Eye Localization Based on Maximal Complexity of Rotating Scan Lines

A novel algorithm based on maximal grayscale complexity of rotating scan lines is proposed to locate the human eye in the detected human face region. First of all, the human face regions are detected from a number of outdoor photos using the skin segmentation algorithm. Secondly, the skin block is retained in the human face region employing the skin block extraction algorithm. Finally, the grayscale complexity is computed by rotating scan line in these skin blocks and the human eyes are located at the straight line corresponding to the maximum of grayscale complexity. The proposed algorithm has a good robustness of the variation of low-resolution faces with head postures and accessories and it is much better than other approaches.