Discriminant projection method based on optimized neighborhood graph for image feature extraction

摘要

The image data collected in reality often has high dimensions, and it contains noise and redundant information. Therefore, it is necessary to extract the compact feature expression of the original perceived image. Based on the above considerations, this paper proposes discriminant projection method based on optimized neighborhood graph model based on graph optimization. The model has the following characteristics: (1) Graph constraint is introduced to maintain the local geometric structure of the data while using the label information and global information. (2) The $L_{2,1}$ norm is used to redefine LDA, and the scale factor is introduced, which improves the robustness of feature learning. The validity and robustness of the proposed algorithm are verified by experiments in 2 image datasets.