Detection Human Faces In Color Images

An approach of human face detection in color images based on skin color feature and face verification is proposed. Firstly, input color image is segmented into skin-like and non skin-like regions in HSI and YCbCr space separately, then using logic 'and' operation to combine them. Secondly, the small objects in the background are removed according to the area, the erosion operation is used to cut the thin connection between skin-like regions. Thirdly, face verification are performing based on a set of rules about the shape and size information of the rectangle around human face, and thus the candidate face regions are selected. Finally, the components of two eyes and mouth in the face regions are used to locate the human faces in the candidate face regions.Light compensation is presented to eliminate color bias. Three skin color model in the HSI and YCbCr color space are constructed and their advantages and disadvantages are analyzed, the strategies that combine the different images after color segmentation is given. The thin connections between skin regions are cut using erosion and how-to-use erosion in human faces detection system is discussed also.The experiments show that the combination of the images after color segmentation based on different skin color models is efficient and that the lessening of the noisiness in the blank and white images is useful.What's more the restriction to this method is small. The system can locate one or more human faces that have small rotation or not in color images. The system has high speed and low misdetection rate.