A processing method for improving video image quality in rainy weather

Abstract

The invention relates to a processing method for improving the color video image quality in rainy days. The processing method comprises a video image conversion step, a video raindrop detection step and a video raindrop removal step, wherein in the video image conversion step, a RGB (Red, Green and Blue) image is converted to a vision cone intensity signal (L, M, S representing long wave, medium wave and shortwave), and then is converted to one no tone function and two opposite tone functions; in the video raindrop detection step, raindrops are detected according to the color difference relation after a pixel (x, y) is shielded by the raindrops, and the raindrops are detected according to the no tone function a (x, y) at the pixel (x, y); and in the video raindrop removal step, a binary image B (x, y) of the raindrops is defined, B (x, y) is processed through Gaussian blur to obtain G (x, y), a weight matrix w (x, y) is built up by G (x, y) and B (x, y), and w (x, y), a<new>(x, y), rg<new>(x, y) and rb<new>(x, y) are processed through inverse transformation to reach a RGB domain so as to obtain the image, in which the raindrops are removed. The processing method has the advantage of improving the raindrop-removing image quality.

Description

technical field [0001] The invention relates to the technical field of machine vision and video image processing, in particular to a processing method for improving the quality of video images in rainy days. Background technique [0002] Video surveillance systems have been widely used in various fields such as transportation, finance, environmental protection, and community life. However, weather conditions such as rainfall often lead to a decrease in the quality of surveillance images, which prevents surveillance personnel from discovering hidden alarm situations, and is not conducive to the implementation of intelligent analysis algorithms such as target recognition. Seriously affect the monitoring system performance. [0003] In the prior art, it is judged whether it is a raindrop by fitting the linear ratio of the brightness difference of the pixels in the same raindrop before and after being affected by the raindrop to the background brightness, and then the rain trac...

AI Technical Summary

Problems solved by technology

[0004] Therefore, the above-mentioned existing video raindrop detection method uses a real number algorithm to process image brightness or color information, which does not conform to human visual perception; Smoothing, if the method of smoothing and filtering the image after rain removal is adopted, the image may be blurred and information will be lost

Images