Principal Component Analysis Method for Muitiplicative Noise Removal
姚莉丽 YAO Lili, 冯象初 FENG Xiang-chu, 李亚峰 LI Yafeng · ACTA PHOTONICA SINICA · 2011
The further application of Radar image system relies on the quality of denoising from images.By analyzing the existing denoising algorithms,a new algorithm was presented using principal component analysis for removing multiplicative noise,based on local similarity of images.Multiplicative noise by logarithmic transformation could be converted into the additive noise for processing.Type analysis of the noise in the logarithmic domain was given.In the image logarithm domain,training sample blocks were selected by nonlocal method,and the principal component analysis was used to extract the main features of image blocks.A threshold principle,was proposed by linear minimum mean-square error estimate,which adapted to the signal message.The denoising images were obtained by biased estimation.Experiment results show that the presented method is valid.Compared with the existing variational methods,the new method has higher peak signal to noise ratio and better visual effect.That the performance of the proposed method is practical at a certain extent.