Face recognition method based on gabor multiorientation fusion feature and singular value decomposition

摘要

Traditional Gabor feature and singular value decomposition (SVD) exist the problem of high dimensionality in characterizing facial features, this paper presents a facial feature extraction method that combine the fused multi-directional Gabor features and SVD of the image, which reduce the number of feature dimensions under the premise of feature information-rich. Firstly, use performance-optimized Gabor filters whose DC component is compensated to extract the multi-scale, multi-directional characteristics of facial images, and integrate the same- scale Gabor feature in different directions as the face image local features; Then extract the SVD feature of the image as the global features of face images; Finally, combine the local features and global features to characterize the primitive face image. Experimental results on ORL face database show that the recognition rate of the proposed method is up to 98.25%. The proposed method has advantages over traditional face recognition method based on Gabor features and SVD in the recognition rate and computational efficiency.