Image processing, model training and live broadcast room background switching method

Abstract

The embodiment of the invention provides an image processing method, a model training method and a live broadcast room background switching method. In the image processing method, input data of the matting model in a feature extraction stage are a target image and a background image thereof, and features on the target image and the background image can be mined. In the decoding stage, the cutout model can decode the feature map output in the feature extraction stage under the assistance of the background image and / or the target image, so that the decoding process can be further focused on the difference between the target image and the background image. In the output stage, the matting model uses the depth guiding filtering network to carry out depth guiding filtering on the decoding result image, and the decoding result can be corrected so that the matting result can be more accurate. Based on the implementation mode of guiding separation of the foreground and the background according to the known background information, a relatively accurate matting result can be obtained under the assistance of the target image and / or the background image under the condition of not carrying out mass calculation, and the matting efficiency is greatly improved.

Description

technical field [0001] The present application relates to the technical field of computer vision, and in particular, to a method for image processing, model training, and background switching between live broadcasts. Background technique [0002] In many scenarios, image matting is required to segment the foreground from the background in the image. In some scenarios, the image needs to be cut out in real time to meet the requirements of background or scene replacement. For example, in a live broadcast scene, foreground objects such as people and objects need to be segmented from the real-time picture of the live video, and the live broadcast environment needs to be replaced in real time. [0003] However, the existing matting algorithms have a large amount of computation and low matting efficiency, and cannot meet the requirements of high real-time matting. Therefore, a new solution needs to be proposed. SUMMARY OF THE INVENTION [0004] Various aspects of the present ...

AI Technical Summary

Problems solved by technology

[0003] However, the existing matting algorithms have a large amount of calculation and low matting efficiency, which cannot meet the needs of real-time matting.

Images