基于图像特征改进的正交匹配追踪算法  

The Improved Orthogonal Match Pursuit Algorithm Based on Image Characteristics

作　　者：王玲[1] 田勇志[1] 王俊俏[1] 臧华平[1] 刘晓旻[1] 赵楠楠[1] 陈宝鑫[1] 梁二军[1] 

机构地区：[1]郑州大学物理工程学院,河南郑州450001

出　　处：《郑州大学学报（理学版）》2015年第2期73-77,共5页Journal of Zhengzhou University:Natural Science Edition

基　　金：国家自然科学基金资助项目;编号61307019;河南省科技厅重点科技攻关项目;编号132102210396;河南省教育厅科学技术研究资助项目;编号13A140693;14A14000;郑州市科技创新团队基金资助项目;编号112PCXTD337

摘　　要：将压缩传感理论应用于成像是光场成像理论的热门研究方向,由此可以设计出更简单、便宜、小巧的光学系统.正交匹配追踪算法是压缩传感理论的重要重构算法,它在重建图像时隐含着整幅图像权重相同的思想,没有体现出图像的固有特征,例如行列突变的剧烈程度,以及经过快速傅里叶变换基、离散余弦变换基、离散小波变换基作用得到的小稀疏系数代表图像的细节、大稀疏系数代表图像的轮廓的特点.使用上述3种变换基作用图像时,可以针对正交匹配追踪算法的固有缺点,提出合理选择逐行或者逐列重构图像和使用自适应迭代次数重构图像两种改进方法.仿真结果表明,改进算法明显提高了图像的质量,能够得到更好的图像视觉效果.Compressive sensing to images has become a hot topic in light field imaging theory due to that it can be applied in simple,cheap and small light systems. Orthogonal match pursuit algorithm was an important recovery algorithm in the compressive sensing theory. It assumed that the weight of the whole imaging was the same when reconstructed,ignoring the inherent characteristics of the imaging,such as the intensity of the mutations among lines and columns,and ignoring the characteristics that the small sparse coefficient represented the detail of the image and the big sparse coefficient represented the outline of the image. The sparse coefficient was acquired by the operation of fast Fourier transform basis,discrete Cosine transform basis,or discrete wavelet transform basis. Two methods were put forward to solve this problem,that is,reasonably choosing the line by line or column by column in reconstruction,and using adaptive iterations to reconstruct image,if the image was affected by the three kinds of transform basis mentioned above. Simulation results showed that the improved algorithm could significantly improve the quality of image.

关 键 词：光场成像 正交匹配追踪算法 压缩传感 稀疏表示 图像重建 

分 类 号：O438[机械工程—光学工程]