Face detection method based on multi-feature fusion in YCbCr color space

Based on multi-feature fusion, we introduce an accurate face detection method to complete face detection in a complex image background. First, the method pre-processes a color image based on self-adaptive luminance compensation and hybrid filtering technology. Then, the pre-processed image is converted from RGB color space to YC_bC_r color space. The converted image is segmented according to a skin color model. In order to acquire possible candidates of the face, median and morphological operations are used to merge or segment skin color areas and non skin color areas. Eyes are detected based on the matching relation between human eyes color and luminance. Mouth is detected based on the values of C_r and C_b. Finally, the face is accurately detected from possible candidates of the face through two eyes and mouth location. Experimental results using VC++ 6.0 show the proposed method has better adaptability, it can effectively complete face detection in a complex image background.