A Study Of The Restoration Method Of Fuzzy And Broken Face Images

Human face image processing are gaining popularity and diverse applications in many domain with the rapid development of image processing and pattern recognition.The task of the research project is looking for an effective method to deal with the broken and fuzzy face images. On one hand, the image would be clarified when it's lesser and fuzzy. On the other hand, it would be adaptive restoration when it's broken. The research comes up against a series of problems owing to disadvantageous influence, such as illumination condition, the background of an image, the pattern of face and so on. The research of the restoration process is challenging.In order to answer the purpose as the above, the paper designs a face image restoration system. The main contents of the system are as follows: Firstly, the image is preprocessed by noise elimination through smoothness and light compensation. The precise location of human face can be detected automatically according to color information and gray-level projection. Secondly, the face rotation degree can be corrected through the iris locating operation. In order to obtain a standardizing face image, the face image is tailoring and zooming according to parameters above. Then, obstruction should be gotten rid of. Thirdly, the positions of eyes, nose and mouth can be demarcated by corresponding methods on the basis of cursory localization. At the same time, a library of five sense organs image is created. Binary images are selected to plus into source image respectively. At last, the image would be restored by adjusting parameters adaptively.The experiment show that the skin-model can reduce computational complexity and the gray-level projection can locate face images practicably and accurately. The combined method based on color information and pixel distribution features for automatic face detection is practicably and efficiently. On the stage of feature extraction, the robustness of the features extraction algorithm which is combined with knowledge and color is very well. The computation of dynamic threshold has little insensitivity in light variations. Theselection of adaptive weighting factor can cut down manual intervention and exact the restoration process.The brow of face would influence the feature extraction. In order to surmount the limitation of brow, the experiment would realize fast pose-varied face detection in a complex background. Otherwise, the experiment would penetrate into other obstruction and the shortage of an organ such as nose.