Defogging method based on image gradient distribution prior

Abstract

The invention discloses a defogging method based on image gradient distribution prior. The defogging method comprises the steps that firstly, a gradient distribution prior model of an image is obtained from a large number of high-quality natural image data sets through training and learning; secondly, changing gradient distribution of the foggy image to enable the foggy image to perform infinite approximation learning to obtain a prior model; and finally, solving the reconstructed image by using a Poisson equation to obtain a defogged image. The defogged image processed by the method is in twoaspects of contrast ratio and average gradient; compared with an existing treatment method, the method is improved to a certain extent; compared with the prior art, the method has the advantages thatedge details and other information in the original image are highlighted, a good visual effect is achieved, the values of the MSSIM and the PSNR are greatly improved, haze is more effectively removedwhile good structural similarity of the processed image is kept, and the image is clearer.

Description

technical field [0001] The present invention relates to a method for defogging, in particular to a method for defogging based on prior image gradient distribution. Background technique [0002] In haze weather, due to the scattering effect of particles such as dust and fog in the air, the acquired image is degraded. The obtained images are often poor in contrast and clarity, which affects the visual effect of the image; the existing defogging methods include the wavelet transform defogging method, the method of enhancing the color contrast of the local area to realize the defogging processing method, and the foggy image processing method. Preprocessing and median filter dehazing method, dehazing method based on dark channel prior, etc. [0003] The defogging method based on wavelet transform, the visual effect of the processed image is better, the disadvantage is that the method collects multiple foggy images for image fusion, and the efficiency of image fusion is low; the ...

AI Technical Summary

Problems solved by technology

[0002] In foggy weather, due to the scattering of particles such as dust and fog in the air, the acquired image is degraded

The obtained images are often poor in contrast and clarity, which affects the visual effect of the image; the existing defogging processing methods include wavelet transform defogging method, enhancing the contrast of local area color to realize defogging processing method, and the foggy image processing method. Preprocessing and median filter dehazing method, dehazing method based on dark channel prior, etc.

[0003] The defogging method based on wavelet transform, the visual effect of the processed image is better, the disadvantage is that the method collects multiple foggy images for image fusion, and the efficiency of image fusion is low; the defogging process is realized by enhancing the contrast of the color of the local area, this method enhances the image contrast At the same time, the noise of the image is also enhanced; by preprocessing the foggy image and median filtering to achieve a certain defogging effect, it meets the requirements of real-time performance. A lot of detailed information may also be lost; the dehazing method based on the dark channel prior uses prior knowledge to estimate the air transmittance, and then optimizes the estimated air transmittance through the soft matting principle to achieve the purpose of image defogging. However, since the soft matting process requires a large amount of calculations, it is difficult for this method to meet the real-time requirements.

Images