BLOCK WISE 3D PALMPRINT RECOGNITION BASED ON TAN AND TRIGGS WITH BSIF DESCRIPTOR

Faced by problems such as lack of robustness from 2D palmprint recognition system which can result to be attacked using a fake palmprint or having the same palmprint as another individual, 3D can present an alternative solution to deal with this problem, hence in this paper we are going to introduce a novel approach based on 3D palmprint recognition system named TT-P-BSIF: first, a preprocessing technique based on Tan and Triggs method was applied on a 3D depth image in order to effectively and efficiently eliminate the effect of low frequency component and at the same time keeping the local statistical properties of the treated image. Then the processed image is divided into a regular number of blocks using two parameters (a and b), after that the Binarized Statistical local features (BSIF) has been applied on each block in order to extract the features vector. These vectors are all combined to produce one larger vector for each processed image. Afterwards nearest neighbor classifier is used to classifier the 3D palmprint images. To examine the proposed method, this latter has been evaluated on a 3D palmprint database that contains 8,000 samples, the obtained results were consistent and promising which proves that the introduced method can massively and effectively improve the recognition results. Therefore, this proposed work using Tan and Triggs method for preprocessing and BSIF for feature extraction was able to generate a recognition rate up to 99.63% and verification rate at 1% up to 100% with EER equals to 0.12%.