Efficient person identification from periocular region using intelligent fusion of local and global features
Suneeta Verma, C. B. B. Singh · 2017
Efficient person identification is extremely desired in the digital era. One of the biometric techniques using the features of periocular region may provide efficient identification result. Periocular region is defined as surrounding of eye that is most discriminative region in nature for feature extraction in unconstrained and non cooperative environment. We have proposed an algorithm which defines region of interest from face images that contains ocular region. The nature of periocular image is so diverse that requires to be used in efficient identification. The proposed algorithm solve both extraction and matching problem in unconstrained environment where the captured image has several challenges like blur, de focus, improper illumination, pose variation, motion and occlusion. The algorithm extracts four features of diverse nature that are LBP, HOG, SIFT and SURF. By using feature reduction, the proposed system estimates of feature vector that is stored in database and later on used for matching during identification. The proposed algorithm performs efficiently and provides better result. We have performed our system on UBIRIS v2 database which contains images of real world in unconstrained environment. The result of the proposed algorithm is very efficient regarding each feature.