Method and system based on portrait segmentation precision improvement, and storage medium

Abstract

The invention provides a method and system based on portrait segmentation precision improvement, and a storage medium. The method comprises the steps of receiving a video stream about a video cloud conference and reading a frame of source image from the video stream; performing face detection on the source image through an RBF face detection network, and detecting to obtain the number of suspected faces in the source image; if the number of the detected suspected faces is not zero, judging whether each suspected face is a real face or not according to the ratio relation between the source image resolution and the coordinate area of each suspected face; according to a judgment result, if a real face exists, marking a coordinate area of the real face; carrying out portrait segmentation processing on the source image based on a real face coordinate area to obtain a corresponding binary grey-scale map; and fusing the binary grey-scale map, the source image and a preset background image to obtain a fused image, and displaying the fused image on an interface of a video cloud conference watching side. According to the invention, the portrait segmentation precision of the video cloud watching end can be improved.

Description

technical field [0001] The present invention relates to the technical field of video processing, in particular to a method for improving the accuracy of segmentation based on portraits. Background technique [0002] The virtual background in video cloud conferencing is mainly used when the user needs to replace the background information with an existing image or video due to personal privacy or sensitive location, referred to as virtualization. The function of the virtual background in the cloud conference is mainly to generate a binary gray-scale image Alpha based on Image Matting (portrait segmentation) for each frame in the video (generally, the portrait area is 255, the background area is 0, and the intersection edge of the portrait and the background is 0-255), fuse the current frame with the background image to be replaced, and finally get the desired result image or video. [0003] Image Matting technology refers to the use of traditional image processing or deep le...

AI Technical Summary

Problems solved by technology

[0007] Due to the limitations of the CNN network itself, it is impossible to achieve 100% accuracy. Therefore, the portrait segmentation network often recognizes some objects that look like people's clothes and outlines in unmanned videos as portraits, such as sofas, high-back chairs, and hangers. The coats on the screen, etc., will be recognized as portraits by the portrait segmentation network and displayed on the viewing end, resulting in a sharp drop in viewing experience

Images