CN108876738A - A kind of mono- scape image internal brightness compensation method of SAR based on gamma correction - Google Patents

Abstract

The invention belongs to the technical field of digital image processing, and specifically relates to a gamma correction-based SAR single-scene image internal brightness compensation method, which includes the following specific steps: S1: find the maximum value of pixel grayscale by traversing the image; S2: obtain Normalized image; S3: Perform gamma correction; S4: Make a difference image, make a difference image between the normalized image and the corrected image; S5: Adjust the gray level, multiply the difference image by the pixel gray level The maximum value, the gray scale range of the obtained difference image is [0-255], and the result image of adjusting the gray scale is the final result image processed by the algorithm; S6: Output image, output the image processed by S5. The gamma correction-based SAR single-scene image internal brightness compensation method of the present invention has the characteristics of being able to realize accurate compensation of image brightness, and its color uniformity effect and information retention are better.

Description

A Gamma Correction-Based Internal Brightness Compensation Method for SAR Single Scene Image

technical field

The invention belongs to the technical field of image processing, and in particular relates to a method for compensating internal brightness of a SAR single scene image based on gamma correction.

Background technique

Gaofen-3 satellite is a remote sensing satellite of China Gaofen Special Project, and it is the synthetic aperture radar (SAR) satellite with the most imaging modes in the world. The satellite imaging has a wide width, combined with the advantages of high spatial resolution, it can not only realize a large-scale general survey, but also a detailed survey of a specific area, which can meet the needs of different users for different target imaging.

However, due to changes in SAR imaging angles, terrain, radar antenna patterns, and strong backscattering of special ground objects, there is a problem of uneven brightness distribution inside a single SAR image, showing obvious bright spots, which seriously affects subsequent image processing and Interpretation, so it is particularly important to be able to effectively process the color of the image.

The Gaofen-3 satellite can obtain high-resolution SAR images, which have obvious details such as edges, points, and textures. Therefore, how to eliminate bright spots while retaining obvious detail information is the color-leveling process of Gaofen-3 SAR images. Another important goal.

In 2015, Wang Bangsong proposed a solution using frequency-domain low-pass filtering (Wang Bangsong. Research on SAR image automatic registration and mosaic method [D]. Wuhan University, 2015), see Figure 4, the solution flow of its frequency-domain low-pass filtering As follows: First, the original image is fast Fourier transformed to obtain a Fourier spectrogram, and a frequency-domain Gaussian filter is obtained by a given cutoff frequency, and the Fourier spectrogram is multiplied by a frequency-domain Gaussian filter to obtain a low-pass filter The spectrogram of the low-pass filtered spectrogram is subjected to inverse fast Fourier transform to obtain the background image, and finally the pixel value of the corresponding pixel of the original image and the background image is subtracted to obtain the result image. From the physical effect, the Fourier transform is to convert the image from the spatial domain to the frequency domain, and its inverse transform is to convert the image from the frequency domain to the spatial domain. In other words, the physical meaning of Fourier transform is to transform the grayscale distribution function of the image into the frequency distribution function of the image, and the inverse Fourier transform is to transform the frequency distribution function of the image into the grayscale distribution function. Before Fourier transform, the image is composed of Sampling in a continuous space obtains a set of points, and a two-dimensional matrix is usually used to represent each point in space. Because the space is three-dimensional and the image is two-dimensional, the relationship of objects in another dimension must be represented by gradients, so that we can know the corresponding relationship of objects in three-dimensional space by observing the image. Two-dimensional Fourier transform is performed on the image to obtain a spectrogram, which is the distribution map of the image gradient.

In image processing, many characteristics of the Gaussian low-pass filter make it widely used. The mathematical expression of the Gaussian low-pass filter in the frequency domain is:

H is a Gaussian low-pass filter, u, v are the horizontal and vertical coordinates in the frequency domain, respectively, D(u, v) is the plane distance from a point (u, v) in the frequency domain coordinate system to the coordinate origin (0,0) , D ₀ is the cutoff frequency of the filter.

It can be seen from the above formula that if the given cut-off frequency _D0 is too large, the value of the Gaussian low-pass filter will be too large; if the given cut-off frequency _D0 is too small, the value of the Gaussian low-pass filter will be too large. On the small side.

Low-pass filtering is to retain the low-frequency components in the image and remove the high-frequency components in the image. Both the edge and the noise in the image correspond to the high-frequency part in the frequency domain of the image, and the low-pass filtering in the frequency domain can remove or weaken the influence of noise and blur the edge contour. In order to better remove or weaken the influence of noise and blur the edge contour, the selection of the cut-off frequency D ₀ should not be too small.

The frequency-domain low-pass filtering algorithm can reduce the bright channel of the image to a certain extent, but it does not effectively eliminate the bright channel phenomenon. As a result, the gray level of the image still shows obvious differences; Larger, multiplied to get the low-pass filtered spectrogram and perform inverse fast Fourier transform to get the background image pixel value is greater than the original image pixel value, then the corresponding pixel gray value of the difference image will appear negative. The image data range is [0-255], and the pixels with negative values obtained by subtraction will be automatically filled to 0, and these pixels will display the original detailed information as black dots with a gray value of 0, causing the resulting image not only to lose the corresponding details, and black spots appear on the brightness compensation image. In short, it has not achieved good results in terms of color uniformity and information retention.

Contents of the invention

In order to solve the above-mentioned problems existing in the prior art, the present invention provides a SAR single-scene image internal brightness compensation method based on gamma correction, which utilizes the gamma correction method in power-law transformation to strictly ensure that the difference image is not There will be negative values, which reduces the loss of information, ensures the accuracy of image brightness adjustment, and has the characteristics of better color uniformity and information retention.

The technical problem to be solved in the present invention is realized through the following technical solutions:

Include the following specific steps:

S1: By traversing the image, find the maximum gray value I _max of the pixel.

S2: Obtain a normalized image

Divide the gray value of each pixel of the original image by the maximum gray value of the pixel I _max to obtain a normalized image, that is, map the gray value of the pixel of the original image to the range [0,1] to obtain the image I _N ; that is, the normalized image I _N ∈ [0,1];

S3: Perform gamma correction

For the normalized image Use gamma correction The corrected image is obtained, where c (0<c<1) is a constant, γ (γ>1) is a gamma factor, since I _N ∈ [0,1], 0<c<1, λ>1, It is easy to know that J _N ∈ [0,1), and only when I _N takes 0, J _N is 0;

S4: make difference image

For the normalized image and the rectified image Make a difference to obtain a difference image: ΔJ=I _N -J _N , where ΔJ≥0;

S5: Adjust the gray level

Multiplying the difference image ΔJ= _{IN-J N by} the maximum value of pixel grayscale I _max to obtain the result image of the adjusted gray level, the gray level range of the result image of the adjusted gray level is 0-255], and the result image of the gray scale adjustment is the final result image of algorithm processing;

S6: output image

Output the final result image processed by the algorithm.

Further, in S1, for the SAR single-view image, the image is in uint8 format, the gray value data range is [0-255], and each pixel is a gray value of [0-255], by looping through all The pixels traverse the original image I to find the maximum gray value I _max of the pixel.

Compared with the prior art, the present invention has the following beneficial effects:

The present invention utilizes the method of gamma correction in power-law transformation to process the normalized image, which not only enables the gamma-corrected image to obtain effective brightness compensation, but also enables the image to obtain a better uniform color effect; at the same time, the The gray value of the image pixel after gamma correction is not greater than the gray value of the original normalized image pixel, which avoids the negative value of corresponding subtraction between pixels, reduces the loss of information, and thus realizes accurate compensation of image brightness.

Description of drawings

Figure 1 is a flow chart of the method of the present invention.

Figure 2(a), Figure 2(b), and Figure 2(c) are image comparisons of a set of high-resolution three SAR images processed by different methods;

Wherein, Fig. 2 (a) is the unprocessed original image; Fig. 2 (b) is the image processed by frequency domain low-pass filtering, wherein, the cut-off frequency is 6; Fig. 2 (c) is after the method of the present invention processes The image of , where (c=0.7, γ=10).

Figure 3(a), Figure 3(b), and Figure 3(c) are image comparisons of another group of high-resolution three-sar images processed by different methods;

Wherein, Fig. 3 (a) is the unprocessed original image; Fig. 3 (b) is the image processed by frequency domain low-pass filtering, wherein, the cut-off frequency is 6; Fig. 3 (c) is after the method of the present invention processes The image of , where, (c=0.99, γ=20).

Fig. 4 is a flow chart of the low-pass filtering method in the frequency domain.

Detailed ways

In order to make the object, technical solution and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings. However, it should be understood that these descriptions are illustrative only and do not limit the scope of the present invention. In addition, in the following description, descriptions of some common knowledge and technologies are omitted to avoid unnecessary confusion of the understanding of the present invention.

Due to the change of SAR satellite imaging angle, terrain, radar antenna map and strong backscattering of special ground objects, there is a problem of uneven brightness distribution inside a single SAR image, showing obvious bright spots, which seriously affects the subsequent processing and processing of images. Interpretation, so it is particularly important to be able to effectively process the color of the image.

The Gaofen-3 satellite can obtain high-resolution SAR images, which have obvious details such as edges, points, and textures. Therefore, how to eliminate bright spots while retaining obvious details is the color-leveling process of Gaofen-3 SAR images. Another important goal.

The following are specific implementation procedures to achieve its desired processing results:

Referring to Fig. 1, a method for compensating internal brightness of a SAR single scene image based on gamma correction in the present invention, the specific steps are as follows:

S1: For a SAR single-scene image, the image is in uint8 format, the gray value data range is [0-255], each pixel is a gray value of [0-255], and the original image is traversed by looping through all pixels I, to find the maximum gray value I _max of the pixel.

S2: Obtain a normalized image

Divide the gray value of each pixel of the original image by the maximum gray value of the pixel I _max to obtain a normalized image, that is, map the gray value of the pixel of the original image to the range [0,1] to obtain the image I _N ; that is, the normalized image I _N ∈ [0,1];

S3: Perform gamma correction

For the normalized image Use gamma correction The corrected image is obtained, where c (0<c<1) is a constant, γ (γ>1) is a gamma factor, since I _N ∈ [0,1], 0<c<1, λ>1, It is easy to know that J _N ∈ [0,1), and only when I _N takes 0, J _N is 0;

S4: make difference image

For the normalized image and the rectified image Make a difference to obtain a difference image: ΔJ=I _N -J _N , where ΔJ≥0;

S5: Adjust the gray level

Multiplying the difference image ΔJ= _{IN-J N by} the maximum value of pixel grayscale I _max to obtain the result image of the adjusted gray level, the gray level range of the result image of the adjusted gray level is 0-255], and the result image of the gray scale adjustment is the final result image of algorithm processing;

S6: output image

Output the final result image processed by the algorithm.

The following is the specific result comparison and analysis:

Comparative example 1:

See Figure 2(a) for the unprocessed original image;

Figure 2(b) is the original image processed by low-pass filtering in the frequency domain, where the cutoff frequency is 6;

Fig. 2 (c) is the image obtained after the method of the present invention is processed, and wherein c=0.7, γ=10;

Comparing Figure 2(a) and Figure 2(b), we can see that, referring to Figure 2(a), the unprocessed original image has the problem of uneven brightness on the SAR image due to radiation correction, showing obvious bright spots Phenomenon. In order to carry out the subsequent processing smoothly, the brightness of the image must be compensated, that is, the bright channels in the image must be uniformly processed. Obviously, referring to Figure 2(b), the image processed by the frequency-domain low-pass filtering method can be seen to reduce the image bright channel to a certain extent, but it does not effectively eliminate the bright channel phenomenon, and the gray level of the result image is still It shows a more obvious difference, that is, it does not achieve better brightness compensation and color uniformity. In particular, since the low-pass filtering method in the frequency domain cannot guarantee that the gray level of the background image is strictly smaller than the original image, the gray level of the difference image will appear negative, thus losing the corresponding detail information and showing a lot of black spot phenomenon.

Among them, color uniformity is to correct the uneven distribution of brightness, contrast, hue, and saturation in one or more images, so that the brightness, contrast, hue, and saturation of each position of the image are basically consistent. Brightness compensation is a kind of uniform color processing.

Fig. 2 (c) is the image processed by the SAR single scene image internal brightness compensation method based on gamma correction in the present invention, referring to Fig. 2 (c), the brightness compensation method based on gamma correction proposed in the present invention, by normalizing The gamma correction of reasonable power-law transformation can effectively eliminate the influence of bright channel phenomenon and obtain a more ideal color uniformity effect. At the same time, the method proposed in the present invention can ensure that the gray level of the difference image will not have negative numbers, so that the detailed information of the image can be effectively preserved, the occurrence of black spots can be avoided, and subsequent processing can be carried out efficiently. In short, the proposed method is significantly better than the frequency domain low-pass filtering method in terms of color uniformity effect and information retention.

In summary, comparing Fig. 2(a), Fig. 2(b) and Fig. 2(c), it can be seen that the unprocessed original image SAR image has the problem of uneven brightness, showing obvious bright channel phenomenon; The image processed by low-pass filtering in the frequency domain can be seen to reduce the image bright channel to a certain extent, but it does not effectively eliminate the bright channel phenomenon. As a result, the gray level of the image still shows a relatively obvious difference; The gray level of the value image will appear negative, and the detailed information of the image cannot be effectively preserved; the image processed by the method of the present invention can effectively eliminate the influence of the bright channel phenomenon, and obtain a relatively ideal color uniform effect. The gray level of the value image will not appear negative, so the detailed information of the image can be effectively preserved, the appearance of black spots can be avoided, and the subsequent processing can be carried out efficiently.

In order to be able to further illustrate that the image effect after the method of the present invention is processed is better than the processing method of frequency domain low-pass filtering mentioned in the paper of Wang Bangsong (Wang Bangsong. Research on SAR image automatic registration and mosaic method [D]. Wuhan University, 2015), The following further comparisons were made:

Comparative example 2:

High score three SAR images of different groups from Comparative Example 1; see Figure 3(a), see Figure 3(b), see Figure 3(c);

Figure 3(a) is the unprocessed original image;

Figure 3(b) is the original image processed by low-pass filtering in the frequency domain, where the cutoff frequency is 6;

Fig. 3 (c) is the image obtained after the method of the present invention is processed, and wherein c=0.99, γ=20;

Comparing Figure 3(a), Figure 3(b) and Figure 3(c), it can also be obtained that the unprocessed original image SAR image has the problem of uneven brightness, showing obvious bright channel phenomenon; It can be seen that the image after low-pass filtering can reduce the bright channel to a certain extent, but it does not effectively eliminate the bright channel phenomenon. As a result, the gray level of the image still shows obvious differences; at the same time, because the difference image Negative numbers will appear in the gray level, and the detailed information of the image cannot be effectively preserved; the image processed by the method of the present invention can effectively eliminate the influence of the bright channel phenomenon, and obtain a relatively ideal color uniformity effect, and the method proposed by the present invention can ensure that the difference image There will be no negative gray level, so the detailed information of the image can be effectively preserved, the appearance of black spots can be avoided, and the subsequent processing can be carried out efficiently.

It can be seen that the method of the present invention can effectively eliminate the bright channel phenomenon and obtain a relatively ideal color uniformity effect; at the same time, it can ensure that the gray level of the difference image will not have negative numbers, so that the detailed information of the image can be effectively preserved and black spots can be avoided. the emergence of the phenomenon.

The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be assumed that the specific implementation of the present invention is limited to these descriptions. For those of ordinary skill in the technical field of the present invention, without departing from the concept of the present invention, some simple deduction or replacement can be made, which should be regarded as belonging to the protection scope of the present invention.

Claims (2)

1. A method for compensating brightness inside a SAR single scene image based on gamma correction, characterized in that, the specific steps are as follows: S1: By traversing the image, find the maximum pixel gray value I _max ; S2: Obtain a normalized image Divide each pixel gray value of the original image by the maximum pixel gray value I _max to obtain a normalized image, that is, map the pixel gray value of the original image to the range [0,1] to obtain the image I _N ; That is, the normalized image I _N ∈ [0,1]; S3: Perform gamma correction For normalized images Use gamma correction The corrected image is obtained, where c (0<c<1) is a constant, γ (γ>1) is a gamma factor, since I _N ∈ [0,1], 0<c<1, λ>1, It is easy to know that J _N ∈ [0,1), and only when I _N takes 0, J _N is 0; S4: make difference image For normalized images and corrected image Make difference image: ΔJ=I _N -J _N , where, ΔJ≥0; S5: Adjust the gray level Multiply the difference image ΔJ=I _N -J _N by the maximum value of pixel grayscale I _max to get the grayscale range of the difference image is [0-255], and the resulting image after adjusting the grayscale is the final result of the algorithm processing result image; S6: output image Output the image processed by S5. 2. A kind of SAR single scene image internal brightness compensation method based on gamma correction according to claim 1, it is characterized in that, described S1 is specifically: For the SAR single-view image, the image is in uint8 format, the gray value data range is [0-255], each pixel is a gray value of [0-255], and the original image I is traversed by looping through all the pixels, Find the maximum pixel gray value I _max .