joint lowrank and sparse principal feature coding for enhanced robust representation and visual classification.ieee trans image process10.116文档.pdfVIP

IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 25, NO. 6, JUNE 2016 2429 Joint Low-Rank and Sparse Principal Feature Coding for Enhanced Robust Representation and Visual Classification Zhao Zhang, Member, IEEE , Fanzhang Li, Mingbo Zhao, Member, IEEE , Li Zhang, Member, IEEE , and Shuicheng Yan, Senior Member, IEEE Abstract — Recovering low-rank and sparse subspaces jointly I. INTRODUCTION for enhanced robust representation and classification is dis- N NUMEROUS practical applications, most real-world cussed. Technically, we first propose a transductive low-rank and sparse principal feature coding (LSPFC) formulation that Idata can be characterized with high-dimensional attributes decomposes given data into a component part that encodes low- or features. But high-dimensional data often have unfavorable rank sparse principal features and a noise-fitting error part. features, redundant information or (grossly) corruptions, so the To well handle the outside data, we then present an inductive study on how to recover the original data accurately by feature LSPFC (I-LSPFC). I-LSPFC incorporates embedded low-rank learning or low-rank/sparse coding has been extracting much and sparse principal features by a projection into one problem for direct minimization, so that the projection can effectively map attention in the past years. Feature learning aims to find a both inside and outside data into the underlying subspaces to transformation or mapping to conver