A novel infrared face recognition method in DCT domain

A novel infrared face recognition method using blood perfusion conversion and DCT is proposed in this paper. Firstly, the blood perfusion image instead of thermal image is used to get the stable biological features. Secondly, due to the low-resolution of the infrared images, the feature extraction method (Linear Discriminative Analysis in DCT domain) is chosen to get the principle features in the blood perfusion image, which is suitable for real-time application. A new feature selection algorithm in DCT domain is applied to lessen the small sample problem of LDA. The experiments illustrate that the feature selection algorithm in DCT domain is effective in getting the discriminant features and the proposed method has better performance compared with the traditional DCT and LDA method.