统]