The Afresh Transform Algorithm Based on Limited Histogram Equalization of Low Frequency DCT Coefficients

ABSTRACT

In reality, face recognition due to illumination, expression, posture, or other objects caused by facial shade seriously affects the recognition rate of the problem, put forward the afresh transform algorithm based on limited histogram equalization of low frequency DCT coefficients. Firstly, divide the image into several non-overlapping local small fragments, and then use limited histogram equalization for local contrast stretching to realize and get rid of the local sub-block noise. Then, eliminate the illumination change in face image by reducing appropriate number of low frequency DCT coefficients. Finally, use kernel principal component analysis for feature extraction, and the nearest neighbor classifier to complete the final face recognition. In ORL, extension Yale B is used in one experiment to verify the effectiveness and the robustness of the proposed algorithm on outdoor facial database. The experimental results show that this algorithm, in dealing with robust face recognition, has achieved higher recognition rate as compared with several linear algorithm.

Keywords:

Adaptive histogram equalization, Coefficient afresh, Face recognition, The low-frequency discrete cosine transform.