Sparse and collaborative representation based kernel pairwise linear regression for image set classification

Abstract As an important part of expert and intelligent systems, image set classification has been widely applied to many real-life scenarios including surveillance videos, multi-view camera networks and personal albums. Compared with single image based classification, it is more promising and therefore has attracted significant research attention in recent years. Traditional pairwise linear regression classification (PLRC) introduces the unrelated subspace to increase the discriminative information, and it shows a demonstrated better performance on image set classification. However, the unrelated subspace constructed by PLRC is not optimal and PLRC may fail for well classifying image sets that are not linear separable, or when the axes of linear regression of class-specific samples of different classes have an intersection. In this paper, two new unrelated subspace construction strategies are proposed based on sparse and collaborative representation, respectively. Then, based on them, a new image set classification framework, kernel pairwise linear regression classification (KPLRC) is developed. KPLRC is a nonlinear extension of PLRC and can overcome the drawback of PLRC. Specifically, KPLRC embeds the related and unrelated gallery sets and probe sets into the high-dimensional Hilbert space, and in the kernel space, the data become more linear separable. Extensive experiments on four well-known databases prove that the classification accuracies of KPLRC are better than that of PLRC and several state-of-the-art classifiers. These quantitative assessments reinforce the significance as well as the importance of embedding the proposed method in other intelligent systems application areas.