Histogram based face recognition system

Abstract

A practical face recognition system using HSI (hue, saturation, and intensity) color space based on color histogram matching is described. The threshold for hue and saturation is found by using 450 face samples from the FERET dataset and 1500 face samples from the Essex University database. Assuming that all face images in a dataset are of the same size, eigenfaces are obtained as the eigenvectors of the covariance matrix of the data points. The histogram-based post invariant face recognition study shows that a typical monochrome image with 8-bit representation has 256 grey levels. The performance of each color channel is found to be different and this observation allows combining the results of different color channels. which may increase the probability of making the correct recognition of a face image.