Wavelet-Based Multi-Level Image Matching with Detail Measure Weight for Face Recognition under Varying Illumination

With illumination varying condition, face features gotten from image is distorted nonlinearly by variant lighting intensity and direction, so face recognition becomes very difficult. According the "common assumption" that illumination vary slowly and the face intrinsic feature (including 3D surface and reflectance) vary rapidly in local area, we can consider that high frequency features represent the face intrinsic structure. As the popular method, wavelet can decompose the image into multi-level detail images representing high frequency features and analogy image representing low frequency features. But we can\´t make a quantitative analysis that how many detail features can be used for eliminate illumination variation. So we propose two measures to quantify the detail features, and with these measure weights, we can do wavelet-based multi-level detail image matching for face recognition under vary illumination. With PCA, the experiments based on Yale face database B and MU PIE face database show the method this paper proposed can get remarkable performance^.