The research based on the video sequence of face key frames extraction

Accurate, real-time as well as high fitness of face detection is always the difficult problem of Face Recognition System. The paper aiming at the problem come up with a solution which can extract key frames of face from the video sequence. The algorithm using skin color and AdaBoost face detection to construct a real-time, high-efficient face key frames extractor. It can reduce AdaBoost cascade classifier training intensity and enhance the efficiency of Face Detection. On the other hand, we combine 40 Gabor filter groups in different scales and directions to detect human eyes, and then verify the extracted face key frames that significantly improve the speed of Gabor eye location. From the above, our method can extract face key frames from video sequence in real-time.