Hair-color modeling and head detection

A method based on hair-color modeling to detect human head in color images was proposed. The presented algorithm compared the performance of four different chrominance spaces for the distribution of hair color and built a human hair color model in YCbCr space. A Gaussian mixture density model was used to describe the distribution of hair color and segmentation threshold value in Cb-Cr plate is obtained though parameter estimation using Expectation-Maximization algorithm. By combining chrominance information with other features of head, color segmentation and subsequent detection of human head detection have been performed in two-dimensional static images. Experimental results show that the presented method can detect human head in complex background effectively and provide a new approach of human head detection.