A Single Image Restoration Method Based on Gradient Field Region Segmentation

Abstract

The present invention is a single image restoration method based on gradient field area segmentation. The scaled image is converted into HSI color space and grayscale space respectively, automatic threshold segmentation is performed on the brightness space, and gradient field segmentation is performed on the grayscale space at the same time. , obtain the bright and non-bright areas of the image, and obtain the value representing the global atmospheric background light based on the bright area; obtain the dark channel image from the scaled image, and use the constructed NSRAD model for smoothing to obtain the optimized transmittance Image, introduce multi-region theory, fuse bright areas and non-bright areas, and further correct the transmittance image; on the basis of the dark primary color theory model, based on the value representing the global atmospheric background light obtained above and the corrected transmittance image to obtain a dehazed image. The invention has strong adaptability to the road environment, good algorithm stability, can effectively eliminate the halo phenomenon, and can effectively restore images in foggy days, allowing monitors to obtain more effective information.

Description

technical field [0001] The invention belongs to the field of image processing, in particular to a method for restoring a single image based on a gradient field area field. Background technique [0002] At present, road traffic monitoring is mainly composed of acquisition equipment, transmission equipment and processing equipment and other units. The acquisition device is mainly a camera, the purpose of which is to convert road and pedestrian information into image information through sensors. Current monitoring equipment is usually suitable for better unobstructed natural environments. When encountering bad weather, such as fog, the images collected by the camera will be blocked, and road monitoring information will be lost. Therefore, it is of great significance to better monitor road information in various natural environments. [0003] At present, there are generally two types of image restoration methods in foggy weather: methods based on dark primary colors, and metho...

AI Technical Summary

Problems solved by technology

Based on the enhanced method, Retinex theory is generally used for processing. The method uses multi-scale Gaussian convolution filter processing. Although it has a good effect on dense fog, it will appear excessive for thin fog. Enhancement phenomenon

[0004] The uniformity of background light in the environment affects the quality of video capture images. The traditional dark channel theory does not take into account the brighter areas in the image, that is, the areas where the background light in the environment affects the image when processing the collected foggy images. , when estimating the transmittance of the image, global processing is used, resulting in a small estimated transmittance. After the restoration of the dark channel model, the overall brightness of the image is dark, and a certain degree of halo appears at the edge of the brighter area. Phenomenon

Images