广东科技馆比特球展项图像处理的研究与实现-通信与信息系统专业论文.docxVIP

华中科技大学硕士学位论文 华 中 科 技 大 学 硕 士 学 位 论 文 II II Abstract This paper started with the Bitball project in the Guangdong Scientific Technical Exhibition Hall. It described, designed and achieved the important modules relate to the project, which contains the Image Real-Time Acquisition, Image Processing and Image Transmission. This paper focused on the study of the Localization algorithm for face detection and face feature points. Based on the Model of face color, we could divide the multicolor image into three parts by the color information, which are skin area, hair area and background area. Then smoothen the skin area of the image so as to succeed in the Localization of face and then eyes, mouth and nose. The article also analyzed the characteristics and disadvantages of the existing face detection algorithm. It amply studied and discussed on the Gaussian Model based on YCbCr space. Introduced the technique of the adaptive thresholds and summed up a few typical techniques of the thresholding segmentation. By comparison, the article ultimately chose the 2-D Maximum Entropy Method as the segmentation algorithm carried out finally. The article stated the overall frame of the whole exhibition project and the structure of each module in the software system. Designed and carried on the modules in the Bitball software system, which contains Image Acquisition Module, Face Detection Module, 2-D Maximum Entropy Thresholding Segmentation Module and Image Transmission Image Transmission and also showed the flow charts and results for the experiments. This software was made by the techniques and algorithms above in the Visual C++ Developing Environment, based on Microsoft Windows 2000 platform. The experimental results showed that this software could make the right detection, localization and processing for clear faces with different expressions and different figures. The last part of the paper summed up the study on this subject and analyzed which aspects of the subject need furthe