Automatic Face Beautification Based On Coordinate Regulation Of Three-Channel And Details Smoothing

The areas of intelligent mobile equipment, the cloud computing and the mobileinternet have undergone a significant evolution within the past few years. Acomponent commonly used in these areas is multimedia systems, which have becomean integral part of our daily lives now. Examples include digital cameras, videoconferencing, and photo self-timer and so on. In all such systems, human face is afrequently occuring entity. Naturally, beauty has fascinated human beings from thevery dawn of mankind, especially the beauty of face. Factors that degenerate facialbeauty comprise undesirable skin components such as acne, spots and so forth.Currently, there are some tools for making alterations to facial profiles, such asPhotoshop, Meitu xiuxiu, etc. However, the use of these tools is cumbersome for mostpeople, as it requires some technical knowledge about the operation of the tools. So anauto-beautify-face system is required.In our research, we aim to remedy this problem by designing and implementing auser-friendly, totally automatic facial beauty enhancement tool. The auto-beautify-system main consist of three parts, which are face detection part, skin colorsegmentation part, and skin beautification part. Firstly, we use the algorithm based onHarr-like and AdaBoost put forward by Viola and Jones to locate the face region inthe input im age. Then the detected face is to be segmented to obtain a skin map.Finally, the segmented skin color is beautified in two steps which are whiting andsmoothing.The contributions of the thesis are as follows:1. We propose a new way toprocessing skin segmentation using the combination of fixed threshold and adaptivethreshold.2. To avoid shaping contrast after beautification, we propose to create aspecial mask which modified with Gaussian kernel.3. We formulate a non-lineartransformation of light to prevent highlighted area, and hue to keep the beautified skinnatural after skin whiting.