UMLBP-A Novel Approach for Face Recognition System using OPENCV
Manish Kumar, Rahul Gupta, Dinesh Kumar, Kota Solomon Raju · 2018
Several face recognition algorithms have been proposedin the last decade which provide a biometric authentication process for various utilities. In the present work, a novel algorithm for face verification considering both lightening and shape information for representing the face images, has been presented. In this approach, the region of interest in face is first divided into 8x8 regionfrom which Uniform Mean Local Binary Pattern (UMLBP) histograms are extracted and concatenated into asingle histogram with enhanced feature efficiently representing the face. The matching is done using a K nearest neighbor classifier with minimum error based similarity measure. A number of performed trials reveal that the given approach is better overall considered methods (basic LBP, MLBP and ULBP methods). All experiments have been performed on ORL database which include evaluating the efficacy of the approach over different face angles, illumination and rotation of the query image. Testing was performed using 10 cross validation scheme and average results have been considered. Recognition rate has been recorded as 90 percent on 3 face training of single person and 99 percent on 9 face training of single person. The proposed method also allows for the fast feature extraction and testing for a given platform using OPENCV.