基于几何特征的人脸美学分析计算机科学与技术专业论文.docxVIP

哈尔滨工业大学工学硕士学位论文 哈尔滨工业大学工学硕士学位论文 － －II－ Abstract Human have analyzed self beauty for thousands of years. There ceaseless aspiring after beauty advances the aesthetic consciousness increasingly and meanwhile boosts the social developing. And facial aesthetics is its essential focus. For the research of facial beauty, there are a lot of quantifiable standards, so that we have reasons to believe that there are some quantitative relations between facial beauty and face geometric features. In this paper, we try to use the research of the face geometric features to analyze the statistical properties of facial aesthetics，in order to derive quantitative evaluation of facial aesthetic criteria systematic. First, feature points were comprehensive defined to depict facial features, and the extracted geometric features were normalized. A normalization approach was proposed to filter out the non-shape variations – translation, rotation, and scale. Second, a image warpping method, which is scattered data interpolation with multilevel B-Splines, was studied. The facial geometric feature deformation can be achieved under the condition of original feature points and target feature points, in order to achieve a target face which is more attractive. The results can be used to guide cosmetology, plastics surgery and so on. Third, the previous theory of aesthetic, such as “san ting wu yan” of ancient China, “Golden Ratio” of ancient Greece, and symmetrical character of facial, were combined. 20 facial features of the geometric ratio was obtained, with the results of analyzing the large amount number of sample data, and the law of these proportions satisfied was found, ultimately a systematic quantified facial aesthetic evaluation criterion was established. Finally, experiments were designed to verify the proposed evaluation criteria of facial aesthetics. One of the experiments is about image classification. The results prove that the proposed ratios of 20 geometrical features make a good