Replacement of Facial Parts in Images

Abstract

It is interesting to edit facial appearance in images to create a desirable facial shape of persons. In this paper, we propose a novel method to modify facial appearance by replacing facial parts between arbitrarily paired images. To this end, our method consists of face segmentation, face reconstruction, mesh deformation and image editing. Given one source and one target image, the target image is first segmented into the front facial region and background image. Secondly, 3D facial models and relevant scene parameters are estimated from both images. Thirdly, the target facial part is replaced with the selected source part on the 3D mesh. Then, the new replaced 3D face is rendered into a facial image. Finally, the new facial image is generated by seamlessly blending the rendered image and background image. The main advantage of this method is that we transfer facial geometric information between images using 3D model, which can deal with arbitrarily paired images with the different facial viewpoint. We present several experimental results to show the effectiveness of our method and comparison with those existing methods to demonstrate that our method is more advantageous and flexible in terms of practical applications.