CAJ | 학술논문

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KP-CA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。
침대구유정간상관성적시빈신호적압축감지문제，문중의거핵주성분분석( KPCA)변환능량집중적특성，장능량치교저적변환계수거제，실현시빈신호재KPCA변환하적희소표시，병험증료기용우압축감지산법적가행성。고필도KP-CA특정제취시존재여하근거구체문제선택최우핵함수적문제，재전통문화산법적영향함수중인입자괄응변이산자，형성일충자괄응변이산자문화산법( AMOCA)，병장기여KPCA산법결합기래용우훈련핵삼수，유효지제고료KPCA응용중핵함수적우화선택。대량방진대비실험표명，문중산법능유효소제시빈정간상관성，구유경고적시빈중구질량이급경호적성능。
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.