Enrollment Image Processing Research And Applications Of Human Face Detection

Face detection is an important research subject among computer vision and pattern recognition areas,as well as a key technology of dealing with face information.It plays an important role in face detection systems,video surveillance systems and content-based retrieval for images,and so on.As the constant expansion of the scope of application and development of the demand of actual systems,the researches of face detection have been paid more and more attention,and developed as an independent research.This thesis discussed face detection in still color images of simple background,which include a single face,proposed the de-nosing method of output feedback and the method of scanning the image partially to detect face,and designed an automation processing system for archives images with strong constraints;further more it made some researches on face detection in color images of complex background,which include multi-faces,and improved the method of face detection based on Adaboost algorithm,designed a face detection system for general color images.The main content of this thesis are as follows:In this thesis,we firstly summarized and analyzed the typical face detection algorithms;the method of single face detection based on color model in simple background images was focused on,In which the basic knowledge of color space,color model and image segmentation was introduced,and on this basis,the de-nosing method of output feedback and the method of scanning the image partially to detect face was proposed.Then,the method was successfully used in archives images automation processing system,which can significantly improve efficiency of setting up the files of students,and with a strong practical.As an expansion of face detection in images of simple background,this thesis then made some researches on how to detect multi-faces in color images of complex background quickly and accurately,and carried out a detailed introduction about Adaboost algorithm,include rect feature,classifier and how to training classifier and so on.At the same time,the thesis proposed a method that combined the image segmentation method based on skin color to test and verify the results of Adaboost algorithm.The face detection system based on this improved method can detect multi-faces in color images of complex background fast and accurately.It can achieve at not only a higher hit-rate,but also a lower false-alarm-rate,and is robust and practical.Because faces' complexity and the diversity of face poses and illumination,the systems in this thesis still generates some undetected and false detected faces.The following research will be how to made the un-detection rate and the false rate lower,and explore a more rapid and accurate face detection method.