Head segmentation method based on position controllability

Abstract

The invention discloses a head segmentation method based on position controllability. The method adopts a human head key point detection module, a position correction module and a human head segmentation module, a user click position is corrected through the position correction module, so that a human head key point position is matched, and a final human head segmentation result is obtained by utilizing key point information and the human head segmentation module. The method has the beneficial effects that the single head in a multi-person scene can be accurately segmented, the method is more flexible, the operation efficiency is improved, and the method is deployed at a mobile phone end.

Description

technical field [0001] The present invention relates to the technical field related to image processing, in particular to a position-controllable head segmentation method. Background technique [0002] Head segmentation is a relatively common function in short video software at present. Its purpose is to provide a basis for users to create special effects such as emoticons, head stickers, and animation bodies. [0003] At present, most of the head segmentation is developed based on deep learning technology, and the semantic segmentation model is usually used to deduct the head part of the image. However, when there are many people in the picture, it is impossible to deduct a single head in a targeted manner, resulting in a greatly reduced playability. Although there are instance segmentation technologies that can differentiate the heads of people in multi-person scenarios, the current instance segmentation technology has little consideration for operation and requires compl...

AI Technical Summary

Problems solved by technology

However, when there are multiple people in the picture, it is impossible to carry out targeted deduction for a single head, resulting in a greatly reduced playability

Although there is an instance segmentation technology that can differentiate the head segmentation in a multi-person scene, the current instance segmentation technology is not considered high in operation and requires complex post-processing, which is not suitable for device-side deployment.

Images