Research On Extracting Skin-Region And Face-Region In Direct From Color Image

As for the basis of face recognition technology, face detection is always research keystone of artificial intelligence and patten recognition in recent years. Face automatic detection takes on important value for applied and theorical research. Firstly, that steady and high-efficiency face automatic detection is the basis of all kinds of face iamge automatic-analysis system, such as face automatic recognition system, intelligent human-computer interface. Secondly, it can be used in many aspects, such as intelligent-scene monitor, face tracking. Moreover, it is a typical representation for non-rigid complex visual object detection academically. Its conclusions and methods can be used for the problems of other complex-visual object detection.On the basis of discussing the domestic and international current status of face detection technology, there are two research emphses on establishing chroma table methods of color image and presenting a new method of face region from a completely new perspective. Firstly, the paper presents a new method for establishing chroma-table called fuzzy enhancement skin color chroma-table algorithm. The method based on the fuzzy character of chroma distributions, uses fuzzy enhancement process into skin chroma probability distributions which are collected from chroma space and then makes the fuzzy enhancement result into skin chroma-table. Secondly, it presents a new method of extracting face-region in color image directly. At the phase of extracting skin region, using the improved region growing algorithm to extract the borderline and then using face-target-extracting operator to extract the skin regions accoding to these borderlines without establishing the skin model and skin chroma-table. At the phase of the face region confirmability, using the other operater called face-region-confirming operator to confirm the face regionswithout converting gray image. Finally, the effectiveness of the algorithms proposed is validated through experiments.Fuzzy enhancement skin color chroma-table algorithm presented by this paper can resovle the traditional chroma-table algorithm disadvantages which are poor in distinguishable capability of"face-nonface"and have high error possibility. A new method of extracting face region in color image directly presented by this paper provides a new view of face detection from color image and has more highly detetion rate and more widely adaptability.