CN-121998865-A - Image defogging method, system and storage medium

CN121998865ACN 121998865 ACN121998865 ACN 121998865ACN-121998865-A

Abstract

The invention discloses an image defogging method, an image defogging system and a storage medium, which relate to the technical field of image defogging, wherein the image defogging method comprises the steps of acquiring linear polarization images and circular polarization images in the same scene at the same moment; the method comprises the steps of preprocessing linear polarization images and circular polarization images under the same scene at the same moment to enable the sizes and target positions of each pair of linear polarization images and circular polarization images to be the same, carrying out polarization difference processing on the preprocessed linear polarization images to obtain linear polarization difference images, constructing initial transmittance, carrying out polarization difference processing on the preprocessed circular polarization images to obtain circular polarization difference images, carrying out stability constraint correction on the initial transmittance to obtain final transmittance, and recovering to obtain defogging images based on the final transmittance. According to the invention, the stability constraint is carried out on the linear polarization defogging result by introducing the circular polarization information, so that the defogging effect and the visual quality of the image in the indoor strong smoke scattering environment are effectively improved under the condition of not depending on a deep learning model.

Inventors

ZHANG JIN
ZHANG HAIHANG
HONG LU
HUANG LIANG
XIA HAOJIE

Assignees

合肥工业大学

Dates

Publication Date: 20260508
Application Date: 20260126

Claims (10)

1. An image defogging method is characterized by comprising the following steps: S1, acquiring a linear polarization image and a circular polarization image in the same scene at the same moment; S2, preprocessing linear polarization images and circular polarization images in the same scene at the same moment to enable the sizes and target positions of each pair of linear polarization images and circular polarization images to be the same; s3, carrying out polarization difference processing on the preprocessed linear polarization image to obtain a linear polarization difference image, and constructing initial transmissivity of linear polarization leading; S4, carrying out polarization difference processing on the preprocessed circular polarization image to obtain a circular polarization difference image, and carrying out stability constraint correction on initial transmittance of linear polarization leading by utilizing the circular polarization difference image to obtain final transmittance; S5, recovering to obtain a defogging image based on the final transmissivity.
2. The image defogging method according to claim 1, wherein in S1, a pair of orthogonal linear polarization images I 1 、I 2 and a pair of orthogonal circular polarization images I 3 、I 4 under the same scene at the same time are acquired by using an amplitude-division imaging system.
3. The image defogging method according to claim 1, wherein the linear polarization image and the circular polarization image are preprocessed in S2 by inversion, sift registration and clipping.
4. The image defogging method according to claim 1, wherein the construction method of the initial transmittance in S3 is as follows: s31, constructing a scattering model; S32, carrying out polarization difference treatment on the preprocessed circularly polarized image to obtain a linear polarization difference diagram; S33, defining linear polarization scattering asymmetry and calibration coefficient by using a linear polarization differential diagram; s34, obtaining initial transmissivity of linear polarization dominant according to the linear polarization scattering asymmetry and the calibration coefficient.
5. The image defogging method of claim 4, wherein said scattering model is as follows: ; Where x denotes the pixel position, J (x) denotes the haze-free target, T (x) denotes the transmittance, A p (x) denotes the scattered light in the polarization state p, and I p (x) denotes the observed image in the polarization state p.
6. The image defogging method according to claim 4, wherein the final transmittance in S4 is calculated as follows: s41, carrying out polarization difference processing according to the scattering model and the preprocessed circular polarization image to obtain a circular polarization difference diagram; S42, defining a circular polarization scattering retention factor by using a circular polarization differential diagram and constructing a stability suppression function; S43, restraining the transmissivity estimation of the high scattering area according to the initial transmissivity and stability restraining function of the linear polarization leading, and obtaining the final transmissivity.
7. The image defogging method according to claim 1, wherein the stability suppressing function in S42 is as follows: ; wherein, beta is expressed as a parameter for adjusting the stability constraint intensity; and/or, the final transmittance is: ; in the formula, Indicating the initial transmission dominated by linear polarization.
8. The image defogging method according to claim 1, wherein the defogging method based on the final transmittance recovery in S5 is as follows, wherein an average value is calculated for the preprocessed linear polarization image and the circular polarization image as a foggy polarization image, a scattered light parameter is estimated according to a scattering region of the foggy polarization image, and the atmospheric scattering imaging model is improved according to the final transmittance to obtain the defogging image: ; Wherein I ref (x) represents a foggy polarized image, A (x) represents a scattered light parameter; represents the final transmittance and J final (x) represents the defogging image.
9. An image defogging system, comprising: the acquisition unit is used for acquiring linear polarization images and circular polarization images in the same scene at the same moment; The first processing unit is used for preprocessing the linear polarization image and the circular polarization image in the same scene at the same moment so that the sizes and the target positions of each pair of linear polarization image and circular polarization image are the same; the second processing unit is used for carrying out polarization difference processing on the preprocessed linear polarization image to obtain a linear polarization difference image and constructing initial transmissivity of linear polarization leading; The third processing unit is used for carrying out polarization difference processing on the preprocessed circular polarization image to obtain a circular polarization difference image, and carrying out stability constraint correction on initial transmittance of linear polarization leading by utilizing the circular polarization difference image to obtain final transmittance; and a fourth processing unit for recovering to obtain a defogging image based on the final transmittance.
10. A computer-readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the image defogging method of any of the claims 1-8.

Description

Image defogging method, system and storage medium Technical Field The present invention relates to the field of image defogging technology, and in particular, to an image defogging method, system and storage medium. Background In recent years, with the development of image processing technology, the requirements for defogging images are higher and higher, and a large amount of interference signals generated by medium scattering can be introduced in the imaging process under the scattering medium such as smoke or haze, so that the problems of overall contrast reduction, unclear edge information, submerged details and the like of the acquired images are caused. Aiming at the phenomenon, defogging treatment is started by utilizing polarization information, and scattering components in an image are distinguished and weakened by analyzing difference characteristics of the scattering light and target reflected light in the polarization state, so that the imaging quality is effectively improved. In the polarization defogging research at the present stage, most of the research is focused on outdoor environment, most of fog is mist, light rays are scattered for a single time, and the scattering components are separated through polarization degree estimation and atmospheric scattering light estimation. However, multiple scattering exists in light in a dense smoke environment, obvious depolarization phenomenon is easy to occur to linear polarization signals, so that the fluctuation of an estimated result of the transmissivity is large, the problems of overlarge local enhancement, distortion of texture details and the like occur in a restoration process due to optimization, and meanwhile, the defogging effect is not obvious by optimizing an atmospheric scattered light calculation mode in an indoor environment. In order to solve the defogging problem in the dense smoke environment, some methods adopt to introduce a complex deep learning model, which improves the defogging effect to a certain extent but also increases the computational burden of the algorithm and the industrial deployability. Therefore, there is a need to design a polarization defogging method with simple structure, light weight and good stability under the scattering of dense smoke in indoor environment. Disclosure of Invention Based on the technical problems in the background art, the invention provides an image defogging method, an image defogging system and a storage medium, which are used for carrying out stability constraint on a linear polarization defogging result by introducing circular polarization information, and effectively improving the defogging effect and the visual quality of an image in an indoor strong smoke scattering environment under the condition of not depending on a deep learning model. The invention provides an image defogging method, which comprises the following steps: S1, acquiring a linear polarization image and a circular polarization image in the same scene at the same moment; S2, preprocessing linear polarization images and circular polarization images in the same scene at the same moment to enable the sizes and target positions of each pair of linear polarization images and circular polarization images to be the same; s3, carrying out polarization difference processing on the preprocessed linear polarization image to obtain a linear polarization difference image, and constructing initial transmissivity of linear polarization leading; S4, carrying out polarization difference processing on the preprocessed circular polarization image to obtain a circular polarization difference image, and carrying out stability constraint correction on initial transmittance of linear polarization leading by utilizing the circular polarization difference image to obtain final transmittance; S5, recovering to obtain a defogging image based on the final transmissivity. Preferably, in S1, a pair of orthogonal linear polarization images I 1、I2 and a pair of orthogonal circular polarization images I 3、I4 are acquired in the same scene at the same time by using an amplitude-division imaging system. Preferably, the linear polarization image and the circular polarization image are preprocessed in S2 by inversion, sift registration and clipping. Preferably, the method for constructing the initial transmittance in S3 is as follows: s31, constructing a scattering model; S32, carrying out polarization difference treatment on the preprocessed circularly polarized image to obtain a linear polarization difference diagram; S33, defining linear polarization scattering asymmetry and calibration coefficient by using a linear polarization differential diagram; s34, obtaining initial transmissivity of linear polarization dominant according to the linear polarization scattering asymmetry and the calibration coefficient. Preferably, the scattering model is as follows: ; Where x denotes the pixel position, J (x) denotes the haze-free target, T (x) denotes the tran