Multimodal Biometric System using Texture Features

In this paper a divided image based Local Gabor Binary Patterns is proposed for image description in multimodal biometrics. The combination of Gabor transform and Local Binary Pattern provides the robustness to variations in the input image. Two well known and widely used biometric traits viz. Face and Finger print are considered. Firstly the images are divided into non over lapping blocks, each of the blocks are treated separately and LGBP is applied over each block. The normalized histogram bin values of each block forms the image descriptor. Feature level fusion of the biometric traits is employed and the classification is carried out using KNN classifier. The modal was trained and tested on SDUMLA-HMT, a homogeneous multimodal biometrics database. The results are compared with multimodal biometric system using LBP and Multimodal Biometric system using Galois operator. The results indicate significant improvement in the recognition ratio of the proposed method.