Online update techniques for projection based Robust Principal Component Analysis

Robust PCA is a modification of PCA, which works well on corrupted observations. Existing robust PCA algorithms are typically based on batch optimization, and have to load all the samples into memory. Therefore, those algorithms have large computational complexity as the size of data increases, and have difficulty with real time processing. In this paper, we propose a projection based Robust Principal Component Analysis (RPCA) in order to use RPCA as an online algorithm for real time processing. The proposed online algorithm in this paper reduces computational complexity significantly, although the proposed algorithm has negligible performance degradation compared to conventional schemes. The proposed technique can be applied to various applications, which need real time processing of RPCA.