PDEs对四种图像降噪降维方法的影响

Effects of PDEs for Four Noise Reduction and Dimension Reduction Methods of Images

针对高维噪声图像(光暗、有遮挡)的分类问题,分类质量的好坏不仅依赖于图像降噪方法的选择,同时还依赖于图像降维方法的选择.目前,大部分的降噪方法不具备降维作用,同样地,大部分的降维方法也不具备降噪作用.为了在降噪降维的同时还能保持良好的分类精度,本文提出了基于偏微分方程组(PDEs)的主成分分析(PCA\PDEs)、基于PDEs的线性判别分析(LDA\PDEs)、基于PDEs的典型相关分析(CCA\PDEs)以及基于PDEs的监督典型相关分析(SCCA\PDEs)四种降噪降维算法,并研究了PDEs的进化对四种算法的影响。

For the classification problems of high-dimensional noisy(dark and covered)images,the classification qualities depend on the selection of noise reduction methods and dimension reduction methods for images.However,the most of noise reduction methods for images have no dimension reduction function,and vice versa.In order to simultaneously noise reduction and dimension reduction with good classification accuracy,this paper presents four noise reduction dimension reduction algorithms based on partial differential equations(PDEs),that is,PCA\PDEs,LDA\PDEs,CCA\PDEs and SCCA\PDEs,and researches the effects of PDEs evolution for them.