An algorithm for face determination based on convolution and average face

Abstract

This paper presents an algorithm for face determination by using the method of convolution and average face template. Firstly, human face images are used to synthesize average face template. Then convolution is applied to each detecting image to get the convolution value. Finally, the threshold decision principle is applied to determine whether the detecting image is human face or non-face. Three different templates of frontal, right, and left are used to test the algorithm. Experimental verifications are conducted to evaluate performance of the algorithm on RGB, Gray, and YCrCb color spaces. The results show that the proposed method can achieve reliable face judgment and R component in RGB has the best effect, and the correct determination rates of front, left, and right face datasets are 96.9%, 98.2%, and 99.1%.