Discriminative Competitive Representation for Image Classification
Zhongli Ma, Xue Qinlin, Tao Jiang, Jing She, Zuoyong Li · 2019
For collaborative representation based image classification methods, it is very important to represent the image well. In this paper, we propose a novel representation method called discriminative competitive representation (DCR) for image classification. It divides the set of all training samples into two competitive subsets to represent a test sample, and then uses the representation result to perform classification. This method can obtain an approximate representation of a test sample and enable representation components to be discriminative and competitive, which is beneficial to correct classification of the test sample. Specifically, the main idea of the DCR method is that it makes test set and training set have low correlation. One notable advantage of the DCR method is that it exploits a mathematically tractable way to obtain discrimination of classes. The image classification experiments show that the DCR method can obtain a higher accuracy than conventional sparse representation methods.