Remote sensing image visualization enhancement method based on guiding filtering

Through the guide filtering method, the problems of excessive exposure areas and low contrast in remote sensing images during global linear compression and enhancement are solved, and the rich color and high contrast visualization enhancement of the image are achieved, and the image details and highlighted areas are retained. Color information.

CN118967455APending Publication Date: 2024-11-15CHANGGUANG SATELLITE TECH CO LTD

Patent Information

Application Number
CN202411037754.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

Existing remote sensing image visualization algorithms are difficult to accurately grasp the amount of data at both ends of the image during global linear compression and enhancement, resulting in excessive exposure areas or low contrast.

Method used

Using a guide filtering method, linear stretching and NDVI index enhancement are performed by counting pixel values, and the basic layer and detail layer of the brightness map are extracted by using a guide filter driven by a local linear model, and logarithmic transformation and dynamic range compression are performed. Finally, visual enhanced image is obtained through exponential transformation.

Benefits of technology

Effectively retain image detail texture and highlighted area color information, achieving rich color and high contrast visualization enhancement of the image, avoiding the problems of overexposure and low contrast of the image.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967455A_ABST
    Figure CN118967455A_ABST
Patent Text Reader

Abstract

The invention relates to a remote sensing image visualization enhancement method based on guiding filtering, and relates to the technical field of remote sensing image enhancement. The method comprises the following steps: counting DN values of an original satellite remote sensing image to be processed; performing linear stretching on the remote sensing image; carrying out NDVI (Normalized Difference Vegetation Index) enhancement on the normalized image; weighting is carried out by using the normalized value of each spectrum segment; guiding filtering processing is carried out on the brightness map; compressing the base layer; performing exponential transformation on the number domain enhancement layer; and calculating by using the enhancement layer, the brightness map and the original satellite remote sensing image to obtain a visual enhanced image map. According to the remote sensing image visualization enhancement method based on guided filtering, detail layer and base layer extraction is carried out on the image brightness map by using the guided filtering algorithm, and only the base layer is compressed, so that the remote sensing image can be visually enhanced under the condition that local detail texture features of the image are ensured and no more exposure areas are generated. Therefore, the color-rich and high-contrast visual enhancement processing of the remote sensing image is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of remote sensing image enhancement, and in particular to a remote sensing image visualization enhancement method based on guided filtering. Background Art

[0002] In remote sensing images, remote sensing images with 12-bit quantization are widely used, which has a quantization accuracy of 4096 levels. The high dynamic range of remote sensing images can obtain more information about ground objects, which is conducive to image recognition and matching. However, ordinary display devices can only display images with 8 bits or less. Therefore, to display 12-bit quantized high-dynamic remote sensing images, it is necessary to compress the remote sensing image bit number, that is, it is necessary to perform visualization enhancement processing on the 12-bit quantized remote sensing images to improve the visual effect of high dynamic range images displayed under low dynamic range conditions. The visualization processing results are displayed through the medium, and under the visual characteristics of the human eye, the effect of restoring the original scene as much as possible is achieved, making the image color natural, real, and rich, which is convenient for visual interpretation, image classification and extraction, etc.

[0003] At present, the traditional methods of remote sensing image visualization mainly include global or local enhancement methods such as histogram truncation, logarithmic stretching, histogram matching, gamma transformation, piecewise linear stretching transformation, and manual adjustment. The above methods will cause the image to lose dark and bright details, resulting in dark shadows and severe exposure. Even through parameter setting and adjustment, due to large scene differences, complex environmental weather, and large dynamic range, there will still be problems such as loss of detail information and low contrast, which places high requirements on the robustness of the visualization algorithm. Summary of the invention

[0004] The present invention aims to solve the technical problem in the prior art that the remote sensing image visualization algorithm is difficult to accurately grasp the amount of data intercepted or retained at both ends of the image during global linear compression and enhancement, resulting in excessive image exposure areas or low contrast, and provides a remote sensing image visualization enhancement method based on guided filtering.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0006] A remote sensing image visualization enhancement method based on guided filtering includes the following steps:

[0007] Step 1: Count the DN values ​​of the original satellite remote sensing images to be processed, and count the thresholds of pixel values ​​less than a certain percentage and greater than a certain percentage;

[0008] Step 2: According to the percentage cutoff threshold calculated in step 1, the remote sensing image is linearly stretched to obtain a normalized remote sensing image;

[0009] Step 3: Using the normalized remote sensing image obtained in step 2 as input, the normalized remote sensing image is enhanced by NDVI index to obtain an NDVI enhancement map;

[0010] Step 4: Take the NDVI enhancement map obtained in step 3 as input, use the normalized value of each spectrum band to weight it, and get the brightness map. At the same time, perform logarithmic transformation on the brightness map to get the logarithmic domain brightness map.

[0011] Step 5: Taking the brightness map obtained in step 4 as input, the brightness map is subjected to guided filtering, the base layer is separated after filtering, and the base layer is logarithmically transformed to obtain a logarithmic domain base layer; then the logarithmic domain brightness map obtained in step 4 is subtracted from the logarithmic domain base layer obtained in the previous step to obtain a logarithmic domain detail layer;

[0012] Step 6: Using the logarithmic domain detail layer, logarithmic domain base layer and dynamic range compression parameters obtained in step 5, compress the logarithmic domain base layer to obtain a compressed logarithmic domain enhancement layer;

[0013] Step 7: Perform an exponential transformation on the logarithmic domain enhancement layer obtained in step 6 to obtain an enhancement layer, and use the enhancement layer and the brightness map calculated in step 4 and the original satellite remote sensing image to calculate to obtain the final visual enhanced image.

[0014] In the above technical solution, in step 5,

[0015] In the guided filtering process, the guided image, the filtered input image, and the filtered output image are represented as logI, p, and q, respectively; the guided filter is driven by a local linear model, so it is assumed that the output and input of the guided filter function satisfy a linear relationship within a two-dimensional window, as shown below:

[0016]

[0017] Where i is the index of the pixel, k is the index of the local square window w with radius r, and ω k is a square window image, a k is the gain factor, b k is the bias coefficient; given the filtered input image p, the reconstruction error between the filtered output image q and p is minimized, and the cost function is:

[0018]

[0019] Among them, ε is the regularization parameter, and the minimum value of the cost function is obtained:

[0020]

[0021] b k =p k-a k u k ;

[0022] Among them, p k and u k Respectively represent p, the mean of logI in the window, σ k is the variance of logI within the window;

[0023] When the brightness image is guided and filtered, the guided filter is used as an edge-preserving filter, and logI=p, a k and b k The result is:

[0024]

[0025] b k =p k -a k p k .

[0026] In the above technical solution, in step 5,

[0027] In the guided filtering process, the guided image, the filtered input image, and the filtered output image are represented as logI, p, and q, respectively; the guided filter is driven by a local linear model, so it is assumed that the output and input of the guided filter function satisfy a linear relationship within a two-dimensional window, as shown below:

[0028]

[0029] Where i is the index of the pixel, k is the index of the local square window w with radius r, and ω k is a square window image, a k is the gain factor, b k is the bias coefficient; given the filtered input image p, the reconstruction error between the output image q and p is minimized, and the cost function is:

[0030]

[0031] Among them, ε is the regularization parameter, and the minimum value of the cost function is obtained:

[0032]

[0033] b k =p k -a k u k ;

[0034] Among them, p k and u k Respectively represent p, the mean of logI in the window, σk is the variance of logI within the window;

[0035] When the brightness image is guided and filtered, the guided filter is used as an edge-preserving filter, and logI=p, a k and b k The result is expressed as:

[0036]

[0037] b k =f mean (logI,r)-a k f mean (logI,r);

[0038] in,

[0039]

[0040] Where M is the input image, f mean (M, r) is the image obtained after mean filtering, s is the set of pixels in the template centered at point (i, j), and the template size is r×r;

[0041] base=f mean (a k ,r)*logI+f mean (b k ,r);

[0042] detail=logI-base;

[0043] Among them, detail is the logarithmic domain detail layer, and base is the logarithmic domain base layer.

[0044] In the above technical solution, in step 2, the percentage cutoff threshold range of the normalized remote sensing image is distributed in the interval [0,1].

[0045] In the above technical solution, the calculation formula for enhancing the NDVI index of the normalized remote sensing image in step 3 is as follows:

[0046]

[0047] Among them, GreenDn, RedDn, and NirDn are the original DN values ​​of the green, red, and near-infrared spectrum bands, respectively; GreenDn′ is the DN value of the green spectrum band after enhancement.

[0048] In the above technical solution, in step 4, the normalized values ​​of each spectrum segment are weighted to obtain a brightness map, and the brightness map is logarithmically transformed to obtain the calculation formulas of the logarithmic domain brightness map as follows:

[0049] I=0.2989*RedDn+0.587*GreenDn+0.114*BlueDn;

[0050] logI=log2I;

[0051] Among them, RedDn, GreenDn, BlueDn are the original DN values ​​of the red, green, and blue spectrum segments respectively, I is the brightness image, and the logarithm of I with base 2 is taken to obtain the logarithmic domain brightness image logI.

[0052] In the above technical solution, the compression parameter calculation formula in step 6 is as follows:

[0053] logEnh=base*compression+detail;

[0054] Among them, detail is the logarithmic domain detail layer, base is the logarithmic domain base layer, logEnh is the logarithmic domain enhancement layer, and compression is the dynamic compression parameter.

[0055] In the above technical solution, the calculation formula in step seven is as follows:

[0056] Perform exponential transformation on the number domain enhancement layer to obtain the enhancement layer. The formula is as follows:

[0057] Enh=2 logEnh ;

[0058] Among them, Enh is the enhancement layer, logEnh is the logarithmic domain enhancement layer;

[0059] Then the enhanced layer and the original satellite remote sensing image are used for calculation to obtain a visualized enhanced image map. The formula is as follows:

[0060] REnh=RedDn*Enh;

[0061] GEnh = GreenDn*Enh;

[0062] BEnh = BlueDn*Enh;

[0063] Among them, RedDn, GreenDn, BlueDn are the original DN values ​​of the red, green, and blue spectral segments, respectively; REnh, GEnh, and BEnh are the enhanced values ​​of the red, green, and blue spectral segments, respectively; and Enh is the enhancement layer.

[0064] The present invention has the following beneficial effects:

[0065] The remote sensing image visualization enhancement method based on guided filtering of the present invention retains a large amount of image detail texture and highlight area color information when the image is normalized by using smaller dark and bright percentage cutoff thresholds, thus providing conditions for subsequent enhancement processing.

[0066] The remote sensing image visualization enhancement method based on guided filtering of the present invention makes the vegetation area of ​​the remote sensing image more colorful through NDVI index enhancement processing.

[0067] The remote sensing image visualization enhancement method based on guided filtering of the present invention can ensure that the color white balance is not destroyed and prevent color cast by converting the three-spectrum RGB color image into a brightness image and converting the multi-spectrum color enhancement problem into a single-spectrum color enhancement problem.

[0068] The remote sensing image visualization enhancement method based on guided filtering of the present invention introduces the concepts of base layer and detail layer, performs visualization enhancement on the base layer in the logarithmic domain, retains the image detail layer information, and realizes image brightness enhancement through reasonable dynamic range compression settings.

[0069] The remote sensing image visualization enhancement method based on guided filtering of the present invention utilizes a guided filtering algorithm to extract detail layers and base layers of an image brightness map, and only compresses the base layer. The remote sensing image can achieve color-rich and high-contrast visualization enhancement processing of the remote sensing image while ensuring the local detail texture features of the image and not generating a large number of exposure areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0071] Figure 1 The figure is a schematic diagram of the overall process of the remote sensing image visualization enhancement method based on guided filtering of the present invention.

[0072] Figure 2 This is a schematic diagram of the original remote sensing image effect.

[0073] Figure 3 Schematic diagram of the normalized remote sensing image effect.

[0074] Figure 4 This is a schematic diagram of the brightness map effect.

[0075] Figure 5 This is a schematic diagram of the base layer effect.

[0076] Figure 6 This is a schematic diagram of the detail layer effect.

[0077] Figure 7 This is a schematic diagram to enhance the effect of the diagram.

[0078] Figure 8 This is a schematic diagram of the final visualization effect after processing.

[0079] Fig. 9 Schematic diagram of the local effect of the original remote sensing image.

[0080] Fig.10 Schematic diagram of the local effect of normalized remote sensing images.

[0081] Fig.11 This is a schematic diagram of the local effect of the brightness map.

[0082] Fig.12 This is a schematic diagram of the local effect of the base layer.

[0083] Fig.13 Schematic diagram of the local effect of the detail layer.

[0084] Fig.14 This is a schematic diagram of the local effect of the enhancement diagram.

[0085] Fig.15 This is a schematic diagram of the local visual effect after final processing. DETAILED DESCRIPTION

[0086] The inventive concept of the present invention is:

[0087] The present invention uses smaller dark and bright percentage cutoff thresholds, so that a large amount of image detail texture and highlight area color information are retained when the image is normalized, providing conditions for subsequent enhancement processing.

[0088] The present invention makes the vegetation area of ​​the remote sensing image more colorful through NDVI index enhancement processing.

[0089] The present invention converts a three-spectrum RGB color image into a brightness image and converts a multi-spectrum color enhancement problem into a single-spectrum color enhancement problem, thereby ensuring that the color white balance is not destroyed and preventing color cast.

[0090] The present invention introduces the concepts of base layer and detail layer, performs visualization enhancement on the base layer in the logarithmic domain, retains the image detail layer information, and realizes image brightness enhancement through reasonable dynamic range compression settings.

[0091] The following is combined with Figure 1 The present invention is described in detail. The remote sensing image visualization enhancement method based on guided filtering of the present invention comprises the following steps:

[0092] Step 1: Count the DN values ​​of the original satellite remote sensing images to be processed, and count the pixel values ​​that are less than a certain percentage and greater than a certain percentage, namely the dark percentage truncation threshold mindn and the bright percentage truncation threshold maxdn.

[0093] Step 2: According to the percentage cutoff threshold calculated in step 1, the original satellite remote sensing image is linearly stretched to obtain a normalized remote sensing image whose value range is distributed in the interval [0,1].

[0094] Step 3: Using the normalized remote sensing image obtained in step 2 as input, perform NDVI index enhancement on the normalized image, i.e. vegetation enhancement, to obtain an NDVI enhancement map. The calculation formula is shown in formula (1):

[0095]

[0096] Among them, GreenDn, RedDn, and NirDn are the original DN values ​​of the green, red, and near-infrared spectrum;

[0097] GreenDn′ is the DN value of the green spectrum after enhancement. After enhancement, the green of the vegetation is more realistic and the image color is fuller and richer.

[0098] Step 4: Take the NDVI enhancement map in step 3 as input, use the normalized values ​​of each spectrum band to weight, and obtain the brightness map. At the same time, perform logarithmic transformation on the brightness map to obtain the logarithmic domain brightness map. The calculation formulas are shown in formula (2) and formula (3):

[0099] I=0.2989*RedDn+0.587*GreenDn+0.114*BlueDn (2)

[0100] logI=log2I (3)

[0101] Among them, RedDn, GreenDn, BlueDn are the original DN values ​​of the red, green, and blue spectrum segments respectively, I is the brightness image, and the logarithm of I with base 2 is taken to obtain the logarithmic domain brightness image logI.

[0102] Step 5: Take the brightness map in step 4 as input, perform guided filtering on the brightness map, separate the base layer after filtering, perform logarithmic transformation on the base layer to obtain the logarithmic domain base layer, and then subtract the logarithmic domain brightness map obtained in step 3 from the logarithmic domain base layer obtained in step 4 to obtain the logarithmic domain detail layer.

[0103] In the guided filtering process, the guided image, the filtered input image, and the filtered output image are represented as logI, p, and q, respectively. The guided filter is driven by a local linear model, so it is assumed that the output and input of the guided filter function satisfy a linear relationship in a two-dimensional window, as shown in formula (4):

[0104]

[0105] Where i is the index of the pixel, k is the index of the local square window w with radius r, represents any i, ω k is a square window image, a k is the gain factor, b k is the bias coefficient. Given a filtered input image p, the reconstruction error between the output image q and p can be defined to be minimized. The cost function is:

[0106]

[0107] Here, is the ε regularization parameter, and we minimize the cost function to obtain:

[0108]

[0109] b k =p k -a k u k (7)

[0110] Among them, p k and u k Respectively represent p, the mean of logI in the window, σ k is the variance of logI within the window.

[0111] When the logarithmic domain brightness image is guided filtered, the guided filter is used for the edge preservation filter, so logI = p, then a k and b k The result can be expressed as:

[0112]

[0113] b k =p k -a k p k (9)

[0114] It can also be expressed as:

[0115]

[0116] b k =f mean (logI,r)-a k f mean (logI,r) (11)

[0117] in,

[0118]

[0119] Where M is the input image, f mean(M, r) is the image obtained after mean filtering, s is the set of pixels in the template centered at point (i, j), and the template size is r×r.

[0120] base=f mean (a k ,r)*logI+f mean (b k ,r) (13)

[0121] detail = logI-base (14)

[0122] Among them, detail is the logarithmic domain detail layer, and base is the logarithmic domain base layer.

[0123] Step 6: Using the logarithmic domain detail layer, logarithmic domain base layer and dynamic range compression parameters obtained in step 5, compress the base layer to obtain a compressed logarithmic domain enhancement layer;

[0124] logEnh=base*compression+detail (15)

[0125] Among them, detail is the logarithmic domain detail layer, base is the logarithmic domain base layer, logEnh is the logarithmic domain enhancement layer, and compression is the dynamic compression parameter.

[0126] Step 7: Perform an exponential transformation on the logarithmic domain enhancement layer in step 6 to obtain an enhancement layer, and use the enhancement layer, the brightness map calculated in step 3, and the original image to calculate to obtain the final visualized image (8-bit true color image). The calculation formula is shown in formulas (15)-(18):

[0127] Enh=2 logEnh (15)

[0128] REnh=RedDn*Enh (16)

[0129] GEnh=GreenDn*Enh (17)

[0130] BEnh=BlueDn*Enh (18)

[0131] Among them, Enh is the enhancement layer, logEnh is the logarithmic domain enhancement layer, RedDn, GreenDn, BlueDn are the original DN values ​​of the red, green and blue spectral segments respectively, and REnh, GEnh, BEnh are the enhanced values ​​of the red, green and blue spectral segments respectively.

[0132] Analysis of specific case images shows that Figure 2This is the original remote sensing image rendering. It can be seen that different objects have large color differences due to differences in DN values. Direct visual viewing of the image will result in brighter or darker areas, and it is difficult to take into account the details and color information of different objects. Fig. 9 This is a local rendering of the original image. It can be seen that there are large-area exposure problems in highlight areas such as the roof. The remote sensing image rendering after normalization and the local remote sensing image rendering are shown in Figure 3 , Fig.10 As shown in the figure, it can be seen that after normalization, the details of the image are retained and there is no large area of ​​exposure, but the image brightness is significantly reduced. The image needs to be enhanced to increase the brightness without losing image details. Figure 4 and Fig.11 It is a remote sensing image brightness map and local brightness map effect map, which converts the 3-channel color image into a single-channel black-and-white image, and uses a single-channel pixel value to represent the image brightness information for subsequent enhancement processing; Figure 5 and Fig.12 It is the extracted remote sensing image base layer and local remote sensing image base layer effect map. Compared with the brightness map, they weaken the image details and maintain the brightness value of the image. They can represent the basic brightness information of the image. At the same time, their texture features are weaker. Figure 6 and Fig.13 It is the rendering of the remote sensing image detail layer and local detail layer, which can clearly display the edge information and texture features of the objects in the image. Fig.13 It can be seen that the texture information in some roof areas is still clear, indicating that the detail layer extraction method can retain the texture features of the highlight area; Figure 7 and Fig.14 The effect diagram of the remote sensing image enhancement layer and the local enhancement layer shows the size of the brightness enhancement coefficient for different objects. The brighter the area, the more brightness enhancement, and the darker the area, the less brightness enhancement. Figure 7 It can be seen that the enhancement coefficients of farmland, bare soil and other areas are large due to the low brightness of the original image, while the enhancement coefficients of urban areas, industrial areas and other areas are small due to the high brightness of the original image. The brightness balance is achieved by calculating different enhancement coefficients for different areas. Figure 8 and Fig.15 These are the final visualization enhancement effect diagrams of the remote sensing image and the local final visualization enhancement effect diagrams. It can be seen that the enhanced image has moderate brightness and clear details. The brightness of low-brightness areas such as farmland is improved, the contrast is high, there is no large-area exposure phenomenon in high-brightness areas such as cities, and the texture is clear. This shows that the visualization method can compress the image dynamic range while retaining the image details, thereby achieving a natural color and high-contrast visualization display effect for remote sensing images.

[0133] In the above specific implementation manner of the present invention, RedDn, GreenDn and BlueDn in different formulas have the same meanings as RedDn, GreenDn and BlueDn, which are the original DN values ​​of the red, green and blue spectrum segments, respectively.

[0134] The remote sensing image visualization enhancement method based on guided filtering provided by the present invention can retain image details while compressing the image dynamic range without excessive exposure, thereby achieving a natural color and high-contrast visualization display effect of the remote sensing image.

[0135] The remote sensing image visualization enhancement method based on guided filtering of the present invention utilizes a guided filtering algorithm to extract detail layers and base layers of an image brightness map, and only compresses the base layer. The remote sensing image can achieve color-rich and high-contrast visualization enhancement processing of the remote sensing image while ensuring the local detail texture features of the image and not generating a large number of exposure areas.

[0136] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A remote sensing image visualization enhancement method based on guided filtering, characterized in that: The following steps are involved: Step 1: Count the DN values ​​of the original satellite remote sensing images to be processed, and count the thresholds of pixel values ​​less than a certain percentage and greater than a certain percentage; Step 2: According to the percentage cutoff threshold calculated in step 1, the remote sensing image is linearly stretched to obtain a normalized remote sensing image; Step 3: Using the normalized remote sensing image obtained in step 2 as input, the normalized remote sensing image is enhanced by NDVI index to obtain an NDVI enhancement map; Step 4: Take the NDVI enhancement map obtained in step 3 as input, use the normalized value of each spectrum band to weight it, and get the brightness map. At the same time, perform logarithmic transformation on the brightness map to get the logarithmic domain brightness map. Step 5: Taking the brightness map obtained in step 4 as input, the brightness map is subjected to guided filtering, the base layer is separated after filtering, and the base layer is logarithmically transformed to obtain a logarithmic domain base layer; then the logarithmic domain brightness map obtained in step 4 is subtracted from the logarithmic domain base layer obtained in the previous step to obtain a logarithmic domain detail layer; Step 6: Using the logarithmic domain detail layer, logarithmic domain base layer and dynamic range compression parameters obtained in step 5, compress the logarithmic domain base layer to obtain a compressed logarithmic domain enhancement layer; Step 7: Perform an exponential transformation on the logarithmic domain enhancement layer obtained in step 6 to obtain an enhancement layer, and use the enhancement layer and the brightness map calculated in step 4 and the original satellite remote sensing image to calculate to obtain the final visual enhanced image.

2. The remote sensing image visualization enhancement method based on guided filtering according to claim 1 is characterized in that: In step five, In the guided filtering process, the guided image, the filtered input image, and the filtered output image are represented as logI, p, and q, respectively; the guided filter is driven by a local linear model, so it is assumed that the output and input of the guided filter function satisfy a linear relationship within a two-dimensional window, as shown below: Where i is the index of the pixel, k is the index of the local square window w with radius r, and ω k is a square window image, a k is the gain factor, b k is the bias coefficient; given the filtered input image p, the reconstruction error between the filtered output image q and p is minimized, and the cost function is: Among them, ε is the regularization parameter, and the minimum value of the cost function is obtained: b k =p k -a k u k ; Among them, p k and u k Respectively represent p, the mean of logI in the window, σ k is the variance of logI within the window; When the brightness image is guided filtered, the guided filter is used for the edge preservation filter, and logI = p, a k and b k The result is: b k =p k -a k p k 。 3. The remote sensing image visualization enhancement method based on guided filtering according to claim 1 is characterized in that: In step five, In the guided filtering process, the guided image, the filtered input image, and the filtered output image are represented as logI, p, and q, respectively; the guided filter is driven by a local linear model, so it is assumed that the output and input of the guided filter function satisfy a linear relationship within a two-dimensional window, as shown below: Where i is the index of the pixel, k is the index of the local square window w with radius r, and ω k is a square window image, a k is the gain factor, b k is the bias coefficient; given the filtered input image p, the reconstruction error between the output image q and p is minimized, and the cost function is: Among them, ε is the regularization parameter, and the minimum value of the cost function is obtained: b k =p k -a k u k ; Among them, p k and u k Respectively represent p, the mean of logI in the window, σ k is the variance of logI within the window; When the brightness image is guided filtered, the guided filter is used for the edge preservation filter, and logI = p, a k and b k The result is expressed as: b k =f mean (logI,r)-a k f mean (logI,r); in, Where M is the input image, f mean (M, r) is the image obtained after mean filtering, s is the set of pixels in the template centered at point (i, j), and the template size is r×r; base=f mean (a k ,r)*logI+f mean (b k ,r); detail=logI-base; Among them, detail is the logarithmic domain detail layer, and base is the logarithmic domain base layer.

4. The remote sensing image visualization enhancement method based on guided filtering according to claim 1 is characterized in that: In step 2, the percentage cutoff threshold range of the normalized remote sensing image is distributed in the interval [0,1].

5. The remote sensing image visualization enhancement method based on guided filtering according to claim 1 is characterized in that: The calculation formula for enhancing the NDVI index of the normalized remote sensing image in step 3 is as follows: Among them, GreenDn, RedDn, and NirDn are the original DN values ​​of the green, red, and near-infrared spectrum bands, respectively; GreenDn′ is the DN value of the green spectrum band after enhancement.

6. The remote sensing image visualization enhancement method based on guided filtering according to claim 5 is characterized in that: In step 4, the normalized values ​​of each spectrum segment are weighted to obtain a brightness map, and the brightness map is logarithmically transformed to obtain the calculation formulas of the logarithmic domain brightness map as follows: I=0.2989*RedDn+0.587*GreenDn+0.114*BlueDn; logI=log2I; Among them, RedDn, GreenDn, BlueDn are the original DN values ​​of the red, green, and blue spectrum segments respectively, I is the brightness image, and the logarithm of I with base 2 is taken to obtain the logarithmic domain brightness image logI.

7. The remote sensing image visualization enhancement method based on guided filtering according to claim 2 or 3, characterized in that: The calculation formula for the compression parameter in step 6 is as follows: logEnh=base*compression+detail; Among them, detail is the logarithmic domain detail layer, base is the logarithmic domain base layer, logEnh is the logarithmic domain enhancement layer, and compression is the dynamic compression parameter.

8. The remote sensing image visualization enhancement method based on guided filtering according to claim 2 or 3, characterized in that: The calculation formula in step seven is as follows: Perform exponential transformation on the number domain enhancement layer to obtain the enhancement layer. The formula is as follows: Unit=2 logEnh ; Among them, Enh is the enhancement layer, logEnh is the logarithmic domain enhancement layer; Then the enhanced layer and the original satellite remote sensing image are used for calculation to obtain a visualized enhanced image map. The formula is as follows: REnh=RedDn*Enh; GEnh = GreenDn*Enh; BEnh = BlueDn*Enh; Among them, RedDn, GreenDn, BlueDn are the original DN values ​​of the red, green, and blue spectral segments, respectively; REnh, GEnh, and BEnh are the enhanced values ​​of the red, green, and blue spectral segments, respectively; and Enh is the enhancement layer.

Citation Information

Patent Citations

  • A Retinex enhancement method of color image based on guiding filter

    CN109544466A

  • Remote sensing surveying and mapping image enhancement method

    CN115908186A

  • Substation main transformer visible light image adaptive contrast enhancement method and system

    CN116681629A

Cited By

  • Multi-level self-adaptive remote sensing image stretching method, equipment and medium

    CN120163704A

  • A multi-level adaptive remote sensing image stretching method, device and medium

    CN120163704B