A method for editing video face fat and thin

Abstract

The invention discloses a video face fat and thin editing method, comprising the following steps: reconstructing a three-dimensional human face model based on a human face video, and outputting three-dimensional human face shape parameters, facial expression parameters and facial posture parameters of each video frame ; Adjust the 3D face model based on the 3D face fat and thin adjustment algorithm, and transfer the adjustment results to each frame to generate a deformed 3D face model for each frame; use the directed distance field to establish the before and after deformation on the 2D plane Dense mapping of the face boundary, and adjust the dense mapping based on the structure of the 3D face model; deform the face video frame based on the dense mapping, and use energy optimization to reduce the background distortion of the video frame caused by the deformation, get The deformed face video frame, and its corresponding replacement back to the original face video. The present invention realizes the automatic generation of human face video conforming to the fat and thin scale, and can still obtain ideal results under the conditions of being blocked, having long hair, wearing glasses, and the like.

Description

Technical field [0001] The present invention relates to the field of portrait editing technology, in particular to a video face fat thin editing method. Background [0002] With the rapid development of social networks and media, more and more people are actively sharing personal videos and pictures on the internet. Image editing techniques are often used to create special face effects, such as exaggeration and beautification of faces. Current research focuses on editing the color, texture, and shape of a human face. [0003] "Deep Shapely Portraits" (In MM′20:The 28th ACM International Conference on Multimedia.2020.1800–1808) exposes an automatic image-based method of fat skinn editing that uses neural networks to automatically identify the most appropriate fat skinny scale for a particular person and deforms the image using rendering fusion techniques. But the technique is unstable in the time domain and doesn't handle the side face very well. [0004] "Motion-AwareGradient Dom...

AI Technical Summary

Problems solved by technology

[0014] However, none of the above video-based reconstruction methods can obtain unique face shape parameters, that is, the face shape parameters obtained in the first frame and the face shape parameters obtained in the last frame are not the same

Images