A low-illumination surveillance image enhancement method based on reference frames

Abstract

The invention provides a reference frame-based low-light surveillance image enhancement method, which includes obtaining a reference frame and a low-light image to be enhanced from a surveillance video, and extracting neighbor frames of the low-light image; calculating the illumination component of the background image, and calculating the low-light image to be enhanced. Enhance the illumination component of the image, extract the reflection components of the low-illumination image and multiple adjacent frames, and then fuse them to obtain the enhancement result after denoising the low-illumination image to be enhanced. Therefore, the present invention can make the illumination component estimation of the image to be enhanced more accurate, the color fidelity of the enhancement result is higher, and the noise interference problem is smaller; thereby improving the clarity and recognition of low-illumination monitoring images, and playing a role in monitoring images. Its criminal investigation value provides strong support.

Description

technical field [0001] The invention belongs to the technical field of image enhancement, and relates to low-illuminance image enhancement related technologies, in particular to a low-illuminance monitoring image enhancement method based on a reference frame. Background technique [0002] As the pace of "Safe City" construction is advancing across the country, video surveillance technology is playing an increasingly critical role in maintaining social public safety. However, because surveillance cameras are often shot in low-quality lighting environments, such as night and daytime rainy weather, the captured surveillance videos are often of low quality and blurred. Images captured in low-quality lighting environments are low-light images. Therefore, in practical applications, there is an urgent need for quality improvement of low-light images in a large number of surveillance scenes to improve their clarity and recognition, and to give full play to the criminal investigatio...

AI Technical Summary

Problems solved by technology

Second, although the reference-based low-illuminance image enhancement method introduces other reference image information, the fusion-based enhancement algorithm adopts the form of dividing the low-illuminance image into two parts, the background and the foreground, which are often enhanced due to the foreground contour extraction. The inaccuracy leads to unnatural outline of the enhancement result; the enhancement algorithm based on daytime reference contains background image texture information in the estimated illuminance component during the enhancement process, which leads to the confusion of foreground and background texture in the enhancement result; the enhancement algorithm based on big data relies on a large number of images Sample data, it takes time to train the enhancement model, and the current enhancement model is good for the simulation of low brightness and noise enhancement, but the enhancement result for the actual low illumination image is still not ideal

Images