Digital Curvelet Transform for Palmprint Recognition

In this paper, we present a new feature extraction method for palmprint recognition The digital curvelet transform is revised here and used to extract the palmprint features In our algorithm, we use the discrete Meyer wavelet transform to replace the “a trous” transform, then apply the ridgelet transform to each block which is subbanded after the discrete Meyer wavelet transform from the palmprint image Our work is carried on the PolyU Palmprint Database Dealing with the palmprint image sized of 64 × 64, our new strategy acquires 4 × 128 × 128 curvelet coefficients Based on the system performance, the best coefficients threshold can be obtained With this threshold the curvelet coefficients are filtered and less than 2% of coefficients are selected With this compressed coefficients set, the correct recognition rate of our palmprint identification experiment is up to 95.25%.