An image restoration method for mine heavy dust environment

Abstract

This invention discloses an image restoration method for heavy dust environments in mines, comprising: acquiring a heavy dust image and performing brightness reduction processing to obtain a preliminary processed image; calculating the ambient light estimate of the preliminary processed image; acquiring an image in a dust-free environment as a reference image; constructing a window function; processing the preliminary processed image and the reference image according to the window function; calculating dust evaluation parameters; calculating a noise term based on the ambient light estimate and the dust evaluation parameters; and processing the preliminary processed image based on the noise term to obtain the final restored image. This invention effectively processes images from heavy dust environments in mines, reducing the large-area distortion or severe information loss that often occurs when using classic dark channel algorithms and other enhancement algorithms to process such images. It preserves more detail of the subject, and the color restoration is more realistic. Furthermore, it can more effectively suppress pixel oversaturation problems when there are light sources in the scene.

Description

Technical Field

[0001] This invention belongs to the field of image restoration technology, and in particular relates to an image restoration method for heavy dust environments in mines. Background Technology

[0002] Imaging is an effective means of acquiring visual information. For imaging systems such as cameras, achieving high-quality information output requires a particularly ideal environment. Mines are precisely such environments, making high-quality imaging impossible. The main reason is the presence of a large amount of dust in the mine environment. The scattering and absorption effects of dust cause the captured images to appear blurred, resulting in reduced image quality and unclear images. Since monitoring the conditions in mines involves mine safety, good monitoring footage and real-time information acquisition can greatly reduce the occurrence of accidents. Therefore, effectively removing the influence of environmental dust from images and restoring them to be closer to the ideal scene is essential.

[0003] Methods for removing dust from images can be broadly categorized into two types: image enhancement algorithms and physical model-based image restoration methods. Image enhancement algorithms, failing to consider the causes of image quality degradation, often result in distortion and loss of detail. Physical model-based image restoration methods, on the other hand, consider the environmental medium as the primary cause of image haziness and blur. The image restoration process involves establishing an image scattering model based on image degradation mechanisms and utilizing prior knowledge and assumptions about image degradation to restore the scene. This type of method is currently a hot research topic and a challenge. A classic example is He Kaiming's dark channel prior algorithm. The dark channel algorithm is primarily used for image restoration in foggy conditions, where the background is generally whitish and the fog is relatively uniform. However, directly applying the dark channel method to mine images often results in darker images, unnatural colors, overexposure, and an inability to suppress light sources. These problems stem from the prior knowledge no longer being applicable to the specific environment, leading to incorrect ambient light values ​​and transmittance. Summary of the Invention

[0004] The purpose of this invention is to provide an image restoration method for heavy dust environments in mines, so as to solve the problems existing in the prior art.

[0005] To achieve the above objectives, the present invention provides an image restoration method for heavy dust environments in mines, comprising:

[0006] Acquire a heavy dust image and perform brightness reduction processing to obtain a preliminary processed image; calculate the ambient light estimate value of the preliminary processed image.

[0007] Obtain an image from a dust-free environment as a reference image;

[0008] Construct a window function, process the preliminary processing image and the reference image according to the window function, and calculate the dust evaluation parameters;

[0009] Based on the ambient light estimate and the dust evaluation parameters, a noise term is calculated, and the preliminary image is processed based on the noise term to obtain the final restored image.

[0010] Optionally, the process of obtaining the ambient light estimate of the preliminary processed map includes:

[0011] Calculate the ambient light values ​​for the three color channels of the preliminary processed image, and take the minimum value as the reference ambient light value; set the ambient light coefficients for the three color channels, and obtain the ambient light estimates for the three color channels based on the reference ambient light value and the ambient light coefficients, and sum them to obtain the ambient light estimate of the preliminary processed image.

[0012] Optionally, the ambient light coefficient for a single color channel can be set based on the sum of the maximum values ​​of each column of a single color channel and the sum of the maximum values ​​of each column of the three color channels.

[0013] Optionally, the process of calculating dust evaluation parameters includes: calculating gradients for the three color channels of the preliminary processed image and the reference image respectively, obtaining horizontal gradient layers and vertical gradient layers and merging them to obtain the merged gradients of each color channel of each image; performing mean filtering on the merged gradients based on the window function, taking the average value of the merged gradients of the three color channels of the preliminary processed image and the reference image as the average gradient, and obtaining dust evaluation parameters based on the average gradient of the preliminary processed image and the average gradient of the reference image, wherein the window function sets the window size to a set value for the mean filtering convolution kernel.

[0014] Optionally, the process of calculating the noise term based on the ambient light estimate and the dust evaluation parameters includes:

[0015] The fitting relationship between transmittance and image pixel distribution is obtained through simulation experiments. The initial transmittance is calculated based on the dust evaluation parameters and the fitting relationship. The initial transmittance is optimized to obtain the final transmittance. The noise term of each color channel is calculated based on the final transmittance, the ambient light estimate, and the brightness distribution of the preliminary processed image.

[0016] Optionally, the noise terms and final transmittance of each color channel in the preliminary processed image are solved, and the results are merged to obtain the final restored image.

[0017] The technical effects of this invention are as follows:

[0018] This invention provides an image restoration method for heavy dust environments in mines. It can more effectively process images of heavy dust environments in mines, reduce the large-area distortion or severe information loss that occurs when classic dark channel algorithms and other enhancement algorithms process such images, retain more detail information of the subject, restore colors in a way that is consistent with the actual situation, and more effectively suppress pixel oversaturation when there are light sources in the scene. Attached Figure Description

[0019] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0020] Figure 1 This is a flowchart of the method in an embodiment of the present invention. Detailed Implementation

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0023] Example 1

[0024] like Figure 1 As shown, this embodiment provides an image restoration method for heavy dust environments in mines, including:

[0025] First, the heavy dust image is preliminarily processed to obtain the preliminarily processed image defog. 1 The initial processing aims to reduce the brightness of the image.

[0026] Calculate the initial processed image defog 1 Ambient light value The minimum value of the three channels is used to obtain the reference ambient light value. To reduce the color cast caused by the light source, an additional coefficient for the ambient light value is set for the three color channels. Based on the pixel characteristics of the image, the brightest pixels often reflect information about the light source. Therefore, the coefficient for a certain color channel 'c' is calculated as the sum of the maximum values ​​in each column of that channel divided by the sum of the maximum values ​​in each column of the other three channels.

[0027]

[0028] x and y are the horizontal and vertical resolutions of the image. Multiplying the coefficients by the reference value yields the ambient light value for each channel.

[0029]

[0030] The next step is to solve the image defog. 1 transmittance We conducted extensive simulation experiments in heavy dust environments, and from these experiments, we fitted the relationship between transmittance and image pixel value distribution. Applying this relationship to mine images, we first need to define a window function F(pic) to process heavy dust and dust-free mine images, where pic is the input image. The execution flow of the window function is as follows:

[0031] ① First, calculate the gradient for each of the three color channels of the input image to obtain two gradient layers: horizontal and vertical. The horizontal gradient of the C color channel is... The gradient in the longitudinal direction is Combine the gradients from the two directions into a single gradient. C :

[0032]

[0033] ② Set a mean-filtered convolution kernel W with a window size of m*m, perform mean filtering on the gradients of the three colors, and take the average value to obtain the average gradient avg of the image:

[0034]

[0035] The window function is used to calculate the average gradient (avg) of pixels. It is used to process the reference image J0 and the pre-processed image defog. 1 ,get and The reference image used is the dust-free image at the same camera position, and the dust parameter V is further calculated:

[0036]

[0037] The initial transmittance is:

[0038] The constants A and B are derived from experimental simulation data, and further... The data in the data is constrained to ensure transmittance is greater than 0 and less than 1, and then stretched to [t]. min ,t max ](0≤t min <t max The final transmittance is obtained within the range of ≤1). Layers:

[0039]

[0040] Where M t and m t They are respectively The maximum and minimum values ​​in; based on the obtained... and Solve for the final processed image J:

[0041]

[0042]

[0043] Where k is the image brightness coefficient, which is related to the overall brightness value of the image, J C This is the final processed image for color channel C, which converts the J values ​​of the three channels... C Merging results in the final processed image J.

[0044] Example 2

[0045] The original image I was subjected to brightness reduction processing to obtain the preliminary processed image defog. 1 Here, a simple dark channel algorithm can be used for preliminary processing;

[0046] Calculate the initial processed image defog 1 Ambient light value The minimum value of the three channels is used to obtain the reference ambient light value. Then calculate the coefficients for the three-channel ambient light values. The coefficient for a certain color channel c is the sum of the maximum values ​​in each column of that channel divided by the sum of the maximum values ​​in each column of the three channels:

[0047]

[0048] x and y represent the horizontal and vertical resolutions of the image. Multiplying the coefficients by the reference value yields the ambient light values ​​for each channel.

[0049]

[0050] The next step is to solve the image defog. 1 transmittance First, define a window function F(pic), where pic is the input image. The execution flow of the window function is as follows:

[0051] ① First, calculate the gradient for each of the three color channels of the input image to obtain two gradient layers: horizontal and vertical. The horizontal gradient of the C color channel is... The gradient in the longitudinal direction is Combine the gradients from the two directions into a single gradient. C :

[0052]

[0053] ② Set a mean-filtered convolution kernel W with a window size of 50*50, perform mean filtering on the gradients of the three colors, and take the average value to obtain the average gradient avg of the image:

[0054]

[0055] The window function is used to calculate the average gradient avg, and is used to process the reference image J0 and the pre-processed image defog. 1 ,get and avg J0 (i,j), the reference image used is the dust-free image under the same camera position, and then the dust parameter V is calculated:

[0056] The transmittance is: To constrain the data in t0 so that the transmittance is greater than 0 and less than 1, the final transmittance is obtained. Layers:

[0057]

[0058] Where M t and m t These are the maximum and minimum values ​​in t0, respectively; based on the obtained... and Calculate the noise term L C :

[0059]

[0060] The three channels are then substituted into the atmospheric scattering model to solve for the original image: J of the three channels C The images are merged into J, and the final processed image J is output.

[0061] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for image restoration for mine heavy dust environment, characterized in that, Includes the following steps: Acquire a heavy dust image and perform brightness reduction processing to obtain a preliminary processed image; calculate the ambient light estimate value of the preliminary processed image. Obtain an image from a dust-free environment as a reference image; A window function is constructed, which is a mean-filtered convolution kernel with a set window size. The preliminary processed image and the reference image are processed according to the window function. The horizontal and vertical gradients of the three color channels of the preliminary processed image and the reference image are calculated respectively. The horizontal and vertical gradients of the same color channel are merged to obtain the merged gradient of each color channel. The merged gradient is then subjected to mean filtering using the window function. The average value of the merged gradients of the three color channels of the processed preliminary processed image and the reference image is taken as their respective average gradients. Dust evaluation parameters are calculated based on the ratio of the average gradient of the preliminary processed image to the average gradient of the reference image. The dust evaluation parameters are used to quantify the degree of image degradation caused by dust. Based on the ambient light estimate and the dust evaluation parameters, a noise term is calculated, and the preliminary image is processed based on the noise term to obtain the final restored image.

2. The image restoration method for heavy dust environments in mines according to claim 1, characterized in that, The process of obtaining the ambient light estimate of the preliminary processed image includes: Calculate the ambient light values ​​for the three color channels of the preliminary processed image, and take the minimum value as the reference ambient light value; set the ambient light coefficients for the three color channels, and obtain the ambient light estimates for the three color channels based on the reference ambient light value and the ambient light coefficients, and sum them to obtain the ambient light estimate of the preliminary processed image.

3. The image restoration method for heavy dust environments in mines according to claim 2, characterized in that, The ambient light coefficient for a single color channel is set based on the sum of the maximum values ​​of each column of a single color channel and the sum of the maximum values ​​of each column of the three color channels.

4. The image restoration method for heavy dust environments in mines according to claim 1, characterized in that, The process of calculating the noise term based on the ambient light estimate and the dust evaluation parameters includes: The fitting relationship between transmittance and image pixel distribution is obtained through simulation experiments. The initial transmittance is calculated based on the dust evaluation parameters and the fitting relationship. The initial transmittance is optimized to obtain the final transmittance. The noise term of each color channel is calculated based on the final transmittance, the ambient light estimate, and the brightness distribution of the preliminary processed image.

5. The image restoration method for heavy dust environments in mines according to claim 1, characterized in that, The noise term and final transmittance of each color channel are used to solve the problem for each color channel of the preliminary processed image. The results are then combined to obtain the final restored image.

Images