Refined facial expression capturing method and device based on RGB image

Abstract

The invention relates to the technical field of 3D expression reconstruction, and particularly discloses a refined facial expression capture method and device based on an RGB image. The method comprises the steps of obtaining a natural expression image and a changed expression image of a target face, and inputting the changed expression image into a face expression capture model to obtain standardexpression features; inputting the natural expression image and the changed expression image into a face key point regression model to obtain key point space coordinates of two groups of target faces, performing coordinate standardization processing, calculating key point space coordinate change data, and substituting the key point space coordinate change data into a key point expression change correlation function to obtain expression change features; and obtaining the expression capture data according to the sensitive parameters. Through the method, the possibility of distortion of expression capture in the face space migration change process is reduced, the effect of local change of expression features is enhanced, and the stability and accuracy of expression capture are improved.

Description

Technical field[0001]The invention relates to the technical field of 3D expression reconstruction, in particular to a method and device for capturing refined facial expressions based on RGB images.Background technique[0002]At present, artificial intelligence technology and AR intelligent technology are developing rapidly. Internet users have more and more demand for content ingestion. The demand for animated characters, AR and other content is becoming stronger. The construction of characteristic actions and expressions based on different characters has become these areas. The fulcrum content inside. Traditional animation and 3D character expression production relies entirely on manual labor. The expression change of the same character requires multiple pictures and changes in multiple models to express. Not only does it require a lot of work, but it is also difficult to guarantee the output of animation or The expression model is consistent with the expected performance, so that th...

AI Technical Summary

Problems solved by technology

The production of traditional animation and 3D character expressions is completely dependent on manual work. The expression changes of the same character need to be represented by multiple pictures and changes in multiple models at the same time, which not only requires a huge workload, but also makes it difficult to guarantee the output of animation The expression model is consistent with the expected performance, so that the construction of character expressions is repeated, which greatly reduces the efficiency of content output in this field

The current existing method of abstracting the changes of facial expression features through character key point positioning also has the problem of key point jitter causing expression jitter; facial space changes, migration, and multiple technical problems such as distortion of the detection effect.

Images