基于视觉显著性的分块自适应压缩感知算法  

Blocked adaptive compressed sensing based on visual saliency

作　　者：王蓉芳[1] 陈佳伟[1] 焦李成[1] 孙奕菲[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：《华中科技大学学报（自然科学版）》2015年第1期127-132,共6页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家重点基础研究发展计划资助项目(2013CB329402);国家自然科学基金资助项目(61271302;61272282;61202176)

摘　　要：针对SAR图像的压缩感知重建问题,在分块压缩感知框架的基础上,提出了基于视觉显著性的分块自适应压缩感知算法.在采样阶段,每个子块的采样率依据显著信息自适应的变化;在重建阶段,根据不同图像显著信息的差异,自适应地滤波.实验结果表明:该方法不仅重建结果的整体质量更优,视觉效果更好,而且在重建后的图像中能更好地保持边缘和目标等重要特征.To solve the compressed sensing（CS）reconstruction problem of synthetic aperture radar（SAR）images,a new blocked adaptive compressed sensing based on visual saliency（VS-BACS）was proposed,based on the basic framework of block-based compressed sensing.The algorithm exploited the property that the visual saliency could found the important target regions of SAR images.VSBACS had mainly two advantages.First,every sub-block sample rate was adaptively changed with saliency in the sample stage.Second,the adaptive filter was realized according to saliency difference among the images in the reconstruction stage.The results show that the improved algorithm can provide superior performance on both the reconstructed image quality and the visual effect,and the edge and the important features of SAR images can be well preserved.

关 键 词：分块压缩感知 视觉显著性 自适应采样 自适应滤波器 图像重建 

分 类 号：TN957.52[电子电信—信号与信息处理]