An effective approach for finding periodicity of a subject in video data

Video based Face Recognition is an emerging research issue which has received much attention during the recent years. In this research, an effective approach for calculating the periodicity of a subject i.e. exact appearance of a subject in different time in video data stream is presented. The system is the combination of two studies: face detection and face recognition. The face detection is performed on video frames. There is a study and implementation of Local Binary Pattern for (4,.5), (8,1), (8,2), (16,2) and (24,3) operators where first value defines neighboring pixels and second denotes radius from centre pixel to neighbor pixels. LBP, HOG and Gradientface methods are implemented for comparing the results and also to compare to show how well these methods can handle variations in expression, pose and illumination. Finally the efficient approach evolved that gives the most effective results 92.3 % result using LBP(24,3), 97 % result using HOG and 100% results by using Gradientface method for captured videos under considerations. For noisy images, Gradientface has achieved 95.7 % result which shows that the method is robust to noise in comparison to LBP and HOG.