期刊名称：International Journal of Computer Science and Information Technologies
电子版ISSN：0975-9646
出版年度：2014
卷号：5
期号：6
页码：8293-8296
出版社：TechScience Publications
摘要：Principal component analysis has been widely used in computer vision tasks. In image processing the outliers typically occur within the sample due to pixels that are corrupted by noise, alignment error, occlusion etc. The conventional PCA is based on the least square approach. However the least squares approach fails to account the outliers and produce the unreliable results. Many robust alternatives are proposed such as M estimator, MCD and S estimators. This paper makes an attempt to perform principal component analysis with most widely used these robust procedures. Also it is proposed a method which is based on MCD approach. The accuracy of the proposed method has been studied with help of an image along with existing algorithms.
关键词：PCA; Least squares; MCD; Robust PCA;Computer Vision.
