基于SDM及加权均值滤波的人脸美化系统  

Face Beauty System Based on SDM and Weighted Mean Filtering

作　　者：秦枫 卢芳芳 林江南 QIN Feng;LU Fangfang;LIN Jiangnan(School of Computer Science and Technology,Shanghai University of Electric Power,Shanghai 200090,China)

机构地区：[1]上海电力大学计算机科学与技术学院,上海200090

出　　处：《上海电力大学学报》2020年第4期374-378,共5页Journal of Shanghai University of Electric Power

摘　　要：提出了一种基于监督下降法(SDM)及加权均值滤波的人脸美化系统。该系统首先采用SDM实现人脸关键点定位,接着采用交互式图像变形实现人脸局部变形,最后采用表面模糊算法实现肤质调整。实验结果表明,与其他人脸美化系统相比,该系统可通过调整修图的力度达到瘦脸、美目、去斑等人脸美化效果,同时美化后的图片保留原图的清晰度和细节特征信息。A new face beauty system based on SDM and weighted mean filtering is presented.Firstly,the system uses Supervised Descent Method to realize the key point positioning of the face.Secondly,it uses Interactive Image Warping to realize the local deformation of the face.Thirdly,it uses the surface blur algorithm to adjust the skin texture.Compared with other facial beauty systems,the system can adjust the intensity of retouching to achieve face beautification effects such as face thinning,eye beautification,and speckle reduction.At the same time,the picture after the beautification of this system can still retain the definition and detail feature information of the primitive image.

关 键 词：监督下降算法 自动人脸美化系统 人脸关键点定位 人脸局部变形 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TP317[自动化与计算机技术—计算机科学与技术]