Image classification via structured dictionary learning
Bowen Lu, Songhao Zhu, Xuewen Ju · 2018
In the past years, dictionary learning has been widely used in many image classification applications. Although lots of dictionary learning methods have been proposed to deal with the issue of image classification, there exists much room for improvement. To further improve the performance of image classification, a novel structured dictionary learning method is here proposed. The proposed method based on the Fisher discriminative criterion aims to learn category-specific dictionaries for all categories in the lower level simultaneously to obtain the specific information of each category, and learn category-shared dictionaries between all categories in the upper level to obtain the common information across different categories. Experiment results on public available databases to specific category tasks are conducted to evaluate the effectiveness and the superiority of the proposed method compared with the state-of-the-art dictionary learning for category methods.