5 / 2016-06-19 16:30:30

Improve the imperfect of facial shape

10523,10522,8554,10521

Abstract—Intuitively, facial shape refers to the contour of face and relates to curvature. In this paper, we propose seven different face template as the target model, which can be improved the imperfect of facial shape and transformed freely shape what you want desired. Firstly, we exploit deep learning technology for extracting feature points that avoids extraordinary sophistication of manual operation. Secondly, we adapt Procrustes analysis to align the original image and target template (translation, scale, rotation) to realize coordinate normalization. Subsequently, we utilize Laplace mesh editing techniques to achieve deformation while preserve the local details. Largely speaking, an important practical aspect of this paper is that we compensate the shortcomings of beautification software that cannot preferably achieve deformation from ordinary facial shape to desired results. Simultaneously, the main contribution of our proposed method can remedy some facial shape deficiencies, such as high convex, leniency, depressions and narrow of face, making unsightly parts and facial proportions more coordinated and symmetrical while largely maintaining the individual characteristics. Experimental results manifest that our proposed method is exceptionally beautiful.