CN111738929A - Image processing method and device, electronic device, and storage medium - Google Patents

Abstract

The embodiments of the present invention disclose an image processing method and device, an electronic device, and a storage medium, which can improve the image quality of radar imaging. The method includes: in at least two original images, obtaining the gray value of each original pixel in each original image; each original pixel is a pixel included in each original image; based on each orientation in each original image The standard deviation of the gray value of the direction is obtained, and each original image is corrected under the action of the preset low-pass filter model to obtain at least two first corrected images; the gray level of each azimuth in each first corrected image is obtained. The mean value, based on the mean value of the gray value of each azimuth, corrects each first corrected image under the action of a preset low-pass filter model, and obtains at least two target images; splicing the at least two target images into one Synthesize the image to complete the image processing process.

Description

Image processing method and device, electronic device, and storage medium

technical field

The present invention relates to the field of synthetic aperture radar imaging, in particular to an image processing method and device, electronic equipment and storage medium.

Background technique

Synthetic Aperture Radar (SAR) has become one of the important means of high-resolution earth observation and global management after 60 years of development since its birth in the late 1950s. Synthetic aperture radar, as an active remote sensor working in the microwave frequency band, compared with optical sensors, is not limited by sunlight and weather conditions, and can observe all-weather, all-day and all-round earth observation. There are important applications in the field of modern microwave remote sensing. In order to further reduce the global observation period and monitor rapidly changing large-scale surface phenomena, Scanning Synthetic Aperture Radar (ScanSAR) adopts the Burst working mode to obtain a larger mapping range of the sea and land environment, and has now become a spaceborne SAR. important direction of technological development.

Radiation correction is a key technology to obtain wide-range ScanSAR images with normal radiation. Early research on ScanSAR radiation correction focused on the signal processing stage, and considered eliminating scallop effects and stripe inhomogeneity in the echo data imaging process. All methods need to rely on a large amount of prior data such as radar parameters, and cannot be corrected only based on the original problem image, and the algorithm complexity is high. Radiation correction methods based on image post-processing have been extensively studied by scholars at home and abroad. That is, after completing the imaging processing of the echo data, only relying on the characteristics and information of the image itself, the peculiar scallop effect and the uneven stripe phenomenon appearing in the image are corrected, and the periodic stripes and large-scale bright bands are eliminated; Before image stitching, adjust the brightness, contrast, and brightness trend of the overlapping area of each two adjacent images to be consistent, so as to avoid seams in the stitched images, which will affect the visual effect and post-processing. However, in the absence of prior information such as radar parameters, or in the case of only correcting a single problem image, the methods of the prior art are not applicable and cannot improve the image quality of radar imaging.

SUMMARY OF THE INVENTION

The embodiments of the present invention are expected to provide an image processing method and apparatus, an electronic device, and a storage medium, which can improve the image quality of radar imaging.

The technical scheme of the present invention is realized as follows:

In a first aspect, an embodiment of the present invention provides an image processing method, including:

In at least two original images, the gray value of each original pixel in each original image is obtained; each original pixel is a pixel included in each original image;

Correcting each original image under the action of a preset low-pass filtering model based on the standard deviation of the gray value of each azimuth in each original image to obtain at least two first corrected images;

Obtain the average gray value of each azimuth in each first corrected image, and based on the average gray value of each azimuth, under the action of the preset low-pass filtering model Correcting the image is performed to obtain at least two target images;

The at least two target images are spliced into a composite image to complete the image processing process.

In the above solution, based on the standard deviation of the gray value of each azimuth in each original image, each original image is corrected under the action of a preset low-pass filter model to obtain at least two thirds. A corrected image, including:

In each original image, the preset low-pass filtering model is used to filter the standard deviation of the gray value of each azimuth to obtain an estimated value of the standard deviation of each azimuth; The gray value standard deviation of the azimuth direction corresponds to the estimated value of the standard deviation of each azimuth direction one-to-one;

Taking the ratio of the standard deviation of the gray value of each azimuth to the estimated value of the standard deviation of the azimuth as the gray value gain of each azimuth;

Correcting each original image according to the gray value gain in each azimuth direction, thereby obtaining the at least two first corrected images.

In the above solution, the correction is performed on each original image according to the gray value gain of each azimuth, so as to obtain the at least two first corrected images, including:

dividing the gray value of each original pixel by the gray value gain of each azimuth to obtain the first corrected gray value corresponding to each original pixel;

Correcting each original pixel according to the first corrected grayscale value to obtain each first corrected pixel, thereby obtaining a first corrected image corresponding to each original image, and further obtaining the at least two The first corrected image.

In the above solution, the average gray value of each azimuth in each first corrected image is obtained, and based on the average gray value of each azimuth, under the action of the preset low-pass filtering model, Each of the first corrected images is corrected to obtain at least two target images, including:

Using the preset low-pass filtering model, filter the mean value of the gray value of each azimuth in each first corrected image to obtain an estimated value of the mean value of each azimuth; The gray value mean value corresponds to the mean value estimated value of each azimuth direction one-to-one;

Taking the difference between the mean value of the gray value of each azimuth and the estimated value of the mean value of the azimuth as the gray value deviation of each azimuth;

Correcting each first corrected image according to the gray value deviation in each azimuth direction, thereby obtaining the at least two target images.

In the above solution, the correction is performed on each of the first corrected images according to the gray value deviation of each azimuth, so as to obtain the at least two target images, including:

acquiring the grayscale value of each first corrected pixel in each of the first corrected images;

The gray value of each first corrected pixel is correspondingly subtracted from the gray value deviation of each azimuth to obtain the target gray value corresponding to each first corrected pixel;

According to the target gray value of each first corrected pixel, the first corrected pixel of each first corrected image is corrected to obtain each target pixel, thereby obtaining the target image corresponding to each first corrected image , and then obtain the at least two target images.

In the above solution, the average gray value of each azimuth in each first corrected image is obtained, and based on the average gray value of each azimuth, under the action of the preset low-pass filtering model, After each first corrected image is corrected to obtain at least two target images, the method further includes:

In each target image, according to the gray value of each target pixel, obtain the average value of the gray value of each distance direction of each target image;

For each target image, filter correction is performed on each target image under the action of a Gaussian filter model based on the average gray value of each distance direction, to obtain a second correction of each target image. image, and then obtain at least two second corrected images;

The at least two second corrected images are spliced into a final composite image to complete the image processing process.

In the above solution, for each target image, based on the average gray value of each distance direction, filter and correct each target image under the action of a Gaussian filter model, and obtain the each target image. A second corrected image of the target image, thereby obtaining at least two second corrected images, including:

In each target image, calculating the mean value of the grayscale values of all target pixels as the overall grayscale value of each icon image;

For each target image, correspondingly calculate the ratio of the overall average gray value of each target image to the average gray value of each distance direction in the target image, and obtain the ratio of the average gray value of each distance direction. ;

Through the Gaussian filter model, filtering the gray value mean ratio of each distance direction to obtain the compensation factor of each distance direction;

In each original image, multiply the corresponding gray value of each target pixel by the compensation factor of each distance direction to obtain the corrected gray value of each target pixel;

Each second corrected pixel is obtained according to the corrected grayscale value of each target pixel, thereby obtaining a second corrected image of each target image, and further obtaining at least two second corrected images.

In the above scheme, the splicing of the at least two target images into a composite image to complete the image processing process includes:

obtaining the geolocation result of each target image;

According to the geolocation result of each target image, in the at least two target images, an overlapping area between every two adjacent target images is calculated to obtain a first overlapping area and a second overlapping area; the The first overlapping area is the area in the previous target image of the every two adjacent target images; the second overlapping area is the area in the next target image of the every two adjacent target images;

Taking the previous target image in each of the two adjacent target images as a standard image, based on the first overlapping area and the second overlapping area, under the action of a color-leveling filter model and a Gaussian convolution operation, the The last target image of each two adjacent target images is adjusted to obtain a final adjusted image corresponding to the last target image;

For every two adjacent target images in the at least two target images, the standard image in the every two adjacent target images and the final adjustment image are continuously image stitched until the last two adjacent target images are processed After completion, the composite image is obtained, and the image processing process is completed.

In the above scheme, the previous target image in each of the two adjacent target images is used as a standard image, and based on the standard image and the overlapping area, under the action of a color uniform filter model and a Gaussian filter algorithm, Adjust the last target image of each two adjacent target images to obtain the final adjustment image corresponding to the last target image, including:

respectively calculating the mean value of the first gray value and the standard deviation of the first gray value of the first overlapping area;

respectively calculating the second gray value mean and the second gray value standard deviation of the second overlapping area;

Based on the first gray value mean, the first gray value standard deviation, the second gray value mean, and the second gray value standard deviation, under the action of the uniform color filter model, the The last target image is subjected to uniform brightness processing to obtain an adjustment image corresponding to the last target image;

calculating the mean value of the gray value of each azimuth in the first overlapping area to obtain a third series of mean value of gray value;

According to the adjustment image, calculate the average value of gray values in each azimuth direction of the overlapping area of the adjustment image and the standard image, and obtain a fourth sequence of average values of gray values;

Through the Gaussian convolution operation, the adjusted image is adjusted according to the ratio of the third gray value mean sequence to the fourth gray value mean sequence to obtain the final adjusted image.

In the above solution, the adjusted image is adjusted according to the ratio of the third average gray value to the first average gray value through Gaussian convolution operation to obtain the final adjusted image, including:

performing a Gaussian convolution operation on the ratio of the third average gray value to the average value of the first gray value with a Gaussian kernel of a preset length to obtain an operation result;

The operation result is multiplied by the gray value of each adjusted pixel in the adjusted image, so as to obtain a final adjusted image; each adjusted pixel is each pixel included in the adjusted image.

In a second aspect, an embodiment of the present invention provides an image processing device, including: an acquisition unit, a correction unit, and a splicing unit; wherein,

The acquiring unit is configured to acquire, in at least two original images, the grayscale value of each original pixel in each original image; each original pixel is a pixel included in each original image;

The correction unit is configured to correct each original image under the action of a preset low-pass filtering model based on the standard deviation of the gray value of each azimuth in each original image, to obtain at least two a first corrected image; obtain the mean value of the gray value of each azimuth in each first corrected image, and based on the mean value of the gray value of each azimuth, under the action of the preset low-pass filter model, Correcting each of the first corrected images to obtain at least two target images;

The splicing unit is used for splicing the at least two target images into a composite image to complete the image processing process.

In a third aspect, an embodiment of the present invention provides an electronic device, including:

memory for storing executable data instructions;

The processor is configured to implement the above-mentioned image processing method when executing the executable instructions stored in the memory.

In a fourth aspect, an embodiment of the present invention provides a storage medium storing executable instructions for implementing the above-mentioned image processing method when a processor is executed.

Embodiments of the present invention provide an image processing method and device, electronic equipment, and storage medium, including: in at least two original images, acquiring the grayscale value of each original pixel in each original image; each original pixel is Pixels contained in each original image; based on the standard deviation of the gray value of each azimuth in each original image, each original image is corrected under the action of a preset low-pass filtering model, and at least two first Correct the image; obtain the average value of the gray value of each azimuth in each first corrected image, and based on the average value of the gray value of each azimuth, perform a pre-set low-pass filtering model on each first corrected image. Correction to obtain at least two target images; splicing the at least two target images into a composite image to complete the image processing process. Using the method provided by the embodiment of the present invention, the image processing apparatus can filter the original image twice and correct the gray value twice based on the gray value of the original image and the first corrected image through a low-pass filter, so as to eliminate the The azimuthal scalloped fringes in the original image improve the image quality of radar imaging.

Description of drawings

Fig. 1 is the working mechanism diagram of the spaceborne ScanSAR imaging system shown in the embodiment of the present invention;

FIG. 2 is an optional schematic flowchart of an image processing method provided by an embodiment of the present invention;

3 is an optional schematic flowchart of an image processing method provided by an embodiment of the present invention;

4 is an optional schematic flowchart of an image processing method provided by an embodiment of the present invention;

Figure 5(a) is a schematic diagram of a ScanSAR image without scallop correction;

FIG. 5(b) is a schematic diagram of an image after scallop effect correction is performed on FIG. 5(a) by using the method in the embodiment of the present invention;

6 is an optional schematic flowchart of an image processing method provided by an embodiment of the present invention;

Figure 7(a) is a schematic diagram of a ScanSAR image that has not been corrected for band inhomogeneity;

FIG. 7(b) is a schematic diagram of an image after correcting the uneven banding phenomenon in FIG. 7(a) by using the method in the embodiment of the present invention;

FIG. 8 is an optional schematic flowchart of an image processing method provided by an embodiment of the present invention;

Figure 9(a) is a schematic diagram of the image stitching result without any processing;

Figure 9(b) is a schematic diagram of the effect of using the classic Wallis filter processing on Figure 9(a);

Fig. 9 (c) is the effect schematic diagram that adopts the method in the embodiment of the present invention to carry out image stitching;

10 is an optional schematic flowchart of an image processing method provided by an embodiment of the present invention;

11 is a schematic structural diagram of an image processing apparatus provided by an embodiment of the present invention;

FIG. 12 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

An image processing method provided by an embodiment of the present invention is applicable to the spaceborne ScanSAR imaging system shown in FIG. 1 . FIG. 1 shows the working mechanism of spaceborne ScanSAR. The satellite 100 carrying the synthetic aperture radar sensor runs along the satellite orbit 120 at a At the height of H, ScanSAR changes the antenna elevation angle within a synthetic aperture time, performs beam switching, and performs multiple scans along the range direction to obtain echo data of multiple subbands 110_1, 110_2...110_n, and then converts the echo data on each subband. The echo data is imaged and stitched to obtain a super wide synthetic aperture image. Since the ScanSAR image is slant range imaging, in this embodiment of the present invention, the direction along the track is called the azimuth direction; the direction along which the radar wave is emitted is called the range direction.

FIG. 2 is an optional schematic flowchart of a method provided by an embodiment of the present invention, which will be described in conjunction with the steps shown in FIG. 2 .

S101. In at least two original images, obtain a grayscale value of each original pixel in each original image; each original pixel is a pixel included in each original image.

In the embodiment of the present invention, the image processing apparatus uses the image obtained by imaging the echo data on each subband of the radar as an original image, thereby acquiring at least two original images.

In the embodiment of the present invention, when the radar is performing beam switching, the ground target can only receive the illumination of part of the directional pattern. Therefore, after the echo data received by the radar is imaged, there will be bright and dark patterns in the azimuth direction of the image. Fine stripes, the scallop effect. The scallop effect has the characteristics of verticality, periodicity and gradual change. The stripes are basically perpendicular to the azimuth direction, and the degree of light and shade changes periodically along the azimuth direction. The error in the middle is large, and the two sides gradually decrease. In addition, the severity of the scallop effect is also related to the imaging algorithm used. Therefore, the image processing device needs to first eliminate the scallop effect in the at least two original images before stitching the at least two original images.

In this embodiment of the present invention, the image processing apparatus may acquire, in each of the at least two original images, the grayscale value of each original pixel in each original image, which is used for subsequent grayscale value processing, wherein. Raw pixels refer to the pixels in the original image.

In the embodiment of the present invention, the gray value represents the color depth and brightness of each pixel in the original image.

S102 , correcting each original image under the action of a preset low-pass filtering model based on the standard deviation of the gray value of each azimuth in each original image to obtain at least two first corrected images.

In this embodiment of the present invention, for each original image, the image processing apparatus may take the standard deviation of the gray value of the original image in each azimuth as an observation value sequence, perform filtering through a preset low-pass filter model, and then filter according to each azimuth. The original image is subjected to gray value correction based on the filtering result of the standard deviation of the gray value in the azimuth direction to obtain a first corrected image corresponding to the original image. The image processing device processes the at least two original images using the same method to obtain at least two first corrected images.

In the embodiment of the present invention, there are obvious differences in the contrast of the stripes due to the scallop effect. Assuming that the stripes of the scallop effect are caused by multiplicative factors, the multiplicative factor relationship model of the gray value between the original image and the scallop effect corrected image can be shown in formula (1), as follows:

S _c (i,j) = g _c (i, j) · S _m (i, j) (1)

In formula (1), i, j are the position coordinates of each pixel in the range direction and azimuth direction in an original image with scallop effect, and S _c (i, j) is each original pixel in the original image. The gray value of point (i, j), g _c (i, j) is the gray value gain existing in the gray value S _c (i, j) of each original pixel point (i, j), S _m (i, j) is the gray value of each original pixel point (i, j) after the gray value S _c (i, j) is corrected by the gray value gain g _c (i, j), that is, the original image The grayscale value of each first corrected pixel in the corresponding first corrected image.

In the embodiment of the present invention, it can be known from formula (1) that the multiplicative factor in the scallop effect can be corrected by the gray value gain g _c (i,j). In a statistical sense, the standard deviation of the gray value of an image can characterize the contrast of an image. Therefore, the image processing device can firstly use the standard deviation of the gray value in each azimuth in the original image as the feature on the position of the azimuth, and then use the standard deviation of the gray value in each azimuth as the observation value, Set a low-pass filter model, such as a first-order Kalman filter for statistical estimation, to obtain the optimal estimate of the standard deviation of the gray value in each azimuth, and then based on the optimal estimate of the standard deviation of the gray value in each azimuth , the gray value gain g _c (i,j) for correction is obtained.

In the embodiment of the present invention, each azimuth direction in the original image corresponds to each pixel column one-to-one. For each original image, the image processing apparatus firstly calculates the standard deviation of the gray value of each column of original pixels according to the gray value of each original pixel, so as to obtain the standard deviation of the gray value of each azimuth.

In some embodiments, the original image S can be represented by an M×N pixel matrix, as shown in formula (2):

In formula (2), M is the number of pixel rows of the original image S, and N is the number of pixel columns of the original image S, then for the original image represented by formula (2), the image processing device can obtain the grayscale of each column of original pixels. value standard deviation, so as to obtain 1×N gray value standard deviations as the gray value standard deviation of each azimuth in the original image S.

In the embodiment of the present invention, the preset low-pass filtering model is used to process the input sequence of observation values containing noise, to make assumptions about the statistical properties of the sequence of observation values, and to output an estimate corresponding to each observation value in the sequence of observation values. value as the filtering result of the preset low-pass filtering model. When the input sequence of observations is the standard deviation of gray values in each azimuth in the original image, the preset low-pass filtering model can output an estimated value corresponding to the standard deviation of gray values in each azimuth. In this way, the image processing apparatus can calculate the adjustable gray value gain in each azimuth based on the gray value standard deviation in each azimuth and the estimated value of the gray value standard deviation in each azimuth.

In some embodiments, the preset low-pass model may be a first-order Kalman filter, or may be a Gaussian filter, an average filter, a median filter, etc. with the same function, which is not limited in this embodiment of the present invention.

In this embodiment of the present invention, after obtaining the gray value gain g _c (i,j), the image processing apparatus may use the gray value gain g _c (i, j for the gray value of each original pixel according to formula (1) ) to perform correction, and obtain a first corrected image according to the corrected grayscale value.

S103: Acquire the average gray value of each azimuth in each first corrected image, and correct each first corrected image under the action of a preset low-pass filtering model based on the average gray value of each azimuth , get at least two target images.

In the embodiment of the present invention, each first corrected image is an image in which the image processing apparatus performs a multiplicative factor correction on the scallop effect in each original image, thereby reducing the contrast difference between the scallop stripes and adjacent areas. After the image processing device obtains each first corrected image, the image processing device may use the preset low-pass model again to filter each first corrected image based on the average gray value of each azimuth direction, and then filter the first corrected image according to the second As a result of the secondary filtering, each first corrected image is corrected again, and finally at least two target images without scalloping effect are obtained.

In the embodiment of the present invention, the image processing apparatus uses the pixels included in the target image as the target pixels.

In this embodiment of the present invention, in each first corrected image, the image processing device calculates the average value of the grayscale values of the first corrected pixels in each column by column, as the grayscale value of each azimuth in each first corrected image mean.

In this embodiment of the present invention, the first corrected pixels are pixels included in each of the first corrected images.

In the embodiment of the present invention, because the stripes in the scallop effect have differences in contrast, the original image usually also exhibits a feature that the brightness of the stripes is inconsistent with that of adjacent areas. Therefore, assuming that the stripes of the scallop effect are also caused by additive factors, the relationship model of the additive factors can be expressed as formula (3), as follows:

S _m (i,j)=S ₀ (i,j)+o _m (i,j) (3)

In formula (2), S _m (i, j) is the gray value of each first corrected pixel in the first corrected image, and o _m (i, j) is the gray value S _m of each first corrected pixel The gray value deviation existing in (i, j), S ₀ (i, j) is the gray value S ₀ (i, j) of each target pixel in the target without the scallop effect.

In the embodiment of the present invention, it can be known from formula (3) that the additive factor in the scallop effect can be corrected by the gray value deviation o _m (i,j). In a statistical sense, the mean value of the gray value of the image can characterize the brightness of the image. Therefore, the image processing device can first use the mean value of the gray value in each azimuth in the original image as the observation value, and perform statistical estimation through a preset low-pass filter model, such as a first-order Kalman filter, to obtain the gray value of each azimuth. Then, based on the optimal estimated value of the standard deviation of the gray value in each azimuth direction, the gray value deviation o _m (i, j) for correction is obtained.

It should be noted that, in the embodiment of the present invention, both g _c (i,j) and om ( _i , j) are only functions of azimuth positions, and remain unchanged for positions with the same azimuth but different distances.

In this embodiment of the present invention, after obtaining the gray value deviation o _m (i,j), the image processing apparatus may determine the gray value S of each first corrected image in each first corrected image according to formula (3). _m (i, j) uses the gray value deviation o _m (i, j) to perform secondary correction, and finally obtains a target image without scallop effect corresponding to each first corrected image.

It should be noted that, in this embodiment of the present invention, the two first-order Kalman filters used for gray value standard deviation and mean filtering may be set to the same parameters for filtering, or different filtering model parameters may be used. This embodiment of the present invention is not limited.

In some embodiments, the image processing apparatus can construct a first-order Kalman filter as a preset low-pass filtering model by formulas (4)-(8), which are used for the filtering process in S102 and S103, formulas (4)- (8) as follows:

P(k|k-1)=P(k-1|k-1)+Q (5)

P(k|k)=[1-Kg(k)]P(k|k-1) (8)

表示从k-1步到k步的状态一步预测值，

表示上一步最小均方误差估计值，P(k|k-1)与P(k-1|k-1)分别是

与

对应的协方差，Q表示过程噪声方差。In the embodiment of the present invention, the Kalman filter is an algorithm for optimally estimating the system state by using the linear system state equation and inputting and outputting observation data of the system. Since the observation data includes the influence of noise and interference in the system, the optimal estimation can also be regarded as a filtering process. In formula (4) and formula (5),

represents the one-step prediction value of the state from k-1 steps to k steps,

Represents the estimated value of the minimum mean square error in the previous step, P(k|k-1) and P(k-1|k-1) are respectively

and

The corresponding covariance, Q represents the process noise variance.

为第k步估计值，即为滤波输出，P(k|k)为第k步估计值对应的协方差。In the embodiment of the present invention, formula (6) is an update equation, and in formulas (6)-(8), Kg represents the Kalman gain, that is, the weight, which represents the ratio of the observed value to the estimated value in formula (7). R is the observed noise variance, Z is the observed value,

is the estimated value of the k-th step, that is, the filtering output, and P(k|k) is the covariance corresponding to the estimated value of the k-th step.

In this embodiment of the present invention, Kalman filters with different parameter settings can be obtained by adjusting the values of Q and R. In some embodiments, Q=1×10 ⁻⁶ , R=0.01, and the initial value of P updated in each iteration is set to 0.01, so as to obtain the corresponding first-order Kalman filter as the preset low-pass filter model , filter each original image or each first corrected image.

S104, splicing at least two target images into a composite image to complete the image processing process.

In the embodiment of the present invention, the at least two target images are images from which the scallop effect has been eliminated. In this way, after the image processing device obtains the at least two target images, each target image can be image-spliced to finally obtain a composite image. In the image processing process, the scallop effect has been removed from the obtained composite image, which improves the image quality of the final composite image.

It can be understood that, in the embodiment of the present invention, the image processing apparatus can filter the original image twice and correct the gray value twice based on the average gray value and the standard deviation of the gray value through a low-pass filter, so as to obtain the desired value. The stripes of the scallop effect in the azimuth direction in the original image are eliminated, and the image quality of the echo data imaging is improved; and the image processing device corrects the original image by filtering twice, which reduces the calculation amount of each filtering operation and improves the image processing. speed.

In the embodiment of the present invention, based on FIG. 2, based on the standard deviation of the gray value of each azimuth in each original image in S102, each original image is corrected under the action of a preset low-pass filtering model, and at least two A first corrected image, as shown in FIG. 3, includes S1021-S1023, as follows:

S1021. In each original image, use a preset low-pass filter model to filter the standard deviation of the gray value of each azimuth to obtain an estimated value of the standard deviation of each azimuth; the gray value of each azimuth The standard deviation corresponds one-to-one with the standard deviation estimate for each bearing.

In the embodiment of the present invention, the image processing apparatus uses a preset low-pass filter model, inputs the standard deviation of the gray value of each azimuth as an observation value sequence and inputs the preset low-pass filter model, and uses the preset low-pass filter model to analyze each The standard deviation of the gray value in the azimuth is estimated by filtering, and the estimation result of each step is retained, and the estimated value of the standard deviation of the gray value in each azimuth can be obtained as the filtering result.

In the embodiment of the present invention, the standard deviation of the gray value of each azimuth is in a one-to-one correspondence with the estimated value of the standard deviation of each azimuth.

In some embodiments, for the 1×N standard deviations D of gray values obtained by formula (2), the image processing apparatus can obtain the corresponding 1×N standard deviations D of gray values by using a preset low-pass filtering model. 1×N standard deviation estimates

S1022 , taking the ratio of the standard deviation of the gray value of each azimuth to the estimated value of the standard deviation of the azimuth as the gray value gain of each azimuth.

In the embodiment of the present invention, after obtaining the estimated value of the standard deviation of each azimuth, the image processing device calculates the ratio of the standard deviation of the gray value of each azimuth to the estimated value of the standard deviation corresponding to the azimuth, as each azimuth The gray value gain of , as shown in formula (9):

为第j列原始像素对应的标准差估计值，g_c(j)为第j列原始像素对应的灰度值增益。图像处理装置可以通过公式(9)，得到该原始图像中每个方位向的灰度值增益。In formula (9), D(j) is the grayscale standard deviation of the original pixel in the jth column of an original image, j is the number of columns of the original pixel,

is the estimated value of the standard deviation corresponding to the original pixel in the jth column, and g _c (j) is the gray value gain corresponding to the original pixel in the jth column. The image processing apparatus can obtain the gray value gain of each azimuth direction in the original image by formula (9).

S1023 , correcting each original image according to the gray value gain of each azimuth, so as to obtain at least two first corrected images.

In the embodiment of the present invention, after obtaining the gray value gain of each azimuth direction, the image processing device can correct each original image according to the gray value gain of each azimuth direction. The contrast difference caused by the gray value gain is removed from the value, and the first corrected image corresponding to each original image is obtained. The image processing device performs the same processing on at least two original images to obtain at least two first corrected images.

In some embodiments of the present invention, S1023 may specifically include S201-S202, as follows:

S201. Divide the gray value of each original pixel by the corresponding gray value gain of each azimuth to obtain a first corrected gray value corresponding to each original pixel.

In the embodiment of the present invention, based on formula (1), it can be known that the scallop effect caused by the multiplicative factor can be eliminated by removing the corresponding gray value gain from the gray value of the original pixel, as shown in formula (10), as follows:

In formula (10), for the original pixel point in the distance direction position i and the azimuth direction position j in the original image, the image processing device combines the gray value S(i, j) of the original pixel point with the pixel where the original pixel point is located. The gray value gain g _c (j) of the column is used as a quotient to obtain the first corrected gray value S _m (i, j) after correcting the gray value S(i, j).

S202 , correcting each original pixel according to the first corrected grayscale value to obtain each first corrected pixel, thereby obtaining a first corrected image corresponding to each original image, and further obtaining at least two first corrected images.

In this embodiment of the present invention, for each original pixel in an original image, the image processing apparatus may replace the original gray value of the original pixel with the first corrected gray value of the original pixel, and convert each gray The original pixel whose value has been replaced is used as each first corrected pixel, so that the overall gray value of the original image is updated to obtain a first corrected image corresponding to the original image.

In this embodiment of the present invention, for at least two original images, the image processing apparatus performs processing in the same manner to obtain at least two first corrected images.

In the embodiment of the present invention, based on FIG. 2 or FIG. 3 , in S103, the average value of the gray value of each azimuth in each first corrected image is obtained, and based on the average value of the gray value of each azimuth, the preset low-pass filter Under the action of the model, each first corrected image is corrected to obtain at least two target images, and the obtained at least two target images can be specifically shown in Figure 4, including S1031-S1033, as follows:

S1031. Using a preset low-pass filter model, filter the mean value of the gray value of each azimuth in each first corrected image to obtain an estimated value of the mean value of each azimuth; the mean value of the gray value of each azimuth direction One-to-one correspondence with the mean estimate for each azimuth.

In the embodiment of the present invention, the image processing device inputs the average gray value of each azimuth as an observation value sequence into a preset low-pass filter model, and filters the average gray value of each azimuth through the preset low-pass filter model Estimate, and retain the filtering result of each step, the gray value estimate of each azimuth can be obtained correspondingly, as the mean value estimate of each azimuth. Wherein, the estimated value of the gray value of each azimuth is an estimated value obtained by the image processing device by filtering and estimating the mean value of the gray value of each azimuth through a preset low-pass filtering model.

S1032 , taking the difference between the mean value of the gray value of each azimuth and the estimated value of the mean value of the azimuth as the gray value deviation of each azimuth.

In this embodiment of the present invention, the image processing apparatus may use the difference between the mean value of the gray value of each azimuth and the estimated value of the mean value of the azimuth as the gray value deviation of each azimuth according to formula (11), as follows :

第j列第一校正像素的均值估计值，即第j个方位向的均值估计值。In formula (11), o _m (j) represents the gray value deviation of the first correction pixel in the jth column, that is, the gray value deviation in the jth azimuth direction; M(j) represents the first correction pixel in the jth column. The mean value of the gray value, that is, the mean value of the gray value in the j-th azimuth;

The mean estimated value of the first corrected pixel in the jth column, that is, the mean estimated value of the jth azimuth.

S1033. Correct each first corrected image according to the gray value deviation of each azimuth, so as to obtain at least two target images.

In the embodiment of the present invention, the image processing apparatus can correct the difference in the stripe brightness of the scallop effect by using the gray value deviation in each azimuth direction, which may specifically include S301-S302, as follows:

S301. Acquire the grayscale value of each first corrected pixel in each first corrected image.

S302 , the gray value of each first corrected pixel is correspondingly subtracted from the gray value deviation of each azimuth to obtain a target gray value corresponding to each first corrected pixel.

In each first corrected image, the gray value of each first corrected pixel is correspondingly subtracted from the gray value deviation of the azimuth direction where the first corrected pixel is located to obtain the target gray level corresponding to each corrected pixel value; wherein, the azimuth direction where the first corrected pixel is located is the pixel column where the first corrected pixel is located.

In this embodiment of the present invention, based on formula (3), formula (12) can be obtained, and the image processing apparatus can use formula (12) to obtain, for each first corrected image, each first corrected pixel in the first corrected image The gray value of the first corrected pixel is subtracted from the gray value deviation corresponding to the first corrected pixel to obtain the target gray value of each first corrected pixel, as follows:

In formula (12), S ₀ (i,j) represents the gray value of the first corrected pixel at position i in the distance direction and the position j in the azimuth direction after the gray value deviation correction,

S303. Correct the first corrected pixels of each first corrected image according to the target gray value of each first corrected pixel to obtain each target pixel, thereby obtaining a target image corresponding to each first corrected image, and then Obtain at least two target images.

In the embodiment of the present invention, after obtaining the target gray value of each first corrected pixel in a first corrected image, the image processing apparatus may use the target gray value of each first corrected pixel to replace the original first corrected pixel to obtain the target image corresponding to the first corrected image. The image processing apparatus uses the same method for at least two first corrected images to obtain at least two target images.

It can be understood that, in this embodiment of the present invention, the image processing apparatus may first perform filtering and correction on the original image based on the standard deviation of the gray value of the original image by using a preset low-pass filtering model to obtain a first corrected image, which weakens the original image. The contrast difference of the scallop effect stripes in the middle is filtered and corrected based on the average gray value of the first corrected image, which reduces the brightness difference of the scallop effect stripes in the original image, thereby improving the image quality; further, due to the preset low-pass The filtering model only needs to filter one parameter of the standard deviation of the gray value or the mean value of the gray value in each filtering, and the calculation amount is small, thereby improving the speed of image processing.

In some embodiments, when the preset filtering model is the Kalman filter, the ScanSAR image without scallop correction can be shown in Fig. 5(a). In the image in Fig. 5(a), it can be clearly observed The scallop stripes that periodically change along the azimuth direction lose the details of the image and interfere with the normal interpretation of the image. From the graph in Figure 5(a), it can be seen that the mean and standard deviation of the pixel gray value fluctuate greatly, and the gray The value distribution is less continuous. Using the method in the embodiment of the present invention to process Fig. 5(a), the image after scallop effect correction as shown in Fig. 5(b) can be obtained. There is no obvious scallop stripe in the image of Fig. 5(b). And recover some details covered by the scallop effect in Figure 5(a), eliminate the jagged fluctuations of the mean and standard deviation of the azimuth position in Figure 5(a) from the graph in Figure 5(a), and correct the image Periodic changes in brightness and contrast.

In the embodiment of the present invention, based on any one of the methods shown in FIG. 2 to FIG. 4 , S103 obtains the average value of the gray value of each azimuth in each first corrected image, and based on the average value of the gray value of each azimuth, After correcting each first corrected image under the action of the preset low-pass filtering model to obtain at least two target images, as shown in FIG. 6 , including S401-S403, as follows:

S401. In each target image, according to the gray value of each target pixel, obtain the average value of the gray value of each distance direction of each target image.

In the embodiment of the present invention, during the ScanSAR imaging process, since the range radiation gain is mainly affected by the slant range and the range antenna pattern modulation, there is a large difference between the range patterns of different beams. Large uniform bands appear in the distance direction of the image.

In the embodiment of the present application, the non-uniformity of the strips does not have periodicity, but is vertical, and the non-uniform strips are almost perpendicular to the distance direction. The unevenness of the band is gradual, and the trend is smooth along the distance direction. Since the non-uniformity of the stripe is independent of the azimuth scallop effect, the image processing device can also use the method in S102-S103 to correct the scallop effect, and can also correct the stripe after ScanSAR imaging. Correction for uneven linearity.

In the embodiment of the present invention, since the unevenness of the stripes is caused by the difference between the distance direction patterns of different beams, the most important thing to correct the unevenness of the stripes is the accurate estimation of the distance direction patterns, Knowing the pattern of each image, the distance direction pattern compensation can be performed on the image, and the scattered radiation distribution of the image can be corrected evenly by multiplying the compensation function with the image distance direction.

In the embodiment of the present invention, for the target image S ₀ (M×N), the image processing apparatus may perform calculation on a row-by-row basis, and obtain the mean value M _r (M×1) of the gray value of the target pixels in each row of the target image, as the The mean gray value of each distance direction in the target image.

In the embodiment of the present invention, the mean value of the gray value of each distance direction represents the brightness distribution of the target image in the distance direction.

S402. For each target image, filter and correct each target image under the action of a Gaussian filter model based on the mean value of the gray value of each distance direction to obtain a second corrected image of each target image, and then obtain at least two a second corrected image.

In the embodiment of the present invention, for each target image, the image processing apparatus may substitute each range direction in the target image into a Gaussian filter model, so as to perform smooth filtering on the brightness distribution of the target image in the range direction through the Gaussian filter model, A second corrected image of the target image is obtained. The image processing apparatus adopts the same method for each target image, so as to obtain the second corrected image of each target image.

In this embodiment of the present invention, S402 may specifically include S4021-S4025, as follows:

S4021. In each target image, calculate the mean value of the grayscale values of all target pixels as the overall grayscale value of each icon image.

In this embodiment of the present invention, for each target image, the image processing apparatus may calculate the average gray value of all target pixels in the target image as the overall average gray value of the target image.

S4022. For each target image, correspondingly calculate the ratio of the overall average gray value of each target image to the average gray value of each distance direction in the target image, and obtain the average gray value ratio of each distance direction.

In the embodiment of the present invention, the image processing device calculates the overall average gray value of each target image with the average gray value of each distance direction in the target image to obtain the ratio of the average gray value of each distance direction. .

S4023 , filtering the mean value ratio of gray values of each distance direction through a Gaussian filter model to obtain a compensation factor for each distance direction.

In the embodiment of the present invention, the image processing device uses a Gaussian filter to filter the mean ratio of the gray value of each distance direction by using the formula (13), so as to obtain the compensation of brightness compensation for each row of target pixels in the target image. factor, as follows:

为卷积运算，g表示长度为L，标准差为σ的离散高斯核，如公式(14)所示：In formula (13), F(n) represents the compensation factor,

is a convolution operation, and g represents a discrete Gaussian kernel with length L and standard deviation σ, as shown in formula (14):

In some embodiments, L may be 800, σ may be 200, and may also be determined according to different image conditions, and the specific selection is made according to the actual situation, which is not limited in this embodiment of the present invention.

S4024. In each original image, multiply the corresponding gray value of each target pixel by the compensation factor of each distance direction to obtain the corrected gray value of each target pixel.

In the embodiment of the present invention, after the image processing device obtains the compensation factor F, it can compensate the grayscale value of each target pixel in the target image by the compensation factor to obtain the corrected grayscale value of each target pixel, as shown in formula (15 ) as shown:

S ₁ (i,j)=F(i)·S(i,j) (i∈[1,M],j∈[1,N]) (15)

In formula (15), S(i,j) is the gray value of the target pixel in the ith row and jth column of the target image, and F(i) is the compensation factor corresponding to each target pixel in the ith row. Image processing The device multiplies S(i,j) and F(i) to obtain the corrected grayscale value S ₁ (i,j) corresponding to the target pixel.

S4025 , obtaining each second corrected pixel according to the corrected grayscale value of each target pixel, thereby obtaining a second corrected image of each target image, and further obtaining at least two second corrected images.

In the embodiment of the present invention, for each target image, after obtaining the corrected grayscale value of each target pixel, the image processing device can replace the original grayscale value of each target pixel with the corrected grayscale value of each target pixel, A second corrected image corresponding to each target image is obtained, and the process of compensating and correcting the unevenness of the banding of the target image is completed.

In this embodiment of the present invention, the image processing apparatus processes the at least two target images using the same method, thereby obtaining at least two second corrected images.

S403 , splicing at least two second corrected images into a final composite image to complete the image processing process.

In this embodiment of the present invention, the image processing apparatus may stitch at least two second corrected images into a final composite image to complete the image processing process.

In this embodiment of the present invention, the image processing apparatus splices at least two second corrected images to obtain a final synthesized image in the same principle as S104, and details are not repeated here.

It can be understood that, in the embodiment of the present invention, the image processing device can further eliminate the uneven striping phenomenon based on the original target image from which the scallop effect has been eliminated, and use a Gaussian filter to filter the average gray value in the distance direction. Correction to get the target image with better image quality. After that, the image processing device can perform image stitching based on the target image from which the scallop effect and the uneven stripe phenomenon are eliminated, thereby further improving the image imaging quality.

In some embodiments, as shown in Fig. 7(a), the ScanSAR image without the correction of band inhomogeneity is shown. It can be seen from the image in Fig. 7(a) that the distance radiates strongly to the upper middle and the image On the bright side, while the distance radiates weakly towards the bottom, the image is on the dark side. It can be observed from the distance-to-brightness distribution curve graph in Figure 7(a) that a higher peak appears in the middle and upper parts of the distance, and decreases towards both sides. The phenomenon of uneven banding affects the overall perception of the image, and it is easy to give wrong radiation intensity distribution, which affects the interpretation and interpretation of the image, and brings great difficulties to the post-processing of the image. Fig. 7(b) shows the image after correcting the unevenness of the stripes by using the method in the embodiment of the present invention. As shown in the distance-direction luminance distribution curve in Fig. 7(b), the distance-direction luminance distribution is uniform, As shown in the image in Fig. 7(b), the large-scale bright band in the image disappears, the upper and lower transitions are uniform, and the uneven banding phenomenon is effectively eliminated.

In the embodiment of the present invention, based on FIGS. 2 to 6 , in S104, at least two second corrected images are spliced into a final composite image to complete the image processing process. Specifically, as shown in FIG. 8 , including S1041 to S1044, as follows:

S1041. Obtain the geographic location result of each target image.

In the embodiment of the present invention, the image processing apparatus may obtain the geographic positioning result of each target image from the satellite parameters corresponding to the target image.

S1042, according to the geolocation result of each target image, in at least two target images, calculate the overlapping area between every two adjacent target images, and obtain a first overlapping area and a second overlapping area; the first overlapping area is the area in the previous target image of every two adjacent target images; the second overlapping area is the area in the latter target image of every two adjacent target images.

In the embodiment of the present invention, due to the influence of various factors such as the slant range of the satellite from the target, the radar angle of view, the antenna pattern, and the atmospheric attenuation, the radiation intensity of the images formed by different sub-swaths will be different. When stitching a single image, there will be obvious seams at the stitching, and the image brightness on both sides of the stitching will be uneven. All the above-mentioned uneven radiation phenomena, if not properly and effectively corrected, will seriously affect the visual effect of the image, interfere with image interpretation, and hinder post-processing such as feature extraction and image stitching.

Therefore, in the embodiment of the present invention, after obtaining the target image, the image processing apparatus can calculate the overlapping area between two adjacent images according to the image geolocation result, and eliminate the splicing seam based on the overlapping area.

In the embodiment of the present invention, in every two adjacent target images, the image processing device takes the geographic location overlapping area of the previous target image and the next target image as the first overlapping area; The target image geolocation overlaps the region as the second overlap region.

S1043, taking the previous target image in every two adjacent target images as a standard image, and based on the first overlapping area and the second overlapping area, under the action of the color leveling filter model and the Gaussian convolution operation, for each two phase The next target image adjacent to the target image is adjusted to obtain the final adjusted image corresponding to the latter target image.

In the embodiment of the present invention, after obtaining the overlapping area of each two adjacent target images, the image processing apparatus may, based on the grayscale values of the first overlapping area and the second overlapping area between each two adjacent target images, through The color filter model adjusts the last target image in every two adjacent target images, and takes the adjusted last target image as the adjustment image.

In the embodiment of the present invention, since the actual splicing process is directional, the image processing apparatus takes the previous target image as a standard image in adjacent target images, and adjusts the latter target image. Exemplarily, if the sequence of image stitching is from left to right, then the left image in the adjacent target images is used as the standard image, and the right image is adjusted, and so on.

In some embodiments of the present invention, the uniform color filter model may be a Wallis filter, and the Wallis filter can map the grayscale mean and variance of a local image in an image to a given grayscale mean and variance, thereby Make the grayscale variance and grayscale mean value at different positions of the image have approximately equal values, and finally increase the contrast in the area with small image contrast, reduce the contrast in the area with large image contrast, and make the small grayscale in the image. Information is enhanced. Other uniform color and uniform light filtering models may also be selected according to different actual filtering needs, which is not limited in the embodiment of the present invention.

In the embodiment of the present invention, if the target image is only corrected by the color-leveling filter model once, the correction may fail when the brightness trend of the overlapping area is exactly opposite to the brightness trend of the target image. Therefore, in the embodiment of the present invention, the adjustment image will be adjusted The Gaussian convolution operation is performed again to make up for the limitation of the classic Wallis filter that fails when the brightness trend of the overlapping area is exactly opposite, and the final adjusted image is obtained.

In some embodiments of the present invention, S1043 may specifically include S501-S506, as follows:

S501. Calculate the mean value of the first gray value and the standard deviation of the first gray value of the first overlapping area respectively.

In the embodiment of the present invention, for the first overlapping area in each two adjacent target images, the image processing apparatus calculates the average value of the gray values of all the pixels in the first overlapping area as the first gray value average value; image processing The apparatus calculates the standard deviation of the grayscale values of all the pixels in the first overlapping area as the standard deviation of the first grayscale value.

S502, respectively calculating the second gray value mean and the second gray value standard deviation of the second overlapping area.

In the embodiment of the present invention, for the second overlapping area of every two adjacent target images, the image processing device calculates the average value of the gray values of all pixels in the second overlapping area as the second gray value average value; the image processing device The standard deviation of the grayscale values of all the pixels in the second overlapping area is calculated as the second standard deviation of the grayscale values.

S503. Based on the mean value of the first gray value, the standard deviation of the first gray value, the mean value of the second gray value and the standard deviation of the second gray value, under the action of the uniform color filter model, perform uniform brightness on the next target image After processing, the adjusted image corresponding to the latter target image is obtained.

In the embodiment of the present invention, the algorithm corresponding to the uniform color filtering model is shown in formula (16). The mean value of the degree value and the standard deviation of the second gray value are processed to uniformize the brightness of each target pixel in the latter target image, and the adjusted gray value of each target pixel in the latter target image is obtained, as follows:

In formula (16), S _β is the next target image in every two adjacent target images, s _α is the standard deviation of the first gray value, m _α is the mean value of the first gray value, and s _β is the second gray value. The standard deviation of the degree value, m _β is the mean value of the second gray value, and S' _β is the adjusted image after S _β is processed by the uniform color filter model.

S504: Calculate the mean value of the gray value in each azimuth direction in the first overlapping area to obtain a third series of mean value of gray value.

In the embodiment of the present invention, the image processing apparatus recalculates the average value of gray values in each azimuth direction in the first overlapping area in the standard image, and obtains a third sequence of average values of gray values.

S505 , according to the adjusted image, calculate the average value of gray values in each azimuth direction of the overlapping area between the adjusted image and the standard image, and obtain a fourth sequence of average values of gray values.

In this embodiment of the present invention, since the second overlapping area is an area in the next target image, after the image processing apparatus adjusts the latter target image to the adjusted image, the image processing apparatus may adjust the overlapping area between the image and the standard image based on the calculation , and recalculate the average gray value of each azimuth in the overlapping area to obtain a third gray value mean sequence.

S506 , adjusting the adjusted image according to the ratio of the third gray value mean sequence to the fourth gray value mean sequence through Gaussian convolution operation to obtain a final adjusted image.

In the embodiment of the present invention, after obtaining the third average gray value, the image processing device may perform Gaussian convolution operation processing on the adjusted image according to the ratio of the third average gray value to the average gray value of the standard image to obtain the final adjusted image .

In the embodiment of the present invention, the Gaussian convolution operation formula may be as shown in formula (17), as follows:

需要与高斯核g(n)卷积运算进行低通滤波，从而得到经过高斯卷积运算处理后的最终调整图像S”_β。In formula (17), S' _β is the adjusted image filtered by the color-leveling filter model in every two adjacent target images. M _α is the third gray value mean sequence, and M _β is the fourth gray value mean sequence. In order to avoid the effect of jitter, the gray value mean ratio

The convolution operation with the Gaussian kernel g(n) is required to perform low-pass filtering, so as to obtain the final adjusted image S" _β after the Gaussian convolution operation.

S1044, for every two adjacent target images in the at least two target images, continuously stitch the standard image and the final adjusted image in each two adjacent target images, until the last two adjacent target images are processed, A composite image is obtained, and the image processing process is completed.

In the embodiment of the present invention, the image processing device performs image splicing between the standard image in each two adjacent target images and the final adjusted image in at least two target images, so as to eliminate image splicing seams, and the image processing device performs image splicing on the at least two target images. After all two adjacent target images included in the target images are processed, a composite image is obtained, and the image processing process is completed.

It can be understood that in the embodiment of the present invention, after completing a Wallis filter, an average ratio is added to supplement the description of the brightness trend of the overlapping area, and make up for the limitation of the failure of the classic Wallis filter when the brightness trend of the overlapping area is exactly opposite. better image stitching effect and improved image quality.

In some embodiments, the image stitching result without any processing is shown in Fig. 9(a). In the image of Fig. 9(a), due to the difference in radiation intensity of the two images before stitching, the upper half of the image is brighter , the lower part is dim, so the seam is very obvious, and a step fault appears on its azimuthal brightness distribution curve. Figure 9(b) shows the effect of filtering Figure 9(a) through classical Wallis. From the azimuthal brightness distribution curve, it has been uniformly distributed, but the seams in the image still exist, and the brightness of the upper half of the image From bright to dark, the lower half of the brightness changes from dark to bright, the brightness trend is just the opposite, but the mean and standard deviation are almost the same, which is the limitation of the classic Wallis filter. Figure 9(c) shows the effect of eliminating image splicing seams by using the method in the embodiment of the present invention. The upper and lower images are properly connected and excessively stable, the azimuthal brightness distribution curve is uniform, and the splicing seams are effectively eliminated.

The embodiment of the present invention provides an image processing method, which can be applied to my country's first C-band multi-polarization high-resolution synthetic aperture radar satellite, the only civilian microwave remote sensing imaging satellite in the "National High-resolution Earth Observation System Major Project" The Gaofen-3 (GF-3) satellite has a variety of imaging modes, including a scanning mode. The high-quality ScanSAR data of the GF-3 satellite can be used for scientific research experiments, which is of great significance for verifying the radiation correction algorithm of on-board ScanSAR images.

In the embodiment of the present invention, the ScanSAR image captured by the GF-3 satellite in 2016 is selected for testing. The selected image is located in Chifeng City, Inner Mongolia Autonomous Region, my country, which is the interception zone between the southwestern section of the Daxingan Mountains and the northern end of the Qilaotu Mountains. The scattering in this area is relatively uniform, and it is easy to observe the radiation inhomogeneity from the image, which is suitable for the radiation correction of the spaceborne ScanSAR image. In the embodiment of the present invention, the methods in S601-S609 can be used to correct and stitch the spaceborne ScanSAR images, as follows:

S601. In at least two original images, obtain the gray value of each original pixel in each original image.

S602. Correct each original image under the action of a preset low-pass filtering model based on the standard deviation of the gray value of each azimuth in each original image to obtain at least two first corrected images.

S603: Acquire the mean value of the gray value of each azimuth in each first corrected image, and correct each first corrected image under the action of a preset low-pass filter model based on the mean value of the gray value of each azimuth , get at least two target images.

S604. In each target image, according to the gray value of each target pixel, obtain the average value of the gray value of each distance direction of each target image.

S605. For each target image, filter and correct each target image under the action of the Gaussian filter model based on the average value of the gray value of each distance direction to obtain a second corrected image of each target image, and then obtain at least two a second corrected image.

S606. Obtain the geolocation result of each second corrected image.

In this embodiment of the present invention, the principle of the method for obtaining the geolocation result of each second corrected image in S606 is the same as that in S1041, and details are not described herein again.

S607. According to the geolocation result of each second corrected image, in at least two second corrected images, calculate the overlapping area between every two adjacent second corrected images to obtain a third overlapping area and a fourth overlapping area area; the third overlapping area is the area in the previous target image of every two adjacent second corrected images; the fourth overlapping area is the area in the second corrected image after every two adjacent second corrected images area.

In this embodiment of the present invention, the principle of the method for obtaining the third overlapping region and the fourth overlapping region in S607 is the same as that for obtaining the first overlapping region and the second overlapping region in S1042, and details are not described herein again.

S608. Use the previous second corrected image in every two adjacent second corrected images as a standard image, and based on the third overlapping area and the fourth overlapping area, under the action of the color leveling filter model and the Gaussian convolution operation, to The last second corrected image of every two adjacent second corrected images is adjusted to obtain a final adjusted image corresponding to the last second corrected image.

In this embodiment of the present invention, the principle of the method for obtaining the final adjusted image corresponding to the next second corrected image in S608 is the same as that for obtaining the final adjusted image corresponding to the next target image in S1043 , and details are not repeated here.

S609. For every two adjacent second corrected images in the at least two second corrected images, perform continuous image stitching between the standard image and the final adjusted image in every two adjacent second corrected images, until the last two After the processing of the adjacent second corrected image is completed, a composite image is obtained, and the image processing process is completed.

In this embodiment of the present invention, the principle of the method for obtaining a composite image in S609 is the same as that in S1044, and details are not described herein again.

It can be understood that, in the embodiment of the present invention, the image processing device can sequentially perform image processing on at least two original images obtained by radar scanning to eliminate the scallop effect in the azimuth direction, correct the uneven phenomenon of the stripe in the distance direction, and eliminate the phenomenon of stitching. Finally, a composite image is obtained, which greatly improves the image quality of radar imaging.

An embodiment of the present invention provides an image processing apparatus 2. As shown in FIG. 11, the image processing apparatus 2 includes: an acquisition unit 200, a correction unit 201, and a splicing unit 202; wherein,

The acquiring unit 200 is configured to acquire, in at least two original images, the gray value of each original pixel in each original image; the each original pixel is a pixel included in each original image;

The correction unit 201 is configured to correct each original image under the action of a preset low-pass filtering model based on the standard deviation of the gray value of each azimuth in each original image, to obtain at least two a first corrected image; obtain the average gray value of each azimuth in each first corrected image, and based on the average gray value of each azimuth, under the action of the preset low-pass filtering model Correcting each of the first corrected images to obtain at least two target images;

The splicing unit 202 is used for splicing the at least two target images into a composite image to complete the image processing process.

In some embodiments of the present invention, the correction unit 201 is further configured to use the preset low-pass filtering model in each of the original images, for the standard deviation of the gray value of each azimuth direction Perform filtering to obtain the estimated value of the standard deviation of each azimuth; the standard deviation of the gray value of each azimuth corresponds to the estimated value of the standard deviation of each azimuth; The ratio of the standard deviation of the gray value to the estimated value of the standard deviation of the azimuth is used as the gray value gain of each azimuth direction; according to the gray value gain of each azimuth direction, each original image is corrected , so as to obtain the at least two first corrected images.

In some embodiments of the present invention, the correction unit 201 is further configured to divide the gray value of each original pixel by the corresponding gray value gain of each azimuth to obtain each original pixel The corresponding first corrected grayscale value; according to the first corrected grayscale value, each original pixel is corrected to obtain each first corrected pixel, thereby obtaining the first correction corresponding to each original image image, and then obtain the at least two first corrected images.

In some embodiments of the present invention, the correction unit 201 is further configured to use the preset low-pass filtering model to filter the mean value of the gray value of each azimuth in each first corrected image , to obtain the mean estimated value of each azimuth; the mean gray value of each azimuth corresponds to the mean estimated value of each azimuth; the average gray value of each azimuth is equal to The difference between the mean estimated values of the azimuth is used as the gray value deviation of each azimuth; using the preset low-pass filtering model, the gray level of each azimuth in each first corrected image is The mean value of each azimuth is filtered to obtain the mean value estimate value of each azimuth direction; the gray value mean value of each azimuth direction corresponds to the mean value estimate value of each azimuth direction one-to-one; The difference between the mean value of the degree value and the estimated value of the mean value of the azimuth direction is used as the gray value deviation of each azimuth direction; according to the gray value deviation of each azimuth direction, each first corrected image is corrected , so as to obtain the at least two target images, and correct each of the first corrected images according to the gray value deviation of each azimuth, so as to obtain the at least two target images.

In some embodiments of the present invention, the correction unit 201 is further configured to acquire the grayscale value of each first corrected pixel in each of the first corrected images; The value corresponding to the gray value deviation of each azimuth direction is subtracted to obtain the target gray value corresponding to each first correction pixel; according to the target gray value of each first correction pixel, for each first correction pixel A first corrected pixel of a corrected image is corrected to obtain each target pixel, thereby obtaining a target image corresponding to each of the first corrected images, and further obtaining the at least two target images.

In some embodiments of the present invention, the image processing apparatus 2 further includes a compensation correction unit, wherein,

The compensation and correction unit is used to obtain the mean value of the gray value of each azimuth in each first corrected image, and based on the mean value of the gray value of each azimuth, the function of the preset low-pass filter model After correcting each first corrected image to obtain at least two target images, in each target image, each distance of each target image is obtained according to the gray value of each target pixel The mean value of the gray value of the distance direction; for each target image, based on the mean value of the gray value of each distance direction, filter correction is performed on each target image under the action of the Gaussian filter model to obtain the A second corrected image of the target image is obtained, and at least two second corrected images are obtained; the at least two second corrected images are spliced into a final composite image to complete the image processing process.

In some embodiments of the present invention, the compensation and correction unit is further configured to, in each of the target images, calculate the average value of the grayscale values of all target pixels as the overall grayscale value of each icon image For each of the target images, correspondingly calculate the ratio of the mean value of the overall gray value of each target image to the mean value of the gray value of each distance direction in the target image, and obtain the mean value of the gray value of each distance direction ratio; through the Gaussian filter model, filter the gray value mean ratio of each distance direction to obtain the compensation factor of each distance direction; in each original image, each The gray value of the target pixel is correspondingly multiplied by the compensation factor of each distance direction to obtain the corrected gray value of each target pixel; each second corrected pixel is obtained according to the corrected gray value of each target pixel , thereby obtaining a second corrected image of each target image, and further obtaining at least two second corrected images.

In some embodiments of the present invention, the splicing unit 202 is further configured to obtain a geographic location result of each target image; according to the geographic location result of each target image, the at least two target images , calculate the overlapping area between every two adjacent target images, and obtain a first overlapping area and a second overlapping area; the first overlapping area is the previous target image in each two adjacent target images The second overlapping area is the area in the next target image of the two adjacent target images; the previous target image in the two adjacent target images is used as the standard image, Based on the first overlapping area and the second overlapping area, under the action of a color-leveling filter model and a Gaussian convolution operation, the last target image of each two adjacent target images is adjusted to obtain the The final adjustment image corresponding to the latter target image; for every two adjacent target images in the at least two target images, the standard image in the every two adjacent target images and the final adjustment image are continuously imaged , until the last two adjacent target images are processed, the composite image is obtained, and the image processing process is completed.

In some embodiments of the present invention, the splicing unit 202 is further configured to calculate the first gray value mean and the first gray value standard deviation of the first overlapping area respectively; Two gray value mean and second gray value standard deviation; based on the first gray value mean, the first gray value standard deviation, the second gray value mean and the second gray value standard deviation, under the action of the color equalization filter model, the brightness of the last target image is uniformly processed to obtain the adjustment image corresponding to the last target image; the grayscale of each azimuth in the first overlapping area is calculated. The average value of the intensity values is obtained to obtain a third sequence of average values of gray values; according to the adjusted image, the average value of gray values in each azimuth direction of the overlapping area between the adjusted image and the standard image is calculated to obtain a fourth sequence of average values of gray values ; Through Gaussian convolution operation, according to the ratio of the third gray value mean sequence and the fourth gray value mean sequence, the adjustment image is adjusted to obtain the final adjustment image.

In some embodiments of the present invention, the splicing unit 202 is further configured to perform Gaussian convolution with a Gaussian kernel of preset length between the ratio of the third average gray value to the first average gray value operation to obtain an operation result; multiply the operation result by the gray value of each adjustment pixel in the adjustment image to obtain the final adjustment image; the adjustment pixel is each adjustment pixel included in the adjustment image. pixel.

An embodiment of the present invention provides an electronic device 5. As shown in FIG. 12, the electronic device 5 includes: a processor 54, a memory 55, and a communication bus 56, and the memory 55 communicates with the processing device through the communication bus 56. communicate with the processor 54, the memory 55 stores one or more programs executable by the processor 54, and when the one or more programs are executed, the processor 54 executes any one of the above. image processing method.

Embodiments of the present disclosure provide a computer-readable storage medium, where one or more programs are stored in the computer-readable storage medium, and the one or more programs can be executed by one or more processors 54 to achieve the following: The image processing method described in any one of the above.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media having computer-usable program code embodied therein, including but not limited to disk storage, optical storage, and the like.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention.

Claims (13)

1. An image processing method, comprising:

acquiring a gray value of each original pixel in each original image from at least two original images; each original pixel is a pixel contained in each original image;

correcting each original image under the action of a preset low-pass filtering model based on the gray value standard difference of each azimuth direction in each original image to obtain at least two first corrected images;

acquiring a mean value of gray values of each azimuth direction in each first correction image, and correcting each first correction image under the action of the preset low-pass filtering model based on the mean value of gray values of each azimuth direction to obtain at least two target images;

and splicing the at least two target images into a composite image to finish the image processing process.

2. The method according to claim 1, wherein the correcting each original image under the action of a preset low-pass filtering model based on the standard deviation of the gray-scale value of each azimuth direction in each original image to obtain at least two first corrected images comprises:

filtering the gray value standard deviation of each azimuth direction in each original image by using the preset low-pass filtering model to obtain a standard deviation estimation value of each azimuth direction; the gray value standard deviation of each azimuth direction corresponds to the standard deviation estimation value of each azimuth direction one by one;

taking the ratio of the gray value standard deviation of each azimuth direction to the standard deviation estimation value of the azimuth direction as the gray value gain of each azimuth direction;

and correcting each original image according to the gray value gain of each azimuth direction so as to obtain at least two first corrected images.

3. The method of claim 2, wherein said correcting each of said original images according to said gray value gain for each orientation to obtain said at least two first corrected images comprises:

correspondingly dividing the gray value of each original pixel with the gray value gain of each azimuth direction to obtain a first correction gray value corresponding to each original pixel;

and correcting each original pixel according to the first correction gray value to obtain each first correction pixel, so as to obtain a first correction image corresponding to each original image, and further obtain the at least two first correction images.

4. The method according to any one of claims 1 to 3, wherein the obtaining of the mean value of the gray-scale values of each orientation in each first corrected image, and based on the mean value of the gray-scale values of each orientation, correcting each first corrected image under the action of the preset low-pass filtering model to obtain at least two target images comprises:

filtering the gray value mean value of each azimuth direction in each first correction image by using the preset low-pass filtering model to obtain a mean value estimation value of each azimuth direction; the gray value mean value of each azimuth direction corresponds to the mean value estimation value of each azimuth direction one by one;

taking the difference value of the gray value mean value of each azimuth direction and the mean value estimation value of the azimuth direction as the gray value deviation of each azimuth direction;

and correcting each first correction image according to the gray value deviation of each azimuth direction, thereby obtaining the at least two target images.

5. The method of claim 4, wherein said correcting each first corrected image according to the gray value deviation of each orientation to obtain the at least two target images comprises:

acquiring a gray value of each first correction pixel in each first correction image;

correspondingly subtracting the gray value of each first correction pixel from the gray value deviation of each azimuth direction to obtain a target gray value corresponding to each first correction pixel;

and correcting the first correction pixel of each first correction image according to the target gray value of each first correction pixel to obtain each target pixel, so as to obtain a target image corresponding to each first correction image, and further obtain the at least two target images.

6. The method according to any one of claims 1 to 5, wherein after obtaining a mean value of gray-level values of each azimuth direction in each first corrected image, and correcting each first corrected image under the action of the preset low-pass filtering model based on the mean value of gray-level values of each azimuth direction to obtain at least two target images, the method further comprises:

in each target image, obtaining a mean value of gray values of each target image in each distance direction according to the gray value of each target pixel;

for each target image, based on the gray value mean value of each distance direction, performing filtering correction on each target image under the action of a Gaussian filtering model to obtain a second corrected image of each target image, and further obtain at least two second corrected images;

and splicing the at least two second correction images into a final composite image to finish the image processing process.

7. The method according to claim 6, wherein the performing, for each target image, filter correction on each target image under the effect of a gaussian filter model based on the mean of the grayscale values of each distance direction to obtain a second corrected image of each target image, and further obtain at least two second corrected images, comprises:

calculating the mean value of the gray values of all target pixels in each target image to serve as the integral gray value of each icon image;

for each target image, correspondingly calculating the ratio of the overall gray value mean value of each target image to the gray value mean value of each distance direction in the target image to obtain the gray value mean value ratio of each distance direction;

filtering the gray value average ratio of each distance direction through the Gaussian filtering model to obtain a compensation factor of each distance direction;

in each original image, multiplying the gray value of each target pixel by the compensation factor of each distance direction correspondingly to obtain the corrected gray value of each target pixel;

and obtaining each second correction pixel according to the correction gray value of each target pixel, thereby obtaining a second correction image of each target image and further obtaining at least two second correction images.

8. The method according to any one of claims 1 to 7, wherein said stitching said at least two target images into a composite image, performing an image processing procedure, comprises:

acquiring a geographical positioning result of each target image;

according to the geographic positioning result of each target image, calculating an overlapping area between each two adjacent target images in the at least two target images to obtain a first overlapping area and a second overlapping area; the first overlap region is a region in a previous target image of the every two adjacent target images; the second overlapping area is an area in a subsequent target image of the every two adjacent target images;

taking the previous target image of every two adjacent target images as a standard image, and adjusting the next target image of every two adjacent target images under the action of a uniform color filter model and Gaussian convolution operation on the basis of the first overlapping area and the second overlapping area to obtain a final adjusted image corresponding to the next target image;

and for every two adjacent target images in the at least two target images, continuously splicing the standard images in every two adjacent target images with the final adjustment image until the last two adjacent target images are processed, obtaining the composite image, and finishing the image processing process.

9. The method according to claim 8, wherein the step of adjusting a subsequent target image of each two adjacent target images based on the overlapping area of the standard image and the overlapping area under the action of a uniform color filter model and a gaussian filter algorithm to obtain a final adjusted image corresponding to the subsequent target image, which is performed by taking a previous target image of each two adjacent target images as a standard image, comprises:

respectively calculating a first gray value mean value and a first gray value standard deviation of the first overlapping area;

respectively calculating a second gray value mean value and a second gray value standard deviation of the second overlapped area;

performing brightness uniformity processing on the next target image under the action of a uniform color filter model based on the first gray value mean value, the first gray value standard deviation, the second gray value mean value and the second gray value standard deviation to obtain an adjusted image corresponding to the next target image;

calculating the gray value mean value of each azimuth direction in the first overlapping area to obtain a third gray value mean value sequence;

calculating the gray value mean value of each azimuth in the overlapping area of the adjusted image and the standard image according to the adjusted image to obtain a fourth gray value mean value sequence; and adjusting the adjusted image according to the ratio of the third gray value mean sequence to the fourth gray value mean sequence through Gaussian convolution operation to obtain the final adjusted image.

10. The method of claim 9, wherein the adjusting the adjusted image according to the ratio of the third gray-scale value mean to the first gray-scale value mean by gaussian convolution operation to obtain the final adjusted image comprises:

performing Gaussian convolution operation on the ratio of the third gray value mean value to the first gray value mean value and a Gaussian kernel with a preset length to obtain an operation result;

multiplying the operation result by the gray value of each adjusting pixel in the adjusting image to obtain a final adjusting image; the each adjustment pixel is each pixel included in the adjustment image.

11. An image processing apparatus characterized by comprising: the device comprises an acquisition unit, a correction unit and a splicing unit; wherein,

the acquiring unit is used for acquiring the gray value of each original pixel in each original image in at least two original images; each original pixel is a pixel contained in each original image;

the correction unit is used for correcting each original image under the action of a preset low-pass filtering model based on the gray value standard deviation of each azimuth direction in each original image to obtain at least two first corrected images; acquiring a mean value of gray values of each azimuth direction in each first correction image, and correcting each first correction image under the action of the preset low-pass filtering model based on the mean value of gray values of each azimuth direction to obtain at least two target images;

and the splicing unit is used for splicing the at least two target images into a composite image to finish the image processing process.

12. An electronic device, comprising:

a memory for storing executable data instructions;

a processor for implementing the method of any one of claims 1 to 10 when executing executable instructions stored in the memory.

13. A storage medium having stored thereon executable instructions for causing a processor to perform the method of any one of claims 1 to 10 when executed.

Images