A relative radiometric correction method based on multi-temporal images

Abstract

This invention discloses a relative radiometric correction method based on multi-temporal imaging. The method first obtains the image to be corrected and the target image, performs channel separation and histogram calculation on the original and target images, and obtains the histogram of each channel. The cumulative distribution function and grayscale mapping are then calculated, and finally, the color channels are merged. The corrected red, green, and blue channels are combined into a corrected RGB image; this corrected RGB image is the final correction result. This method can reduce the impact of atmospheric transmission on images, thereby improving image quality and clarity. The corrected image will be more realistic, clear, and have a better visual effect.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing, and in particular relates to a relative radiation correction method based on multi-temporal images. Background Art

[0002] In the fields of computer vision and image processing, numerous techniques are used to improve image quality, enhance image features, and solve various image processing problems. In remote sensing image processing, radiometric brightness distortion often occurs due to occlusion and scattering effects caused by atmospheric transport, necessitating radiometric correction. Currently, several approaches to radiometric correction exist. The first involves the use of colorimetric colorimetry, which involves capturing images under standard lighting conditions using a standard color chart and then correcting the camera image to the standard color space based on the color chart. However, in the field of remote sensing imagery, field sampling of the scene at the time of capture is difficult to obtain. Furthermore, this method is complex, highly dependent on the light source, and suffers from large errors. A second approach involves white balance correction based on theoretical assumptions, including the grayscale world assumption and the perfect world reflection assumption. These assumptions modify image statistics to bring the image's color space closer to the neutral color seen by the naked eye. A third approach involves finding a mapping function between images based on pixel values. This function is then applied to the camera image to be corrected, transforming it to the target camera's color space before output. However, these methods suffer from limitations in accuracy, stability, and time complexity, and have thus far remained largely unacceptable for industrial use. Therefore, it is necessary to develop a highly accurate and stable radiation correction algorithm and reduce its time complexity. Summary of the Invention

[0003] To overcome the shortcomings of the existing technology, the present invention provides a relative radiometric correction method based on multi-temporal imagery. The method first obtains the image to be corrected and the target image. Channel separation and histogram calculation are performed on the original and target images to obtain the histograms of each channel. The cumulative distribution function and grayscale mapping are then calculated. Finally, the color channels are merged to form a corrected RGB image based on the corrected red, green, and blue channels. This corrected RGB image is the final correction result. The method can reduce the impact of atmospheric transmission on images, thereby improving image quality and clarity. The corrected image will be more realistic, clearer, and have a better visual effect.

[0004] The technical solution adopted by the present invention to solve the technical problem includes the following steps:

[0005] Step 1: Data preparation stage;

[0006] Step 1-1: Obtain the image to be corrected;

[0007] Acquire an image to be corrected from an image acquisition device or a satellite; the RGB image to be corrected is the original image to be corrected;

[0008] Step 1-2: Acquire the target image;

[0009] Obtaining a target image from a rectified image database, the target image serving as a reference standard for calibration;

[0010] Step 2: Calculate the histogram;

[0011] Step 2-1: First, perform channel separation on the original image and extract the pixel values of the red channel, green channel, and blue channel respectively; then, calculate the histogram of the pixel values of each channel and count the number of pixels with different gray levels in the image;

[0012] Step 2-2: Perform channel separation and histogram calculation on the target image to obtain the histogram of each channel of the target image;

[0013] Step 3: Calculate the cumulative distribution function;

[0014] Step 3-1: Normalize the histogram of each channel of the original image to obtain the probability density function of each channel of the original image; through normalization, map the pixel value to the range of [0,1];

[0015] In step 3-2, the histogram of each channel of the target image is also normalized to obtain the probability density function of each channel of the target image;

[0016] Step 4: Grayscale mapping value calculation;

[0017] Step 4-1: In each channel, find the closest mapping value based on the cumulative distribution function of each channel of the original image and the cumulative distribution function of each channel of the target image; the mapping value represents the grayscale value corresponding to the current pixel grayscale value in the target image;

[0018] Step 4-2: Based on the found mapping value, each pixel of the original image is mapped to the grayscale value of the target image channel according to the grayscale mapping to achieve correction; this process is applied to the red, green and blue channels respectively;

[0019] Step 5: Color channel merging;

[0020] According to the corrected red, green and blue channels, they are merged into a corrected RGB image; this corrected RGB image is the final correction result.

[0021] Preferably, the original image is first subjected to channel separation, and pixel values of the red channel, green channel and blue channel are extracted respectively as follows:

[0022] L=L B +L G +L R

[0023] L B =L[:,:,0]

[0024] L G =L[:,:,1]

[0025] L R =L[:,:,2]

[0026] Where L represents the grayscale image, which is used to represent the grayscale information of the original image; the formula is obtained by adding the pixel values of the red channel, green channel and blue channel; L B Represents the pixel value of the blue channel, which represents the blue component of all pixels in the original image; L G Represents the pixel value of the green channel, which represents the green component of all pixels in the original image; L R Represents the pixel value of the red channel, which represents the red component of all pixels in the original image; by adding these channels, the grayscale image L is obtained.

[0027] Preferably, the step 3 is specifically as follows:

[0028] Using the histogram data, the cumulative distribution functions of the red, green, and blue channels of the original image and the target image are calculated respectively; the cumulative distribution function is used to represent the cumulative distribution of pixel values in different ranges in the image;

[0029] Assume that the histogram of the original channel is H Corrected , the histogram of the target channel is H Target , where CDF represents the cumulative score. Assume that the histogram of the original channel is H Corrected , the histogram of the target channel is H Target , i and j are each represented by a pixel of the image, where CDF represents the cumulative distribution function:

[0030]

[0031] Preferably, the step 4 is specifically as follows:

[0032] Create an empty output image channel with the same size as the original image; perform grayscale mapping on each pixel of the original image; for each pixel value of each channel, find the closest mapping relationship of pixel values by comparing the cumulative distribution function (CDF) of the original image with the cumulative distribution function (CDF) of the target image, and fill the mapped pixel values into the output image channel:

[0033] map_value_R=Min(|CDF CR (i)-CDF TR (j)|)

[0034] map_value_G=Min(|CDF CG (i)-GDF TG (j)|)

[0035] map_value_B=Min(|CDF CB (i)-CDF TB (j)|)

[0036] where CDF CR Represents the cumulative distribution function of the red channel of the original histogram, CDF CG Represents the cumulative distribution function of the green channel of the original histogram, CDF CB represents the cumulative distribution function of the blue channel of the original histogram;

[0037] Correspondingly, CDF TR Represents the cumulative distribution function of the red channel of the target histogram, CDF TG Represents the cumulative distribution function of the green channel of the target histogram, CDF TB represents the cumulative distribution function of the blue channel of the target histogram;

[0038] For the red channel C R For each pixel value i in , fill the mapping value map_value into the output image channel:

[0039] C R [R=i]=map_value_R(i∈[0,256])

[0040] For the green channel C G For each pixel value i in , fill the mapping value map_value into the output image channel:

[0041] C G [G=i]=map_value_G(i∈[0,256])

[0042] For the blue channel C B For each pixel value i in , fill the mapping value map_value into the output image channel:

[0043] C B [B=i]=map_value_B(i∈[0,256])

[0044] The above formula represents the process of grayscale mapping based on the CDF of the original image and the target image, and filling the mapped pixel values into the output image channel to achieve image correction.

[0045] Preferably, the step 5 is specifically as follows:

[0046] Merge the grayscale-mapped red, green, and blue channels to get the corrected RGB image:

[0047] L=Stack(L B +L G +L R ).

[0048] The beneficial effects of the present invention are as follows:

[0049] 1. Improved image quality. This method reduces the impact of atmospheric transmission on images, thereby improving image quality and clarity. The corrected image will be more realistic, clearer, and have a better visual effect.

[0050] 2. Color correction accuracy. This method leverages image statistical information through histogram matching and cumulative distribution function mapping to achieve accurate image correction. Compared to traditional colorimetry methods and theoretical assumptions, this method offers higher correction accuracy and stability.

[0051] 3. Multi-scenario applicability. This method does not rely on specific scenes or light source conditions and can be applied to various image processing fields, especially for scenes that require atmospheric correction, such as aviation and aerospace remote sensing images.

[0052] 4. Simple and efficient. Compared with the traditional color card colorimetry method, this method is simpler to operate, does not require field sampling and strict control of light source conditions, and saves calibration costs and time.

[0053] 5. Innovation. This method combines histogram matching and maps the target image pixel values to the image to be corrected through the cumulative distribution function, avoiding the complex manual adjustment process. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 Flow chart of the method of the present invention.

[0055] Figure 2 1 is an example of an original image and a target image according to an embodiment of the present invention.

[0056] Figure 3 This is a comparison diagram of the original image before and after correction and the target image according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The present invention will be further described below with reference to the accompanying drawings and examples.

[0058] In response to the problems existing in the above-mentioned methods, the present invention discloses a relative radiation correction method based on multi-temporal images, which mainly solves the problems existing in traditional radiation correction methods, such as complex operations, strong dependence on light sources and large errors, as well as the challenges of accuracy, stability and time complexity in the current image processing field.

[0059] The main advantages of the present invention are concentrated in the following aspects.

[0060] 1. No need to sample color cards on site or rely on specific lighting conditions, making operation simpler and more convenient.

[0061] 2. Images can be corrected through pixel value mapping, which can accurately adjust the image tone while maintaining the image features, achieving the goal of correcting multi-phase images to a uniform tone.

[0062] The present invention relates to the field of image processing, and in particular to a relative radiometric correction method based on multi-temporal images, which aims to reduce the impact of atmospheric transmission on images and improve image quality, comprising the following steps:

[0063] Step 1: Data preparation phase

[0064] a. Acquire the image to be calibrated: First, acquire the image to be calibrated from an image acquisition device or satellite, such as infrared remote sensing images or aerospace images. These images may be affected by atmospheric transmission, resulting in color deviation or blurring.

[0065] b. Obtain a target image: Obtain a target image from a database of known good quality, rectified images. This target image serves as a reference for calibration. The target image should have similar scenes and features and have been previously atmospherically corrected, so its color and quality should be accurate.

[0066] Step 2: Calculate the histogram

[0067] a. For the image to be corrected, first perform channel separation, extracting pixel values for the red, green, and blue channels. Then, calculate a histogram for each channel's pixel values to count the number of pixels at different grayscale levels in the image. These histograms will be used for subsequent color correction.

[0068] b. Perform channel separation and histogram calculation on the target image in the same way to obtain the histogram of each channel of the target image.

[0069] These target histograms will serve as a reference to help correct images that need correction.

[0070] Step 3: Calculate the cumulative distribution function

[0071] a. Normalize the histogram of the original image channel to obtain the probability density function of the original channel. Normalization maps the pixel values to the range of [0, 1] to facilitate subsequent calculations.

[0072] b. Normalize the histogram of the target image channel in the same way to obtain the probability density function of the target channel.

[0073] These normalized histograms are used to calculate the cumulative distribution function.

[0074] Step 4: Grayscale mapping calculation

[0075] a. For each pixel grayscale value in the original image, within each channel, find the closest mapping value based on the cumulative distribution function of the original channel and the cumulative distribution function of the target channel. This mapping value represents the grayscale value corresponding to the current pixel grayscale value in the target image.

[0076] b. Based on the found mapping value, each pixel of the original image is mapped to the grayscale value of the target image channel according to the grayscale mapping to achieve correction. This process will be applied to the red, green and blue channels separately.

[0077] Step 5: Color channel merging

[0078] a. Combine the corrected red, green, and blue channels into a corrected RGB image. This corrected RGB image is the final correction result, effectively correcting any color deviations.

[0079] Example:

[0080] Step 1, prepare image data:

[0081] Before implementing the image correction method of the present invention, two sets of image data need to be prepared: the RGB image to be corrected and the target image. The RGB image to be corrected is the original image that needs to be corrected, and the target image is an image that is similar to the image to be corrected and has been corrected, that is, it has accurate color information. The image format includes common image formats, such as PNG images and TIF images. Here we define the two images as the image file to be processed named Corrected.tif and the target image file named Target.tif image. Figure 2 shown.

[0082] Step 2: Extract channel data:

[0083] Perform channel separation on the RGB image to be corrected and extract the pixel value data of the red channel, green channel and blue channel.

[0084] L=L B +LG +L R

[0085] L B =L[:,:,0]

[0086] L G =L[:,:,1]

[0087] L R =L[:,:,2]

[0088] Step 3, calculate the cumulative distribution function:

[0089] Using the histogram data, calculate the cumulative distribution function of the red, green and blue channels of the image to be corrected, as well as the cumulative distribution function of the red, green and blue channels of the target image. The cumulative distribution function is used to represent the cumulative distribution of pixel values in different ranges in the image. Assume that the histogram of the original channel is H Corrected , the histogram of the target channel is H Targdt . Where CDF represents the cumulative distribution function.

[0090]

[0091]

[0092] Step 4: Perform grayscale mapping:

[0093] To store the corrected pixel values, create an empty output image channel of the same size as the image to be corrected. Grayscale mapping is performed on each pixel of the image to be corrected. For each pixel value in each channel, the CDF of the image to be corrected is compared with the CDF of the target image to find the closest pixel value mapping relationship. The mapped pixel value is then filled into the output image channel.

[0094] map_value_R=Min(|CDF CR (i)-CDF TR (j)|)

[0095] map_value_G=Min(|CDF CG (i)-CDF TG (j)|)

[0096] map_value_B=Min(|CDF CB (i)-CDF TB (j)|)

[0097] For each pixel value i in the red channel, fill the output image channel with the mapping value map_value.

[0098] CR [R=o]=map_value_R(i∈[0,256])

[0099] C G [G=i]=map_value_G(i∈[0,256])

[0100] C B [B=i]=map_value_B(i∈[0,256])

[0101] The above formula represents the process of grayscale mapping based on the CDF of the image to be corrected and the target image, and the mapped pixel values are filled into the output image channel to achieve image correction.

[0102] Step 5: Merge channels:

[0103] The grayscale-mapped red, green, and blue channels are merged to obtain the corrected RGB image.

[0104] L=Stack(L B +L G +L R )

[0105] The corrected RGB image can be displayed to observe the correction effect. At the same time, the corrected RGB image can also be saved as a file for subsequent use. Figure 3 shown.

[0106] The color distribution of the corrected image obtained in the embodiment is nearly consistent with the target image, and the clarity and details are improved. Compared with other image correction methods, the method of the present invention is simple to operate and has accurate correction effects while maintaining image quality. It is a radiation correction method that is more suitable for the field of remote sensing images. This embodiment demonstrates a relative radiation correction method based on multi-temporal images, which eliminates color deviations caused by atmospheric transmission through grayscale mapping, effectively improving image quality and visual quality. In summary, the present invention has broad value and application prospects.

Claims

1. A relative radiometric correction method based on multi-temporal images, characterized in that: The steps include: Step 1: Data preparation stage; Step 1-1: Obtain the image to be corrected; Acquire an image to be corrected from an image acquisition device or a satellite; the RGB image to be corrected is the original image to be corrected; Step 1-2: Acquire the target image; Obtaining a target image from a rectified image database, the target image serving as a reference standard for calibration; Step 2: Calculate the histogram; Step 2-1: First, perform channel separation on the original image and extract the pixel values of the red channel, green channel, and blue channel respectively; then, calculate the histogram of the pixel values of each channel and count the number of pixels with different gray levels in the image; Step 2-2: Perform channel separation and histogram calculation on the target image to obtain the histogram of each channel of the target image; Step 3: Calculate the cumulative distribution function; Step 3-1: Normalize the histogram of each channel of the original image to obtain the probability density function of each channel of the original image; through normalization, map the pixel value to the range of [0,1]; In step 3-2, the histogram of each channel of the target image is also normalized to obtain the probability density function of each channel of the target image; Step 4: Grayscale mapping value calculation; Step 4-1: In each channel, find the closest mapping value based on the cumulative distribution function of each channel of the original image and the cumulative distribution function of each channel of the target image; the mapping value represents the grayscale value corresponding to the current pixel grayscale value in the target image; Step 4-2: According to the found mapping value, each pixel of the original image is mapped according to the grayscale mapping, and it is mapped to the grayscale value of the target image channel to achieve correction; This process is applied separately to the red, green, and blue channels; Step 5: Color channel merging; According to the corrected red, green and blue channels, they are merged into a corrected RGB image; this corrected RGB image is the final correction result; Create an empty output image channel with the same size as the original image; perform grayscale mapping on each pixel of the original image; for each pixel value of each channel, find the closest mapping relationship of pixel values by comparing the cumulative distribution function (CDF) of the original image with the cumulative distribution function (CDF) of the target image, and fill the mapped pixel values into the output image channel: map_value_R=Min(|CDF CR (i)-GDF TR (j)|) map_value_G=Min(|CDF CG (i)-CDF TG (j)|) map_value_B=Min(|CDF CB (i)-CDF TB (j)|) where CDF CR Represents the cumulative distribution function of the red channel of the original histogram, CDF cG Represents the cumulative distribution function of the green channel of the original histogram, CDF CB represents the cumulative distribution function of the blue channel of the original histogram; Correspondingly, CDF TR Represents the cumulative distribution function of the red channel of the target histogram, CDF TG Represents the cumulative distribution function of the green channel of the target histogram, CDF TB represents the cumulative distribution function of the blue channel of the target histogram; By calculating the mapping value, the pixel value of each channel of the original image is mapped to the grayscale value corresponding to the output image channel to obtain a reference image that has been corrected for radiation.

2. The relative radiometric correction method based on multi-temporal images according to claim 1, characterized in that: The original image is first subjected to channel separation, and the pixel values of the red channel, green channel and blue channel are extracted respectively as follows: L=L B +L G +L R L B =L[:,:,0] L G =L[:,:,1] L R =L[:,:,2] Where L represents the grayscale image, which is used to represent the grayscale information of the original image; the formula is obtained by adding the pixel values of the red channel, green channel and blue channel; L B Represents the pixel value of the blue channel, which represents the blue component of all pixels in the original image; L G Represents the pixel value of the green channel, which represents the green component of all pixels in the original image; L R Represents the pixel value of the red channel, which represents the red component of all pixels in the original image; By adding these channels, a grayscale image L is obtained.

3. The relative radiometric correction method based on multi-temporal images according to claim 2, characterized in that: The step 3 is specifically as follows: Using the histogram data, the cumulative distribution functions of the red, green, and blue channels of the original image and the target image are calculated respectively; the cumulative distribution function is used to represent the cumulative distribution of pixel values in different ranges in the image; Assume that the histogram of the original channel is H Corrected , the histogram of the target channel is H Target , where CDF represents the cumulative score. Assume that the histogram of the original channel is H Corrected , the histogram of the target channel is H Target , i and j are each represented by a pixel of the image, where CDF represents the cumulative distribution function:

4. The relative radiometric correction method based on multi-temporal images according to claim 3, characterized in that: The step 4 is specifically as follows: For the red channel C R For each pixel value i in , fill the mapping value map_value into the output image channel: C R [R=i]=map_value_R(i∈[0,256]) For the green channel C G For each pixel value i in , fill the mapping value map_value into the output image channel: C G [G=i]=map_value_G(i∈[0,256]) For the blue channel C B For each pixel value i in , fill the mapping value map_value into the output image channel: C B [B=i]=map_value_B(i∈[0,256]) The above formula represents the process of grayscale mapping based on the CDF of the original image and the target image, and filling the mapped pixel values into the output image channel to achieve image correction.

5. The relative radiometric correction method based on multi-temporal images according to claim 4, characterized in that: The step 5 is specifically as follows: Merge the grayscale-mapped red, green, and blue channels to get the corrected RGB image: L=Stack(L B +L G +L R )。

Images