A Single Image Restoration Method Based on Gradient Field Region Segmentation

Abstract

The invention discloses a gradient field region segmentation-based single image restoration method. The method comprises the following steps of: respectively converting a scaled image into an HIS color space and a gray space, carrying out automatic threshold value segmentation on a brightness space, carrying out gradient field segmentation on the gray space to obtain a bright area and a non-brightarea of the image, and obtaining a value representing global atmospheric background light according to the bright area; obtaining a dark channel image of the scaled image, carrying out smoothing by using a constructed NSRAD model so as to obtain an optimized transmissivity image, importing a multi-area theory to fusing the bright area and the non-bright area, and further correcting the transmissivity image; and obtaining a defogged image according to the obtained value representing the global atmospheric background light and the corrected transmissivity image on the basis of a dark primary color theoretical model. The method is strong in road environment adaptability and food in algorithm stability, and is capable of eliminating halo phenomenon and effectively restoring images in foggy days so as to enable monitors to obtain more pieces of 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