The Study On Face Detection System Based On Color Image Sequences

A study on the algorithms to face detection is presented in this paper. Face detection has been a research focus of computer vision pattern recognition and human -computer interaction for a long time .It also has been applicated in a wide range of fields such as automatic face recognition , visual surveillance and content-based retrieval. However, face detection is a most challenging task because of the complicated pattern and the frangibility of human face. So there is some weakness such as large computation, low efficiency and many false reports among the detection result in most face detection methods.In this paper, skin color information application was mainly researched in face detection algorithms based on some existed methods. Color information is used in segment application. Skin color in human face recognition is especially important. Skin color is obviously shift-invariant, rotation-invariant and even emotion-invariant, and therefore very stable.Skin color usually is far different from background object. However, there exists the possibility that some background color components are similar to skin color. This may leads to false recognizing. A way of face detection algorithms based on inter-frame-differential and skin colorspace was proposed and acted out. That is, human face can be extracted from complex background by means of moving and skin-color information.To improve the performance of face detection, the contributions were made as follows:(l)The moving information to calculate the binary inter-frame differential image and extract edge and locate a gross position of the target was utilized.(2)The skin color model and using the color space transform to accurately locate human-face was designed.(3)Based on human face property, face component geometric distribution relationships and lips color locate and extract features. Experiment results show as follows:(l)Using inter-frame-differential algorithm can remove possible disturbing background in complicate environment.(2)Using skin-color can further restrict the interest region to face.(3)Doing feature extraction after the two-step preprocessing could not only remove environmental noise and decrease false identification, but also, as it largely down-size the searching-area, it actually increased the total processing speed.Generally, this is a stable, simple and applicable algorithm , so it has some research and application value in face recognition.