Performance Evaluation of PCA Based Back Propagation over PCA Based Euclidian Distance for Video Images

Abstract

Key frame selection aims at reducing amount of data and retrieve information desired from a video. Video summarization aims at reducing the amount of data in order to retrieve information from a video. In this paper, we present an innovative approach for key frame selection; and a face detection and recognition from video sequence. For face detection from video, first we select the key frames and then detect multiple faces from key frames. The key frames are selected by using the Canny edge difference between two consecutive frames of the video sequence. Face detection from key frames are performed using Viola–Jones algorithm. It processes images extremely rapidly and provides high detection rates. For face recognition, principal component analysis (PCA) algorithm with Euclidian distance is presented. PCA is mainly used for feature extraction and applies linear projection to the original image space for dimensionality reduction. Here we present a comparison between PCA using Euclidian distance, and PCA based BPNN face recognition. The face features are extracted from database trained images using PCA and recognition of these images by using Euclidian distance and BPNN. For face recognition, we have used YALE face database. In this case, PCA based BPNN shows better performance than PCA based Euclidian distance and the performance decreases with the increase in the number of training images.

Cite this Article
Nazmul Shahadat, Dardina Tasmere Tonu, Shyla Afroge. Performance Evaluation of PCA Based Back Propagation over PCA Based Euclidian Distance for Video Images. Journal of Image Processing & Pattern Recognition Progress. 2016; 3(1): 24–31p.