Posture migration method based on three-dimensional human body parameterized model

Abstract

The invention discloses a posture migration method based on a three-dimensional human body parameterized model. The method comprises the following steps: 1) obtaining a target person three-dimensional human body parameterized model consistent with a source person in posture and expression; 2) binding different parts of the three-dimensional human body parameterized model of the target person with different colors to be rendered, and obtaining a posture image; 3) performing image operation on the posture image to obtain a head and body posture image; 4) inputting the head and body posture images into a human body part generation module to obtain head and body posture migration images; 5) performing image operation on the head and body posture migration image to obtain a head and body posture migration intermediate image; 6) inputting the head and body posture migration intermediate image into a fusion module to obtain a result image after the posture migration of the target person. According to the invention, while the result quality is ensured, the demand of training data is reduced, the collection difficulty of the training data is reduced, and the training time is shortened.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a posture transfer method based on a three-dimensional human body parametric model. Background technique [0002] The main task of human pose transfer is to transform the pose of the target person to make it perform the same action as the source person, and finally generate a pose picture of the target person that is consistent with the pose of the source person. This topic has a good application prospect, such as entertainment industry application scenarios such as generating picture frames of specific actions for actors in movies. In addition, as the research on human behavior understanding using deep learning becomes mainstream, the demand for human data has surged, and human body pose transfer can play a role in data expansion, providing a large amount of data for further human behavior research. [0003] At present, the task of human pose transfer is mainly...

AI Technical Summary

Problems solved by technology

But the disadvantage is that for a new target person, the generative confrontation network needs to be retrained, and a large number of images of the target person are required

Images