Defogging processing method for single color image based on DCP and NL_MTV model

Abstract

The invention belongs to the technical field of image processing, and relates to a defogging processing method for a single color image based on DCP (Dark Channel Prior) and an NL_MTV (Non-Loal_Multi-channel Total Variation) model. The defogging processing method comprises the steps of: acquiring an accurate atmosphere light value and a transmissivity through a layer-by-layer searching method anda quick guided filtering method; building an energy functional model of an atmosphere illumination value A and an atmosphere transmission function t; and recovering defogging images in different scenes by utilizing a rule item and a data item in the NL_MTV model. Through adoption of the defogging processing method of the invention, while a clear image is obtained, a noise problem existing after the defogging processing can be effectively alleviated, and at the same time, original textural features of the scene also can be maintained.

Description

Technical field: [0001] The invention belongs to the technical field of image processing, and relates to a single color image defogging processing method based on a dark channel prior and a non-local MTV model. Background technique: [0002] Fog is a common natural weather phenomenon in nature. It is caused by the scattering of suspended particles when atmospheric light travels through the air. Under such conditions, the quality of images acquired by outdoor imaging equipment deteriorates seriously, and details cannot be preserved. The color saturation is reduced, which greatly limits the work efficiency of outdoor imaging equipment, such as video surveillance, navigation, transportation, satellite remote sensing, etc. Therefore, it is of great research value to improve the working efficiency of outdoor imaging equipment to study how to obtain a clear fog-free image in a foggy environment. [0003] At present, according to the differences in the principle of image dehazing,...

AI Technical Summary

Problems solved by technology

[0003] At present, according to the differences in the principle of image dehazing, the existing methods can be divided into three categories: the first category, based on foggy image enhancement methods, this type of method does not need to understand the cause of image degradation, and mainly uses targeted image processing methods Improve its contrast and detail features, improve the visual effect of the image, the representative algorithm is: histogram equalization algorithm, its principle is to optimize the gray value distribution range of the image through histogram equalization, so that the gray value is evenly distributed in the In a higher dynamic range, the algorithm can improve the contrast and image details of the image, but sometimes it will cause noise phenomenon: Multi-scale Retinex algorithm (Multi-Scale Retinex, MSR) and its improved algorithm are based on color constancy Theoretical processing of foggy images, the algorithm can not only improve the contrast and brightness of the image, but also can preferentially adjust the gray dynamic range of the color image to remove the fog; Too bright; the homomorphic filtering algorithm is an image enhancement method that combines frequency-domain filtering and grayscale transformation. Although this method can improve image quality to a certain extent, it requires two Fourier transforms and one exponential operation for processing. Logarithmic operation, the amount of calculation is relatively large; the basic principle of the wavelet transform algorithm is similar to that of homomorphic filtering, using wavelet transform to obtain images with different frequency characteristics, and enhancing the image details of non-low frequency sub-blocks to achieve image defogging, but for bright or The processing effect of too dark foggy image is not obvious; the second type is based on the image restoration method

This type of image defogging method needs to understand the cause of image quality attenuation, including establishing a mathematical model of related fog image attenuation, using inverse transformation to restore the scene, and repairing image distortion caused by haze and other weather as much as possible. This method is aimed at different conditions. The effect of the foggy image is obvious and natural, and the loss of detail features is less. The key to this method is the accurate calculation of each unknown quantity in the relevant model. The representative algorithms are: Research, extending degraded image restoration to color images, but it is easy to oversaturate the color of the image after defogging and cause distortion; the Fattal algorithm assumes that the transmittance of light and the chromaticity of the object surface are not cross-correlated in the local area, and estimate the color of the scene color information, but when processing images with dark colors, it is easy to cause inaccurate color estimation; the Tarel algorithm performs white balance processing on foggy images to obtain atmospheric light values, and uses the derivative algorithm of median filtering to estimate the atmospheric dissipation equation. The speed is fast, but the effect is not ideal; the He algorithm obtains a dark channel prior theory that can estimate the haze concentration through the statistics of a large number of outdoor images in sunny weather, and combines it with the atmospheric scattering model, effectively However, this algorithm is easy to cause halo or white light effect. Although the soft matting method used in the later stage achieves the purpose of optimizing the transmittance, it has a large amount of calculation and low real-time performance; the third category , based on the image fusion method, image fusion is a method of combining images containing relevant information of multiple source channels into a high-quality image, representative methods include: multispectral image fusion method, this method does not need to detect fog and atmosphere, does not need Depth map, but possible haloing

Images