An DLMNN Classifier Based Efficient Pose Invariant Face Recognition System

Abstract

Human face identification and also recognition together are the most prominent areas in image processing and analysis. Amongst the numerous issues that affects the face recognition (FR) technique; Pose and Illumination are the two major challenges. In recent years, many researches focused on FR system as the existing methods provide very less accuracy. To overcome such difficulties, this paper proposed a DLMNN classifier based efficient Pose Invariant Face Recognition (PIFR) system. The suggested system encompasses of the succeeding steps. First, the input video is transmuted into frames and subsequently the unwanted frames are eliminated using DETC method. Then, the noises are removed from the converted frames using IADF method. Next, the facial points are extorted from the pre-processed image using Modified SDM method. Next, the extracted facial points are offered as input to the DLMNN classifier. In DLMNN, the modification is done by using Entropy-CSA which classifies the frames into recognized face and non-recognized face. Experimental results ascertained that the proposed system provides better performance than the existing system.