Research On Makeup Transfer Model Via Bat Algorithm

With the development of artificial intelligence,automatic face beautification technology has been widely used in the area of beauty camera,live broadcast and shopping software.Makeup transfer is an automatic face beautification technology based on the sample,and it aims to transfer the makeup style from a given example face image to another non-makeup one while preserving face identity.This technology can provide users with the preview makeup effect without physically applying the makeup.However,most of the existing research can only provide users with a single makeup transfer effect.Therefore,this paper studied makeup transfer model based on bat algorithm,and the main contributions are as follows.In the existing work of makeup transfer,the characteristics of example makeup and subject image are not considered,and only a single makeup effect can be transferred.This paper proposes a skin tone tuning model based on bat algorithm,which optimizes the weight value of color layer transfer in process of makeup transfer.Experimental results show that the model proposed is effective,and the beauty value is 10.77% higher than Guo’s method.At the same time,this model is adaptive and can give the corresponding optimal weight value for different example makeup or subject image.What’s more,the model proposed can be applied in beauty cameras and other applications to provide users with appropriate makeup and the lightness of makeup.The blemish or the sudden change of lightness in the lip region of example image can cause the unnatural lip makeup transfer effect.To address this problem,this paper proposes a lip color tuning model based on bat algorithm,which optimizes the weight values of the three channels after converting the lip region to CIELAB colorspace.The experimental results show that this model is effective,the beauty value is improved by 6.63% compared with Guo’s method,and the effect of lip makeup transfer is more natural than original model.What’s more,the model proposed can provide the most suitable lightness of lip color when user wears a certain lip makeup,and it also can give the guiding suggestions about dosage of lipstick for users.In order to solve the problem that the combination of skin tone and lip color will affect the harmony of facial makeup,this paper proposes a skin tone and lip color tuning model based on bat algorithm.The model optimizes both the weight value in the color layer transfer and the weight values of lip color.At the same time,three experiments were designed to prove the effectiveness and adaptability of the proposed model.Compared with Guo’s method and skin tone tuning model,the beauty value of proposed model has increased by 13.33% and 1.94%,respectively.This model can provide users with personalized service of matching skin tone and lip color.Aiming at dealing with the problem of inaccurate or even invalid positioning of facial feature points in makeup transfer caused by blurred or incomplete face,this paper proposes a makeup transfer model for burred face image and blemished face image.The super-resolution network SRWGAN-GP is used to reconstruct the burred face image,and the globally and locally consistent image completion network is used to complete the blemished face image.After that the face image processed goes into the makeup transfer model,and at last it results in the automatic beautification of burred face image and blemished face image.Experimental results show that this model is effective and can provide users with a better preview effect of makeup.What’s more,this model extends the application scene of the makeup transfer model from multiple perspectives,and provides a new idea for the face image beautification in old photos and films.