Method and device for updating background image in video coding

Abstract

The invention relates to video processing, and discloses a background image updating method and device in video coding. The background image updating method in the invention comprises the following steps of: if a background image needs to be updated, judging whether a current frame is a still frame or not; if so, updating a first background image into the current frame; and if not, processing the background image according to a candidate background image obtained through background modelling. Background details can be well reserved in the background image; a good reference image is provided for extracting a foreground image and a subsequent coding background area; and furthermore, the background image extraction efficiency can be effectively increased.

Description

technical field [0001] The invention relates to video processing, in particular to a method and device for updating background images in video coding. Background technique [0002] The general idea of ​​the structured coding technology based on foreground and background separation is that the image to be coded is divided into two parts, the foreground and the background, through the background reconstruction technology. form a stream. [0003] Background reconstruction technology is a commonly used method in the field of video analysis. There are many mainstream methods, such as Gaussian background modeling, codebook model-based background reconstruction, non-parametric learning methods, and median filtering. The goal of these methods is to extract a clean moving foreground, and most of the time, in order to remove the influence of disturbance, multiple background models will be maintained, and the computational complexity is high. In video coding applications, most of the...

AI Technical Summary

Problems solved by technology

The goal of these methods is to extract a clean moving foreground, and most of the time, in order to remove the influence of disturbance, multiple background models will be maintained, and the computational complexity is high.

In video coding applications, most of the processed image resolutions are large resolutions such as 19201088. Direct application of mainstream methods cannot meet the high performance requirements of real-time coding. In addition, the background image in structured coding is expected to be best preserved. Background details, overcome instantaneous light interference, and provide reference images for subsequent coding background area coding

[0004] The background image in semi-structured coding needs to be generated by the background modeling module. Since the coding resolution often reaches 1920*1088 and other large resolutions, and the coding requires real-time operation, this requires the background modeling module to run in real time, which requires a huge The amount of calculation seriously affects the coding efficiency

Images