Photometric Linearization by Robust PCA for Shadow and Specular Removal

Abstract

In this paper, we present an efficient method to remove shadow and specular components from multiple images taken under various lighting conditions. We call the task photometric linearization because Lambert’s law of diffuse component follows linear subspace model. The conventional method[1] based on a random sampling framework make it possible to achieve the task, however, it contains two problems. The first is that the method requires manual selection of three images from input images, and the selection seriously affects to the quality of linearization result. The other is that an enormous number of trials takes a long time to find the correct answer. We therefore propose a novel algorithm using the PCA (principal component analysis) method with outlier exclusion. We used knowledge of photometric phenomena for the outlier detection and the experiments show that the method provides fast and precise linearization results. Additionally, as an application of the proposed method, we also present a method of lossless compression of HDR images taken under various lighting conditions.