基于人脸检测的自动红眼消除算法研究-计算机应用技术专业论文.docxVIP

Abstract Abstract Redeye is a phenomenon that human eye pupil in photo shows a red dot in the flash mode. The color of red-eye appears quite different from that of eyes perceived by people. It often leaves regret to photographers and is regarded as a faulty stroke. As the photo opportunities are often unable to repeat, eliminating redeye in photos is essential. This paper proposes an automatic redeye removal method based on face detection. The method is divided into three parts: face detection, redeye detection, redeye removal. In face detection part, with the consideration of calculation speed and robustness, skin segmentation technique is used to obtain skin areas in the image at first. Then, skin areas are judged by the following methods: calculate length-width ratio of the region, calculate region size and determine whether there is hair on the region, to screen out the possible skin area which contains human face initially. After that, the skin area is detected by Support Vector Machine method and the face area in the image is obtained. Redeye detection is to locate accurate position of redeye in the face region. First, redeye area is marked roughly according to the geometric features of human face and red degree. Then, redeye region is processed using mask algorithm and pupil location algorithm, the exact location of red-eye is gotten. In redeye removal part, adjustment of three channels is made in RGB space of redeye pixels. In order to recover the actual color of pupil region, the value of R channel is decreased while the values of G channel and B channel are tuned as well. At last, a smooth of edge pixels in redeye region is made to restore natural color of pupil. In this paper, an automatic redeye removal system based on the above theory and method is realized, and a large number of experiments and tests are conducted. The results show that the method proposed in the paper can not only eliminate redeye phenomenon, but also improve processing efficiency, an