Adaptive PCA-SIFT Matching Approach for Face Recognition Application
In this paper, the authors present a novel human face identification approach. This approach consists of three parts: de-noised face database, Adaptive Principle Component Analysis based on Wavelet Transform (APCAWT), and the Scale Invariant Feature Transform approach, (SIFT). The main idea is to extend SIFT features by using a APCAWT on compressed and de-noised ORL database, JPG file format is used for compressing and double wavelet filters (Bior 1.1 and Haar both are at level 10 of decomposition) is used for denoising process. For feature extraction the Eigen face of PCAWT entered to SIFT algorithm, and thus only the SIFT features that belong to clusters, where correct matches may be expected are compared according to a specific threshold.