Image-based Jointly Sparse Maximum Margin Criterion with Pairwise Constrain for Feature Extraction

In this paper, we propose an image-based feature extraction framework. To our best knowledge, there are few methods to combine image feature extraction and pairwise constrain (PC) jointly and directly. We fill the gap between them. With the PC, we devise an objective function, taking the PC and image feature extraction consideration jointly. Besides, we introduce maximum margin criterion into the loss function, making all positive constrains and negative constrains take part in the process of feature extraction. Finally, the solution of the new framework can be deduced into two sperate individual problems. Comprehensive experiment results show our framework effective.