Video noise estimation method based on human eye visual characteristics

Abstract

The invention relates to a video noise estimation method based on human eye visual characteristics. In the existing method, a large amount of false detection and missing detection is caused, particularly in coded and decoded noise videos. According to the method, firstly, a video bank is prepared, in addition, the manual labeling is carried out, MOS is obtained, the over-dark and over-bright regions of videos are determined through the analysis on the video brightness distribution, the video interframe object movement and the random variation, the interframe non-movement change quantity is extracted as the reference quantity of the noise evaluation, an appreciable model is introduced in the noise evaluation state for measuring the visual visible degree of the noise, the interframe structure change degree is used for measuring the visual effect of the noise, and meanwhile, the relationship model among the MOS, u and v is estimated. The video noise estimation method has the advantages that the problems of the exiting method can be effectively solved, a better effect is realized on the non-Gaussian model noise, even the noise in non-independent identical distribution, and meanwhile, high correlation on estimation results and human eye visual effects is realized.

Description

technical field [0001] The invention belongs to the field of video analysis and the field of video monitoring technology, and relates to a video noise estimation method based on human visual characteristics. Background technique [0002] With the development and popularization of computer technology, network communication technology, video surveillance technology and consumer electronics, video has been used more and more widely. However, in the process of video acquisition, encoding, and transmission, it may be affected by various factors such as equipment aging, low illumination, unclean power supply, encoding quantization, electromagnetic interference, etc., and it is subject to different degrees and forms of noise pollution, which makes the video quality The decline not only affects the visual effect of the video, but also affects the subsequent more advanced video processing to a certain extent. It can be seen that a reliable noise estimation method plays a very meanin...

AI Technical Summary

Problems solved by technology

[0009] Aiming at the deficiencies existing in the current video noise detection, the present invention proposes a noise estimation method based on human visual characteristics. The method mainly includes: preparing a video library and manually labeling to obtain MOS, by analyzing the brightness distribution of the video, Object movement and random changes between video frames Determine the dark and bright areas of the video, and extract the non-motion changes between frames as a reference for noise evaluation , introduce just perceptible model (JND) in the noise evaluation stage to measure the visual visibility of noise , the visual effect of noise is measured by the degree of structural change between frames , while estimating the MOS and relationship model

Images