Robust face and facial feature localization using the dual skin model
Computer School, China West Normal University, Shida Rd. 1, Nanchong, Sichuan, China
A fast and adaptive face and facial feature localization algorithm for colour images with sophisticated background is present. In this algorithm, a self-adaptive pre-processing method was provided to depress the colour bias and the high light. Then the CbgCbr-YIQ dual skin model was proposed to acquire the integrated skin similarity for improving the quality of skin segmentation and extraction. After the morphological post-processing, by using the Adaboost classifier and the information of spatial position, the facial feature positioning was fast realized finally. Experimental results showed the robustness and good performance of the proposed algorithm.