Data-driven Enhancement Of Facial Attractiveness

Since ancient times,“beauty” has been a fascinating, but complicated and confusingconcept. In most people’s opinion, facial beauty is an abstract concept that is inherentlydifficult to quantify and evaluate. The research in this area has never been stopped, such asLeonardo da Vinci’s “Code of Beauty”. Although this description of the beauty rules is quitesimple, but it still partly reflects the consistency of the human perception for facial beauty.Benefiting from the development of computer technology and digital image processingtechnology, we have found that machines could learn to evaluate facial attractiveness by usingmachine learning methods.On this basis, our project is committed to find out the potential law of the humanperception for facial attractiveness, in order to make the facial beautification using machinelearning methods and image processing techniques.The major works of this paper are listed as below:First, we introduced the concepts and the basic framework of the facial attractivenessevaluating algorithm, and proposed a novel texture-based facial attractiveness evaluatingalgorithm by introducing Gabor feature for describing facial texture characteristics. Thereafter,we combined Gabor features with geometric facial features, and a high correlation of0.93with average human ratings was achieved.Second, we introduced the facial image morphing algorithm and the facial texture blendingalgorithm. Our system uses the Multilevel Free-Form Deformation (MFFD) for morphingfacial images. In the facial texture blending module, we introduced the WLS-basededge-preserving smoothing algorithm and its improved version, and introduced the facialimage decomposition algorithm based on it; meanwhile, we proposed a simplified algorithmto reduce its calculation complexity.Finally, we proposed a data-driven facial beautification system which has the facial texturebeatification feature. This system uses the KNN-based beautification engine to guide thefacial image morphing module and the facial texture blending module to morph the input faceand modify its texture, in order to get a beautified version of the input face. There are so many applications of the face beatification system. For instance, aids offace-lifting in hospitals, providing the best scheme for facial surgery; Face beatification canalso be used for vitual character designing in cartoons or games. For digital photo processing,our facial beatification system can also provide satisfactory beatification results for peoplewho are not good at image processing.