A method for stylized generation of facial contour features based on unsupervised learning

Abstract

The present invention provides a stylized generation method of human face contour features based on unsupervised learning, which includes obtaining a human face picture, and using the human face contour and key feature points to perform feature extraction in the human face picture, and combining color feature information to define After the key face is stylized, the sampled image of the feature area is generated by local area binarization processing; the stylized image generation model is constructed based on the CycleGAN generation confrontation network, and the sampled image is imported into the stylized image generation model , to obtain a cartoon stylized image of the feature region of the face contour. The implementation of the invention solves the problems existing in the prior art, and improves the image stylization result and robustness.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a stylized generation method of human face contour features based on unsupervised learning. Background technique [0002] In recent years, the stylization technology of key areas of the face based on feature extraction has received great market attention. A large number of application technologies based on feature extraction of face regions have emerged, such as face image editing and beautification, video face style mapping, "face changing", etc., which have been widely used in the fields of film and television entertainment, animation games, and advertising. [0003] Using traditional image processing technology, the style change and migration of key features of the face can be realized, so as to achieve the purpose of beautifying the face image. However, with the development of the field of computer vision, more intelligent and rapid facial feature stylization and fe...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the embodiments of the present invention is to provide a method for stylized generation of facial contour features based on unsupervised learning, which can solve the problem of rough image edge information, insufficient smooth transition, and insufficient fusion of contour images in local feature regions in the prior art. Naturally, there is a lack of effective filtering algorithms to deal with the edge fusion of images, which improves the image stylization results and robustness

Images