Evaluation of image enhancement quality measure in Robust PCA for Image Specularities Removal

Recent studies in matrix rank minimization problem has shown that a data matrix can be decomposed into low-rank and sparse matrix convex programming. This pattern of decomposition has wide range of applications, and also known as a form of Robust PCA method. In this paper, we focus on the face recognition application where the objective is to remove specularities which corrupt the images. This paper emphasizes that by using a quantitative image enhancement quality measure, an optimal low-rank approximation is obtained with lower dimensionality, shorter computational time, and comparable approximation quality compared with the approximation performed by the convex program.