Fusion of probabilistic collaborative and sparse representation for robust image classification
Zhangdan Chi, Shaoning Zeng, Jianping Gou · 2017 International Conference on Security, Pattern Analysis, and Cybernetics (SPAC) · 2017
The image representation model determines the robustness of image classification. The sparse model obtained by Probabilistic Collaborative representation based Classification (ProCRC) calculates the probability that a test sample belongs to the subspace of classes, to find out which class has the most possibility. Previous studies showed that the distances obtained by different models may have some complementary in the image representation. For this motivation, we proposed a novel image classification method that fusing two distances obtained by ProCRC and conventional sparse representation based classification (SRC). Therefore, we named it ProSCRC. In the fusion, a weight factor A was introduced to balance contributions from the two distances. In order to evaluate the robustness, we conducted plenty of experiments on prevailing benchmark databases. The experimental results showed that our method has a higher accuracy in image classification than both ProCRC and SRC.