An enhanced viola–jones face detection method with skin mapping & segmentation

Abstract

People count is one of the tedious tasks if it involves human intervene. As an alternative, automation of the process is done using image processing as it provides high accuracy. The face detector is based on a state-of-the-art cascade of boosted integral feature. This algorithm overcomes the drawbacks of Viola–Jones algorithm and relatively faster than Hog detection. The proposed system makes use of Haar-cascade classifiers to detect the face in real time, and this alone is error prone. These errors are eliminated by eliminating the non-face regions detected by mapping the skin colour of image with detected regions. Two different threshold methods are used to reduce the search space. Algorithm is evaluated on many databases. Experiments on these databases reveal good performance of the proposed algorithm.