Image segmentation and image segmentation network training method, device and computer equipment

Abstract

The present application relates to an image segmentation method, an image segmentation network training method, an image segmentation device, an image segmentation network training device, computer equipment and a readable storage medium. The method includes: acquiring an image to be segmented; inputting the image to be segmented into a preset image segmentation network for image segmentation processing to obtain a target segmentation result of the image to be segmented; wherein the image segmentation network is based on the subject loss and edge loss of the training sample image Obtained by training; the main loss is the loss between the segmentation and prediction results of the training sample image for the main region of the target object and the segmentation prediction result, and the edge loss is the segmentation and prediction results of the training sample image for the edge region of the target object. According to the target segmentation result, the image for the main area of ​​the target object is segmented from the image to be segmented. Using this method can improve the precision of edge segmentation.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular to an image segmentation method, an image segmentation network training method, an image segmentation device, an image segmentation network training device, computer equipment and a readable storage medium. Background technique [0002] In many image or video application scenarios such as live broadcast, video, and blurring, it is often necessary to cut out the portrait part of the image to extract the image of the portrait part; the portrait part can be a head portrait, a half-length portrait, a full-body portrait Portraits, etc., to support users' personalized operations. [0003] However, due to the endless human body postures, the unpredictable spatial relationship between people, the erratic spatial position between people and objects, the background noise is complex, the shooting angle is variable, the light intensity is variable, and the infrared intensity ...

AI Technical Summary

Problems solved by technology

[0003] However, due to the endless human body postures, the unpredictable spatial relationship between people, the erratic spatial position between people and objects, the background noise is complex, the shooting angle is variable, the light intensity is variable, and the infrared intensity is different. Factors, the traditional portrait matting method performs very badly on the edge of the portrait, and there are problems such as blurred edges and errors in the matting accuracy of the portrait.

Images