A piecewise smoothing adaptive quantization method for SAR images

Abstract

The present invention relates to a SAR image segmented smoothing adaptive quantization method, comprising the following steps: (1) global adjustment of pixel values: adjusting the pixel values ​​of the original image matrix to a reasonable range; (2) obtaining a segment of pixels: sliding upward in distance and azimuth to obtain a pixel segment; (3) calculating a divisor factor: calculating the divisor factor of the segment of pixels; (4) truncation of the divisor factor: truncation of the divisor factor greater than a threshold; (5) smoothing the divisor factor: smoothing the divisor factor of three adjacent segments in azimuth; (6) quantization: dividing the divisor factor of each pixel obtained by the corresponding pixel value and rounding; (7) repeating steps (2)-(6) until the entire image is traversed. The method of the present invention can reduce the pixel value of the strong scattering area while maintaining or improving the pixel value of the weak scattering area, enhance the detail information of the entire image, and provide core technical support for the SAR imaging field.

Description

Technical Field

[0001] The invention relates to the field of signal and information processing, and in particular to a SAR image piecewise smoothing adaptive quantization method. Background Art

[0002] The purpose of SAR (Synthetic Aperture Radar) quantization is to quantize the floating point data output by the SAR imaging algorithm into 8-bit image data for subsequent interpretation by image judgment experts or interpretation algorithms. The dynamic distribution of SAR raw floating point data is extremely large. The floating point pixel values ​​of strong scattering areas such as buildings and walls and weak scattering areas such as grasslands and farmlands often differ by thousands to tens of thousands of times. The extremely high pixel value range makes quantization extremely difficult. Most traditional quantization methods use global quantization, which makes the texture details of the strong scattering area clearly visible in the image, while also reducing the pixel value of the weak scattering area, so that the weak scattering area appears as a dark area in the image, and the detail information is lost. Conversely, while making the texture details of the weak scattering area clearly visible in the image, the pixel value of the strong scattering area of ​​the image is increased, resulting in saturation, which appears as a white area in the image, and the detail information is lost. Therefore, it is a very difficult and valuable technology to quantize SAR floating point data into 8-bit image data while maintaining the texture structure of the strong scattering area and the weak scattering area to achieve high dynamic display of SAR images. Summary of the invention

[0003] Aiming at the problem that the existing SAR image quantization method cannot satisfy the problem of simultaneously maintaining detail information of weak scattering regions and strong scattering regions after quantization, the present invention provides a SAR image piecewise smoothing adaptive quantization method.

[0004] The specific content of the present invention is as follows: A SAR image piecewise smoothing adaptive quantization method comprises the following steps:

[0005] (1) Global pixel value adjustment: adjust the pixel values ​​of the original image matrix to a reasonable range;

[0006] (2) Get a segment of pixels: Slide upwards one unit at a time and get the pixel segment;

[0007] (3) Calculate the divisor factor: calculate the divisor factor of the segment of pixels;

[0008] (4) Divisor factor truncation: truncate the divisor factor that is greater than the threshold;

[0009] (5) Smoothing divisor factor: Smoothing is performed using the divisor factors of the three adjacent segments in the azimuth direction;

[0010] (6) Quantization processing: Divide the corresponding pixel value by the divisor factor of each pixel and round it up;

[0011] (7) Repeat steps (2)-(6) until the entire image is traversed.

[0012] Furthermore, in step (1), the average pixel value of the image matrix is ​​adjusted to λ by the following formula:

[0013]

[0014] Among them, m x Represents the mean of the image matrix X. The value of λ needs to be adjusted according to specific needs. The larger the λ, the brighter the quantized image will be overall, and vice versa.

[0015] Furthermore, step (2) is quantized by segmented processing, with distance being processed upward by distance units and azimuth being processed upward by sliding line segments.

[0016] Furthermore, in step (3), the divisor factor calculation formula is as follows:

[0017]

[0018]

[0019] in, is the pixel mean of the jth segment of the ith range gate, m x is the mean of the adjusted image matrix X, υ is the power exponential factor, which is used to adjust the decoupling degree of the divisor factors of the strong scattering area and the weak scattering area. is the proportional coefficient, which is the ratio of the pixel mean of the segment to the global mean of the image matrix. is the divisor factor of the line segment, α is the weight factor, and β is the bias factor.

[0020] Furthermore, in step (4), the divisor factor is truncated using the following formula:

[0021]

[0022] Among them, r m is the maximum scale factor.

[0023] Furthermore, the method for smoothing the divisor factor in step (5) is as follows:

[0024] The divisor factors of the j-1th, jth and j+1th segments of the i-th range gate are and The divisor factor for the first pixel of the jth segment is:

[0025]

[0026] The divisor factors for the L / 2th and L / 2+1th pixels of the jth segment are:

[0027]

[0028] Where L is the number of pixels in the pixel segment taken out,

[0029] The divisor factor for the Lth pixel in the jth segment is:

[0030]

[0031] The divisor factors are obtained through two linear interpolations:

[0032]

[0033] l∈{1,2,...,L}

[0034] When j=1, the j-1 segment is replaced by the j segment; when the j segment is the last segment of a range gate, the j+1 segment is replaced by the j segment.

[0035] Furthermore, in step (6), the quantization process is truncated with a threshold of 255, including the following steps:

[0036] Divide the pixel value by the divisor factor obtained for each pixel:

[0037]

[0038] Round the adjusted pixel values ​​to integers:

[0039]

[0040] Finally, pixel values ​​greater than 255 are truncated to obtain the final 8-bit unsigned integer pixel value:

[0041]

[0042] Compared with the existing SAR quantization method, the SAR image piecewise smoothing adaptive quantization method of the present invention has the following significant advantages: adopting the method of unit-by-unit in range direction and piecewise processing in azimuth direction, dividing the pixels in each segment by their own divisor factors, realizing piecewise adaptive quantization, and having a particularly good effect on SAR images with serious azimuth side lobes; the method realizes decoupled quantization processing of strong scattering areas and weak scattering areas, reduces the pixel value of the strong scattering area while maintaining or increasing the pixel value of the weak scattering area, and realizes clear details of the whole image; adopts linear smoothing processing of the azimuth divisor factor, so that the quantized SAR image looks more natural and smooth. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The specific implementation of the present invention will be further explained below in conjunction with the accompanying drawings.

[0044] Figure 1 It is a flowchart of the SAR image piecewise smoothing adaptive quantization method of the present invention;

[0045] Figure 2 Get pixel diagram for segmentation;

[0046] Figure 3 The figure is a comparison chart of quantitative results. DETAILED DESCRIPTION

[0047] Combination Figure 1 and Figure 2 The present invention provides a SAR image piecewise smoothing adaptive quantization method, which specifically comprises the following steps:

[0048] (10) Global adjustment of pixel values: Assume that the original matrix of the SAR image is Where m and n are the range gate and azimuth gate numbers of the SAR image, respectively. Adjust the average pixel value of the image matrix to λ,

[0049]

[0050] Among them, m x represents the mean value of the image matrix X. The value of λ needs to be adjusted according to specific needs. The larger the λ is, the brighter the quantized image will be. Conversely, the darker it will be. In this embodiment, λ is 40.

[0051] (20) Obtain a segment of pixels: the distance is upward by distance unit, and the azimuth is upward to obtain a line segment for quantization processing. The sliding step size and the length of the intercepted line segment are both L. In this embodiment, L is 512.

[0052] (30) Calculate the divisor factor: The divisor factor is the divisor of each pixel. The divisor factor of the strong scattering area is larger, which can suppress the side lobes and enhance the texture details; the divisor factor of the weak scattering area is smaller, which can increase the pixel value of the weak scattering area. In order to achieve the above purpose, the strong scattering area and the weak scattering area need to be decoupled and quantized. When suppressing the side lobes of the strong scattering area, the pixel value of the weak scattering area cannot be reduced; when increasing the pixel value of the weak scattering area, the pixel value of the strong scattering area cannot be increased.

[0053] In order to achieve the above quantization effect, the divisor factor is designed as follows:

[0054]

[0055]

[0056] in, is the pixel mean of the jth segment of the ith range gate, m xis the mean of the adjusted image matrix X, υ is the power exponential factor, which is used to adjust the decoupling degree of the divisor factors of the strong scattering area and the weak scattering area. is the proportional coefficient, which is the ratio of the pixel mean of the segment to the global mean of the image matrix. is the divisor factor of the line segment, α is the weight factor, and β is the bias factor.

[0057] When the extracted line segment is scattered more strongly, The value is large, usually above 5. In this case, the weight factor α has a significant effect on the divisor factor. plays a leading role, and the bias factor β has little effect on the divisor factor. When the extracted line segment scatters weakly, The value is small, usually below 0.3. At this time, the bias factor β has a significant effect on the divisor factor. The weight factor α plays a leading role, and the weight factor α has little effect on the divisor factor. Therefore, the side lobes of the strong scattering area can be suppressed and the brightness of the weak scattering area can be improved by appropriately increasing the value of the weight factor α and reducing the value of the bias factor β. In this embodiment, the weight factor α is 1.3, the bias factor β is 0.5, and the power exponent factor υ is 1.5.

[0058] (40) Divisor factor truncation: When the pixel values ​​in the strong scattering area are large overall, the proportional coefficient of the segment is If the scale factor is too large, the pixel value of the area will be too low after quantization, resulting in over-suppression. Do truncation processing, as described in the following formula,

[0059]

[0060] Among them, r m is the maximum proportionality coefficient, which is 20 in this embodiment.

[0061] (50) Smoothing divisor factor: When the divisor factors of two adjacent segments are very different, after quantization, there will be obvious discontinuity at the junction of the two segments, making the image look unnatural. Therefore, by performing a linear transformation on the divisor factor, the divisor factors between adjacent pixel segments are smoothed, so that the quantized SAR image is smooth and natural. The specific method is as follows:

[0062] The divisor factors of the j-1th, jth and j+1th segments of the i-th range gate are and The divisor factor of the first pixel of the jth segment is,

[0063]

[0064] The divisor factors for the L / 2th and L / 2+1th pixels of the jth segment are,

[0065]

[0066] Wherein, L is the number of pixels in the extracted pixel segment, which is 512 in this embodiment.

[0067] The divisor factor for the Lth pixel in the jth segment is,

[0068]

[0069] The divisor factors are obtained through two-segment linear interpolation.

[0070]

[0071] l∈{1,2,...,L}(9)

[0072] When j=1, the j-1 segment is replaced by the j segment; when the j segment is the last segment of a range gate, the j+1 segment is replaced by the j segment.

[0073] (60) Quantization processing: Divide the corresponding pixel value by the divisor factor obtained for each pixel.

[0074]

[0075] The adjusted pixel value is rounded to the nearest integer.

[0076]

[0077] Finally, pixel values ​​greater than 255 are truncated to obtain the final 8-bit unsigned integer pixel value.

[0078]

[0079] (70) Repeat steps (20)-(60) until the entire image is traversed.

[0080] like Figure 3 The figure shows the comparison of SAR image quantization results, where the three figures on the left are images processed by the traditional quantization method, and the three figures on the right are the corresponding images processed by the quantization method of the present application. It can be seen that after quantization by the present method, the pixel value texture structure of the strong scattering area and the weak heat dissipation area is maintained at the same time, and the image is more natural and smooth.

[0081] The SAR image piecewise smoothing adaptive quantization processing method of the present invention adopts the method of piecewise processing in distance and azimuth, and divides the pixels in each segment by their own divisor factors to realize piecewise quantization, which is particularly effective for SAR images with serious azimuth sidelobes. The divisor factor includes two adjustment factors, weight and bias. The weight factor is used to reduce the pixel value of the strong scattering area, which is not strongly associated with the weak scattering area, and the bias factor is used to increase the pixel value of the weak scattering area, which is not strongly associated with the strong scattering area. The divisor factor realizes the decoupling quantization processing of the strong scattering area and the weak scattering area, reduces the pixel value of the strong scattering area while maintaining or improving the pixel value of the weak scattering area, and realizes SAR adaptive quantization processing. The azimuth divisor factor is used for linear smoothing processing, so that the quantized SAR image looks more natural and smooth. This method can reduce the pixel value of the strong scattering area while maintaining or improving the pixel value of the weak scattering area, enhance the detail information of the whole image, and provide core technical support for the SAR imaging field.

[0082] Many specific details are described in the above description to facilitate a full understanding of the present invention. However, the above description is only a preferred embodiment of the present invention. The present invention can be implemented in many other ways different from those described herein, so the present invention is not limited to the specific implementation disclosed above. At the same time, any person familiar with the art can make many possible changes and modifications to the technical solution of the present invention using the methods and technical contents disclosed above without departing from the scope of the technical solution of the present invention, or modify it into an equivalent embodiment of equivalent changes. Any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of protection of the technical solution of the present invention.

Claims

1. A piecewise smoothing adaptive quantization method for SAR images. Features: The steps include: (1) Global pixel value adjustment: adjust the pixel values ​​of the original image matrix to a reasonable range; (2) Get a segment of pixels: Slide upwards one unit at a time and get the pixel segment; (3) Calculate the divisor factor: Calculate the divisor factor of the segment of pixels; (4) Divisor factor truncation: truncate the divisor factor that is greater than the threshold; (5) Smoothing divisor factor: Smoothing is performed using the divisor factors of the three adjacent segments in the azimuth. (6) Quantization processing: Divide the divisor factor of each pixel by the corresponding pixel value and round it up; (7) Repeat steps (2)-(6) until the entire image is traversed; In step (3), the divisor factor is calculated as follows: ， ， in, For the The distance gate The pixel mean of the segment, is the adjusted image matrix The mean of is a power exponential factor, which is used to adjust the decoupling degree of the divisor factors of the strong scattering area and the weak scattering area. is the proportional coefficient, which is the ratio of the pixel mean of the segment to the global mean of the image matrix. is the divisor factor of the segment, is the weight factor, is the bias factor; The method for smoothing the divisor factor in step (5) is as follows: Record The distance gate , and The segment divisors are , and , so that The divisor factor for the first pixel of the segment is: No. The first paragraph and The divisor factor for pixels is: in, is the number of pixels in the pixel segment taken out, No. The first paragraph The divisor factor for pixels is: The divisor factors are obtained through two linear interpolations: When At that time, Paragraph Paragraph replaced; when When the segment is the last segment of a range gate, Paragraph Segment instead.

2. The SAR image piecewise smoothing adaptive quantization method according to claim 1, Features: In step (1), the average pixel value of the image matrix is ​​adjusted to : ， in, Represents the image matrix The mean of The value of needs to be adjusted according to specific needs. The larger the value, the brighter the quantized image will be overall, and vice versa.

3. The SAR image piecewise smoothing adaptive quantization method according to claim 1, Features: Step (2) uses a segmented processing method to perform quantization, with distance units upward and azimuth sliding upward to take line segments for quantization.

4. The SAR image piecewise smoothing adaptive quantization method according to claim 1, Features: In step (4), the divisor factor is truncated using the following formula: ， in, is the maximum scale factor.

5. The SAR image piecewise smoothing adaptive quantization method according to claim 4, Features: In step (6), the quantization process is performed with a threshold of 255 and includes the following steps: Divide the pixel value by the divisor factor obtained for each pixel: Round the adjusted pixel values ​​to integers: Finally, pixel values ​​greater than 255 are truncated to obtain the final 8-bit unsigned integer pixel value: 。

Images