Texture Based Palmprint Identification Using DCT Features

In the palmprint recognition application utilizing more information other than principle lines or minutiae is much helpful. In this paper we proposed discrete cosine transform (DCT) based feature vector for palmprint representation and matching and compared with DFT and wavelet transform. Here the central part of the palmprint image of size 128 times 128 is resized to the size of 64 times 64 and divided into four non overlapping sub-images. The transform is applied on each sub-image directly without any preprocessing. By dividing the transformed sub-image into nine blocks, standard deviation is calculated for each block and such in total 36 (9 times 4=36) standard deviations will form the feature vector. This feature vector is used in matching stage. Total 10 images per person are taken from standard database available. Training set is prepared with the help of k images where k varies from 1 to 8. Results are checked against remaining images image in identification mode. Results are represented in terms of Genuine acceptance rate(%). In identification mode 97.5% recognition rate is obtained. The work is preliminary but recognition rate is promising and without any pre-processing.