A method and equipment for generating a virtualized body

Abstract

The invention discloses a method and equipment for generating a virtualized body. The method comprises the following steps: detecting whether a target object of a to-be-virtualized body in a picture wears a preset shielding object or not; When the target object wears at least one occlusion object, removing the detected occlusion object from the image of the target object by using a pre-trained neural network model, and generating a corresponding virtual body according to the obtained image without the occlusion object; Selecting a shielding object image matched with the feature from a preset shielding object 3D image library according to the preset external feature of each removed shielding object, and loading the shielding object image to the corresponding position of the virtualized bodyto obtain the virtualized body of the target object; And when the target object does not wear the shielding object, directly generating a corresponding virtualized body according to the image of thetarget object. By adopting the method and the device, the similarity between the virtualized body and the real image can be effectively improved.

Description

technical field [0001] The invention relates to image processing technology, in particular to a method and device for generating virtual avatars. Background technique [0002] With the increasing popularity of models such as avatars (virtual characters, Avatar) in terminal devices such as mobile phones, the method of model generation based on human facial expressions and actions has occupied the mainstream. The generation of virtual avatars is mostly carried out by taking selfies or choosing to take selfies. [0003] Existing avatar generation schemes are usually based on the user's selfie and directly perform model matching, that is, for facial features, load corresponding facial feature textures from the model and combine them to generate avatars. [0004] The inventor finds in the process of realizing the present invention: above-mentioned scheme can produce the mistake of three-dimensional virtual avatar portrait in many cases. Especially after the user takes a photo w...

AI Technical Summary

Problems solved by technology

[0004] The inventor finds in the process of realizing the present invention: the above-mentioned scheme can produce the error of the head portrait of the three-dimensional virtual avatar in many cases

Especially after the user takes a photo while wearing glasses, earrings and other accessories, the virtual avatar generated based on the corresponding photo has many errors, and the similarity with the user is low, which makes the virtual avatar accurately reflect the user's appearance characteristics and affects the virtual avatar. legibility

Images