一种基于新型KPCA算法的视频压缩感知算法  

A Video Compressed Sensing Algorithm Based on Novel KPCA

作　　者：钱阳[1] 李雷[1] 

机构地区：[1]南京邮电大学理学院,江苏南京210023

出　　处：《计算机技术与发展》2015年第10期101-106,共6页Computer Technology and Development

基　　金：国家自然科学基金资助项目(61070234;61071167)

摘　　要：针对具有帧间相关性的视频信号的压缩感知问题,文中依据核主成分分析(KPCA)变换能量集中的特性,将能量值较低的变换系数去除,实现视频信号在KPCA变换下的稀疏表示,并验证了其用于压缩感知算法的可行性。考虑到KPCA特征提取时存在如何根据具体问题选择最优核函数的问题,在传统文化算法的影响函数中引入自适应变异算子,形成一种自适应变异算子文化算法(AMOCA),并将其与KPCA算法结合起来用于训练核参数,有效地提高了KPCA应用中核函数的优化选择。大量仿真对比实验表明,文中算法能有效消除视频帧间相关性,具有更高的视频重构质量以及更好的性能。Aiming at the compressed sensing problems of video signal, which has a strong inter-frame correlation, remove the lower trans- form coefficients according to the energy concentration characteristics of KPCA transform. Therefore, the sparse representation of the vide- o signals in the form of KPCA transform is achieved and the feasibility of the transform being used in compressed sensing is verified. Tak- ing into account the problem of how to choose the best kernel function according to the specific problems when KPCA applied to extract nonlinear feature components,adopt an adaptive mutation operator in the influence function of traditional culture algorithm,forming an A- daptive Mutation Operator Cultural Algorithm （AMOCA）, and then combine it with KPCA to train kernel function. Most comparative simulation results show that the proposed algorithm can effectively eliminate the inter-frame correlation of the video sequence with higher reconstructed quality and better performance.

关 键 词：压缩感知 文化算法 核主成分分析 帧间相关性 稀疏表示 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]