Collaborative Representation Based Fisher Discrimination Dictionary Learning For Image Classification
Rongrong Yao, Yintao Song, Baoqing Yang, Xinfeng Zhang, Yu Zhao · 2023
Recent developments of deep neural networks based image classification have attracted much attention because of their top performance. However, the volume of data for training seriously affected the classification effect of neural networks, thus it remains an interesting problem to estimate the accuracy of classification efficiently and accurately on a small-size dataset. In this article, a novel collaborative representation dictionary learning is presented, named collaborative representation based fisher discrimination dictionary learning (CRFDDL), in which under collaborative representation, this work solves the coefficients efficiently so that efficient learning model can be obtained. In consideration of the fact that both of the reconstruction residual and coefficients of each sample are discriminative, the classification scheme for fusing these two discriminative clues is naturally adopted for the image classification. Experiments shows that the proposed CRFDDL achieves competitive recognition rate compared with the existing classic image classification methods.