Abstract: Finger vein recognition technology, also known as a biometric identify recognition technology, is more superior to fingerprint identification technology and has a broad application prospect. KPCA is a nonlinear feature extraction method, and it overcomes the drawback that linear feature extraction method cannot effectively deal with the higher-order statistics information and nonlinear correlation between pixels. K2DPCA can solve the problem of excessive computation in the one-dimensional matrix operation, but it would require more coefficients for image representation. So the compression effect is far less than the one-dimensional method. Based on KPCA、LDA and MMC,this paper studies the images of the vertical and horizontal directions respectively, obtaining optimal results on the finger vein recognition.