摘要
贝叶斯压缩感知是一种基于统计分析的压缩感知算法,具有很好的鲁棒性,能够充分利用信息间的相关性,它的重构依赖于图像的稀疏性表达.针对贝叶斯压缩感知的深层次稀疏化问题,笔者结合自适应字典学习思想,提出一种冗余自适应字典表示的稀疏贝叶斯学习算法.该算法对图像进行局部分块,从待重建图像的迭代中间图像分块中学习字典,并以该字典作为图像的稀疏变换基,通过稀疏贝叶斯学习算法获得稀疏解.实验结果表明,基于自适应字典的贝叶斯学习算法能提高稀疏化,明显改善图像的重构质量.
Bayesian compressive sensing （BCS）, a kind of compressive sensing algorithm based on statistical analysis, uses information correspondence to get robust performance in image reconstruction. But it depends on image sparsity strongly. In order to solve further level sparsity of BCS, this paper presents a novel self-adaptive Bayesian compressive sensing algorithm combined with redundancy self-adaptive dictionary learning. The algorithm firstly decomposes an image into local patches and builds the dictionary from the iterating transition image. Then the image is represented by this dictionary space. Finally, the image is reconstructed using the sparse Bayesian learning algorithm. Experimental results show that the proposed algorithm obtains deep sparse representation and improves image reconstruction quality.
出处
《西安电子科技大学学报》
EI
CAS
CSCD
北大核心
2016年第4期1-4,122,共5页
Journal of Xidian University
基金
国家自然科学基金资助项目(61271296)
陕西省自然科学基础研究计划资助项目(2016JM6012)
中央高校基本科研业务费专项资金资助项目(JB150218)
西安电子科技大学教育教学改革研究资助项目(B1311)
西安电子科技大学新实验开发与新实验设备研制及实验教学改革资助项目(SY1354)
关键词
稀疏贝叶斯学习
自适应字典
贝叶斯压缩感知
sparse Bayesian learning
self-adaptive dictionary
Bayesian compressive sensing
