Research on clustering-weighted SIFT-based classification method via sparse representation

In recent years, sparse representation-based classification (SRC) has received significant attention for its high recognition rate. However, the original SRC method requires rigid alignment. By further considering the robustness of scale and affine in this paper, we explore the relationship of the similarity of the SIFT descriptors to a recognition task and propose a clustering-weighted SIFT-based SRC algorithm (CWS-SRC). The SIFT descriptors extracted from the samples are first clustered according to similarity. Next, the weight of each feature is calculated for a weighted classifier. Finally, the SRC method is operated on the SIFT descriptors extracted from a probe image, and its identity can be implemented via the weighted classifier. Using two public face databases (AR, Yale face database) and a self-built car-model database, the performance of the proposed method is evaluated and compared with that of SRC, SIFT matching and MKD-SRC method. The proposed CWS-SRC exhibits better performance for sufficient samples in the misalignment scenario.