An infrared image multi-scale stripe noise removing method, system and storage medium

Abstract

The present application relates to the technical field of image processing, more particularly to a kind of infrared image multiscale stripe noise removal method, system and storage medium. Including the first infrared image is carried out N times downsampling, obtains the first infrared image;The first infrared image is carried out one-dimensional single side gray filtering 1*n in the first direction, obtains the low-frequency first infrared image, by filtering out low-frequency first infrared image, the stripe amplitude in the second direction of high-frequency first infrared image is obtained;High-frequency first infrared image second direction stripe amplitude is upsampled, and the stripe amplitude of the original width of first infrared image is obtained;The pixel point in the first direction of first infrared image is subtracted from the original width of the stripe amplitude of the corresponding position of first infrared image, and the first infrared image after removing vertical stripe is obtained. The quality of infrared image is effectively improved by the present application.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to a method, system, and storage medium for multi-scale stripe noise removal of infrared images. Background Technology

[0002] Infrared imaging electronics are a crucial component of thermal imaging systems. They are responsible for fully utilizing the performance of infrared focal plane detectors and processing the electrical signals output by these detectors into video signals or other system-defined signal formats. Infrared imaging electronics consist of two main parts: the hardware system and the image processing algorithm. Given the current maturity of hardware platforms, infrared image processing technology has become a key research area for imaging electronics. In infrared focal plane detectors, the output circuit typically shares the same amplifier circuit for pixels in the same column or row. However, the bias voltages output by different amplifier circuits are not entirely consistent, resulting in stripe noise in the infrared image along the column or row direction.

[0003] In existing technologies, there are two main methods for eliminating non-uniform noise: one is to filter out horizontal stripes based on notch filters. This method is very effective for periodic stripe noise and can remove fine stripe noise in a single column, but it cannot remove wide stripe noise. The other method is to use the correlation between adjacent rows of the image and estimate the noise based on the difference between the row mean of the adjacent rows and the row mean of the current row. This method is very effective when the row mean of the image itself does not change much. However, if the row mean of the image itself changes very much, it will introduce stripe noise.

[0004] To address the above problems, this invention proposes a method, system, and storage medium for multi-scale stripe noise removal in infrared images, thereby improving the quality of infrared images. Summary of the Invention

[0005] To better address the noise problem of wide stripes in infrared images, this invention provides a multi-scale stripe noise removal method for infrared images.

[0006] As a preferred technical solution of the present invention, the following steps are included:

[0007] Step S1: Downsample the first infrared image I by a factor of N to obtain the first infrared image I. down ;

[0008] Step S2: For the first infrared image I down A one-dimensional unilateral grayscale filter 1*n is performed in the first direction to obtain the low-frequency first infrared image I. low By filtering out the low-frequency first infrared image I low Acquire the high-frequency first infrared image I highThe fringe amplitude in the second direction, where n represents the width of the one-dimensional grayscale filter window, and the first direction and the second direction are perpendicular;

[0009] Step S3: Process the first infrared image I at high frequency. high The stripe amplitude in the second direction is upsampled to obtain the stripe amplitude of the original width of the first infrared image I;

[0010] Step S4: Subtract the original width of the stripe amplitude at the corresponding position in the first infrared image I from the pixel values ​​in the first direction of the first infrared image I to obtain the first infrared image I after removing the stripes in the second direction. dst .

[0011] As a preferred embodiment of the present invention, in step S1, the first infrared image I is obtained. down The expression is as follows:

[0012]

[0013] Where I represents the first infrared image, and x and y represent the coordinates of the first infrared image in two-dimensional space.

[0014] As a preferred embodiment of the present invention, in step S2, the low-frequency first infrared image I low The expression is as follows:

[0015]

[0016]

[0017] Where w() represents a one-dimensional Gaussian function, abs() represents calculating the absolute value, and t represents the index of the pixel within the filtering window. As a preferred embodiment of the present invention, the high-frequency first infrared image I... high The expression is as follows:

[0018] I high (x,y)=I down (x,y)-I low (x,y)

[0019] Among them, I down (x, y) represents the first infrared image I after downsampling. low I low (x,y) represents the first low-frequency infrared image, I high (x, y) represents the first high-frequency infrared image I. high .

[0020] As a preferred embodiment of the present invention, the high-frequency first infrared image I high The expression for the fringe amplitude in the second direction is as follows:

[0021]

[0022] Here, StrValue() represents the stripe amplitude function.

[0023] As a preferred embodiment of the present invention, in step S3, the expression for the stripe amplitude of the original width of the first infrared image I is as follows:

[0024] StrValue(x,1)=StrValue N (x / N,1).

[0025] As a preferred technical solution of the present invention, the first infrared image I after removing the second directional stripes... dst The expression is as follows:

[0026]

[0027] Among them, I dst The first infrared image I after removing the second directional stripes. dst .

[0028] The present invention also provides an infrared image multi-scale stripe noise removal system as described above, the system comprising the following modules:

[0029] The downsampling module is used to downsample the first infrared image I by a factor of N to obtain the first infrared image I. down ;

[0030] The stripe amplitude calculation module is used to calculate the first infrared image I. down Perform one-dimensional unilateral grayscale filtering 1*n in the first direction to obtain the low-frequency first infrared image I. low By filtering out low-frequency first infrared images I low Acquire the first high-frequency infrared image I high The fringe amplitude in the second direction, where n represents the width of the one-dimensional grayscale filter window, and the first direction and the second direction are perpendicular;

[0031] The upsampling module is used to upsample the high-frequency first infrared image I. high The stripe amplitude in the second direction is upsampled to obtain the stripe amplitude of the original width of the first infrared image I;

[0032] The stripe removal module is used to subtract the original width of the stripe amplitude at the corresponding position in the first infrared image I from the pixels in the first direction of the first infrared image I, thereby obtaining the first infrared image I after removing the stripes in the second direction. dst .

[0033] The present invention also provides a computing device, the device comprising:

[0034] Memory and processor;

[0035] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the above-described method for multi-scale stripe noise removal in infrared images is implemented.

[0036] The present invention also provides a storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method for multi-scale stripe noise removal in infrared images.

[0037] Compared with the prior art, the beneficial effects of the present invention are at least as follows:

[0038] 1. The technical solution of the present invention obtains the first infrared image I by downsampling the first infrared image I by N times. down ; For the first infrared image I down Perform one-dimensional unilateral grayscale filtering 1*n in the first direction to obtain the low-frequency first infrared image I. low By filtering out low-frequency first infrared images I low Acquire the first high-frequency infrared image I high Acquire the first high-frequency infrared image I high The fringe amplitude in the second direction; for the first infrared image I at high frequency high The fringe amplitude in the second direction is upsampled to obtain the fringe amplitude of the original width of the first infrared image I; the fringe amplitude at the corresponding position of each pixel in the first direction of the first infrared image I is subtracted to obtain the first infrared image I after removing the fringe in the second direction. dst It enables the removal of wide stripes at small scales and fine stripes at large scales in infrared images. It is also applicable to the removal of horizontal and vertical stripes in infrared images, effectively improving the quality of infrared images.

[0039] 2. The technical solution of the present invention transforms wide stripes into thin stripes by downsampling the infrared image, thereby achieving the purpose of removing vertical stripes from the infrared image. This improves the efficiency of processing infrared images. The technical solution has good real-time performance and is suitable for parallel processing and porting to FPGA. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating the steps of a multi-scale stripe noise removal method for infrared images according to the present invention.

[0041] Figure 2 This is the infrared image of the present invention without vertical stripes removed;

[0042] Figure 3 This is the infrared image after the vertical stripe removal process of this invention;

[0043] Figure 4 This is a structural diagram of an infrared image multi-scale stripe noise removal system according to the present invention;

[0044] Figure 4 As shown: 100, Infrared image multi-scale stripe noise removal system; 101, Downsampling module; 102, Stripe amplitude calculation module; 103, Upsampling module; 104, Stripe removal module. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0046] In infrared focal plane detectors, the output circuit typically shares the same amplifier circuit for pixels in the same column or row. However, the bias voltages output by different amplifier circuits are not entirely consistent, resulting in stripe noise in the infrared image along the column or row direction. As mentioned above, improving the quality of infrared images is the problem that this invention aims to solve.

[0047] To address the aforementioned technical problems, the present invention proposes the following... Figure 1 The method for multi-scale stripe noise removal in infrared images, as shown, includes the following steps:

[0048] Step S1: Downsample the first infrared image I by a factor of N to obtain the first infrared image I. down ;

[0049] Step S2: Process the first infrared image I down Perform one-dimensional unilateral grayscale filtering 1*n in the first direction to obtain the low-frequency first infrared image I. low By filtering out low-frequency first infrared images I low Acquire the first high-frequency infrared image I high The fringe amplitude in the second direction, where n represents the width of the one-dimensional grayscale filter window, and the first and second directions are perpendicular;

[0050] Step S3: Analyze the first high-frequency infrared image I highThe fringe amplitude in the second direction is upsampled to obtain the fringe amplitude of the original width of the first infrared image I;

[0051] Step S4: Subtract the original width of the stripe amplitude at the corresponding position in the first infrared image I from the pixels in the first direction of the first infrared image I to obtain the first infrared image I after removing the stripes in the second direction. dst .

[0052] Specifically, the infrared image is downsampled sequentially by a factor of N, N / 2, and N / 4. This downsampling transforms wide stripes into thinner stripes, removing wide stripes at a small scale and thin stripes at a large scale. It is also applicable to removing horizontal and vertical stripes, effectively improving the quality of the infrared image. Downsampling of the infrared image includes, but is not limited to, the following methods: 1) simultaneous downsampling in both width and height directions; 2) downsampling only in the width direction; 3) downsampling only in the height direction. Methods for obtaining low-resolution images include, but are not limited to, the following: arithmetic mean, weighted average, and row or column-interval sampling. The downsampling scale N includes, but is not limited to, 8, where N is a positive integer and must be a common divisor of W and H. Methods for calculating stripe amplitude through filtering include, but are not limited to, the following: spatial filtering, grayscale filtering, bilateral filtering, and guided filtering. These filtering methods are existing technologies and will not be elaborated upon here.

[0053] Depending on the width of the vertical stripes in the infrared image, the downsampling width can be flexibly selected. Here, the downsampling scale values ​​are 8, 4, and 2 respectively. This method is suitable for reducing the computational resources required for processing while preserving image features. A one-dimensional single-sided grayscale filter (1*n) is performed on the downsampled infrared image along the row direction to obtain a low-frequency infrared image. This step helps eliminate noise and non-stripe interference in the infrared image. Subtracting the low-frequency infrared image from the downsampled infrared image to obtain a high-frequency infrared image helps highlight the vertical stripe features in the infrared image. Calculating the mean value along the column direction of the high-frequency infrared image yields the stripe amplitude value in the corresponding column. This step helps quantify the amplitude of each vertical stripe. Upsampling the stripe amplitude values ​​along the column direction of the high-frequency infrared image yields the stripe amplitude value of the original width of the infrared image. This allows for the recovery of the vertical stripe features in the infrared image while maintaining the original resolution. Subtracting the stripe amplitude value of the original width of the infrared image from the original infrared image yields the de-striped infrared image, which effectively eliminates vertical stripe interference in the infrared image, making the image clearer and facilitating further analysis and processing. For example, different downsampling and upsampling factors can be selected based on the infrared image quality and computing resources, or appropriate filters and thresholds can be selected based on the analysis objectives.

[0054] Furthermore, in step S1 above, the first infrared image I is obtained. down The expression is as follows:

[0055]

[0056] Where I represents the first infrared image, and x and y represent the coordinates of the first infrared image in two-dimensional space.

[0057] Specifically, the infrared image is downsampled by a factor of N. First, the width W and height H of the infrared image, the downsampling factor N, and the coordinates (x, y) of the infrared image in two-dimensional space are determined. Then, the first downsampled infrared image I is obtained using the above expression. down It's important to note that downsampling will result in the loss of some information from the original infrared image. Therefore, it's necessary to weigh the relationship between the downsampling factor and the quality of the infrared image and choose an appropriate downsampling factor.

[0058] Furthermore, in step S2 above, the low-frequency first infrared image I low The expression is as follows:

[0059]

[0060]

[0061] Where w() represents a one-dimensional Gaussian function, abs() represents the calculation of absolute value, and t represents the index of the pixel within the filtering window.

[0062] Specifically, low-frequency infrared imaging refers to imaging techniques that utilize the low-frequency region of the infrared spectrum (typically 3-5 micrometers or 8-14 micrometers). This technique can be used to detect thermal radiation, as the temperature of objects typically affects their radiation intensity in the infrared spectrum. A bilateral filter is used to filter the original infrared image in the row direction to obtain the low-frequency information. In the above method, the bilateral filtering method allows high-frequency information to be extracted from the original infrared image while preserving effective edge information. The one-dimensional Gaussian function is: (σ is the standard deviation, x is the horizontal axis, w(x) is the calculation result, the larger σ is, the wider the curve fluctuation). The low-frequency infrared image can be obtained through the above expression. This step helps to eliminate noise and non-striped interference in the infrared image.

[0063] Furthermore, the high-frequency first infrared image I high The expression is as follows:

[0064] I high (x,y)=I down (x,y)-I low (x,y)

[0065] Among them, I down (x, y) represents the first infrared image I after downsampling. low I low (x,y) represents the first low-frequency infrared image, I high (x,y) represents the first high-frequency infrared image I. high .

[0066] Specifically, subtracting the low-frequency infrared image from the downsampled infrared image to obtain the high-frequency infrared image is a common image processing method. Extracting high-frequency information from an image is very useful for many applications, as high-frequency information typically represents details, textures, and other variations in an image. It is important to note that when subtracting two infrared images, their size and type must be identical. Additionally, if the low-frequency infrared image is obtained from the output of some filters, its negative value can be used to perform high-pass filtering during subtraction. This technique helps to highlight vertical stripe features in the infrared image.

[0067] Furthermore, the high-frequency first infrared image I high The expression for the fringe amplitude in the second direction is as follows:

[0068]

[0069] Here, StrValue() represents the stripe amplitude function.

[0070] Specifically, fringe amplitude refers to the difference between the maximum and minimum brightness values ​​of interference fringes in optical interference. In an interferometer, interference fringes are produced by the interference of two beams of light. One beam undergoes a phase change after passing through an optical path difference, and interference fringes appear when the two beams meet. The brightness of these interference fringes changes with the optical path difference, and the fringe amplitude is the difference between the maximum and minimum brightness values ​​of the interference fringes. In optical interference, fringe amplitude is an important parameter. It can be used to measure the phase difference between two beams of light, and the shape and surface roughness of a sample can be measured by measuring the amplitude changes of these interference fringes. For example, in surface topography measurement, an interferometer can be used to measure the shape of a sample surface, and the height difference of the sample surface can be calculated by measuring the amplitude changes of the interference fringes. Furthermore, fringe amplitude can also be used to measure parameters such as the surface quality and transparency of optical components, as these parameters affect the brightness and shape of the interference fringes. Therefore, in optical manufacturing and inspection, fringe amplitude is an important indicator used to evaluate the performance and quality of optical components.

[0071] Calculating the mean value along the column direction of a high-frequency infrared image yields the fringe amplitude along that column direction. In infrared image processing and analysis, this is used to remove or analyze texture information. The fringe amplitude is calculated and then plotted as a grayscale image for display. It can be seen that fringe amplitude is very suitable for describing the texture features of an image. In specific implementations, the calculation of fringe amplitude can be adjusted according to the actual situation; for example, mean filtering can be replaced with Gaussian filtering, or other statistics along the column direction can be calculated. Furthermore, fringe amplitude can also be used in target detection, image denoising, and other fields.

[0072] Furthermore, in step S3 above, the expression for the stripe amplitude of the original width of the first infrared image I is as follows:

[0073] StrValue(x,1)=StrValue N (x / N,1).

[0074] Specifically, upsampling the stripe amplitude along the column direction of a high-frequency infrared image involves inserting new elements between pixels using a suitable interpolation algorithm, essentially expanding a 3x3 structure to a 4x4 or 5x5 structure before performing the corresponding calculations. The specific implementation method uses the same upsampling and downsampling coefficients, both being N. This means inserting N-1 points between points k and k+1 in the original image, creating an N-segment. There are many interpolation methods, generally considered from both the time and frequency domains. For time-domain interpolation, linear interpolation is the simplest. Other methods, such as Hermite interpolation and spline interpolation, can be found in numerical analysis textbooks and directly substituted into the formulas for calculation. This step directly calculates the stripe amplitude of the original width of the infrared image using the aforementioned expression.

[0075] Furthermore, the first infrared image I after removing the second directional fringes dst The expression is as follows:

[0076]

[0077] Among them, I dst Indicates the first infrared image I after removing the second directional fringes. dst .

[0078] Specifically, the final infrared image can be obtained through the above technical solution. For example... Figure 3 As shown, the infrared image after vertical stripe removal is as follows: Figure 2 As shown in the image, the infrared image without vertical stripes is shown. By comparing these two images, it can be seen that the image processed by this invention has significantly removed the stripe information.

[0079] The present invention also provides, for example Figure 4The infrared image multi-scale stripe noise removal system 100 shown includes the following modules:

[0080] The downsampling module 101 is used to downsample the first infrared image I by a factor of N to obtain the first infrared image I. down ;

[0081] Stripe amplitude calculation module 102 is used to calculate the first infrared image I down Perform one-dimensional unilateral grayscale filtering 1*n in the first direction to obtain the low-frequency first infrared image I. low By filtering out low-frequency first infrared images I low Acquire the first high-frequency infrared image I high The fringe amplitude in the second direction, where n represents the width of the one-dimensional grayscale filter window, and the first and second directions are perpendicular;

[0082] Upsampling module 103 is used for upsampling the first high-frequency infrared image I. high The fringe amplitude in the second direction is upsampled to obtain the fringe amplitude of the original width of the first infrared image I;

[0083] The stripe removal module 104 is used to subtract the stripe amplitude value of the original width at the corresponding position in the first infrared image I from the pixels in the first direction of the first infrared image I, so as to obtain the first infrared image I after removing the stripes in the second direction. dst .

[0084] The present invention also provides a memory and a processor;

[0085] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the above-mentioned method for multi-scale stripe noise removal of infrared images is implemented.

[0086] The present invention also provides a computer storage medium storing program instructions, wherein, when the program instructions are executed, the device where the computer storage medium is located executes the above-described infrared image multi-scale stripe noise removal method.

[0087] In summary, the present invention obtains the first infrared image I by downsampling the first infrared image I by N times. down ; For the first infrared image I down Perform one-dimensional unilateral grayscale filtering 1*n in the first direction to obtain the low-frequency first infrared image I. low By filtering out low-frequency first infrared images I low Acquire the first high-frequency infrared image I high The fringe amplitude in the second direction; for the first infrared image I at high frequency highThe fringe amplitude in the second direction is upsampled to obtain the fringe amplitude of the original width of the first infrared image I; the fringe amplitude of the corresponding position in the first infrared image I is subtracted from the fringe amplitude of the corresponding position in the first infrared image I to obtain the first infrared image I after removing the fringe in the second direction. dst It achieves the removal of wide stripes at small scales and fine stripes at large scales in infrared images. It is also applicable to the removal of horizontal and vertical stripes in infrared images, effectively improving the quality of infrared images.

[0088] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0089] The above-described embodiments are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0090] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for multi-scale stripe noise removal in infrared images, characterized in that, Includes the following steps: Step S1: Process the first infrared image The first infrared image is obtained by downsampling by N times. The first infrared image The expression is as follows: ; in, Let x and y be the first infrared image, respectively. In two-dimensional space, W represents the coordinate position of the first infrared image. The width of the first infrared image is H. Height; Step S2: For the first infrared image A one-dimensional unilateral grayscale filter (1 * n) is performed in the first direction to obtain the low-frequency first infrared image. By filtering out the low-frequency first infrared image The high-frequency first infrared image is obtained in the second direction, where n represents the width of the one-dimensional grayscale filter window, and the first direction and the second direction are perpendicular. The first infrared image at low frequency The expression is as follows: ； ； Where w() represents a one-dimensional Gaussian function, abs() calculates the absolute value, and t represents the index of the pixel within the filtering window. The standard deviation; The first high-frequency infrared image The expression is as follows: ; in, This is the first infrared image after downsampling. The first infrared image is low frequency. , The first infrared image is of high frequency. ; Step S3: Process the first infrared image at high frequency. The fringe amplitude in the second direction is upsampled to obtain the first infrared image. The original width of the stripe amplitude; Step S4: The first infrared image The number of pixels in the first direction minus the first infrared image The first infrared image after removing the second direction stripes is obtained by taking the original width stripe amplitude at the corresponding position. ; The first infrared image after removing the second directional stripes The expression is as follows: ; in, The first infrared image after removing the second directional stripes. , This represents the stripe amplitude function.

2. The method for multi-scale stripe noise removal in infrared images according to claim 1, characterized in that, The first high-frequency infrared image The expression for the fringe amplitude in the second direction is as follows: .

3. The method for multi-scale stripe noise removal in infrared images according to claim 1, characterized in that, In step S3, the first infrared image The expression for the original width of the stripe amplitude is as follows: .

4. A multi-scale stripe noise reduction system for infrared images, used to implement the method as described in any one of claims 1-3, characterized in that, The system includes the following modules: The downsampling module is used to process the first infrared image. The first infrared image is obtained by downsampling by N times. ; The stripe amplitude calculation module is used to calculate the first infrared image. Perform one-dimensional unilateral grayscale filtering (1 * n) in the first direction to obtain the low-frequency first infrared image. By filtering out low-frequency first infrared images Acquire the first high-frequency infrared image The fringe amplitude in the second direction, where n represents the width of the one-dimensional grayscale filter window, and the first direction and the second direction are perpendicular; The upsampling module is used for the first infrared image at high frequencies. The fringe amplitude in the second direction is upsampled to obtain the first infrared image. The original width of the stripe amplitude; Destriping module for the first infrared image Pixels in the first direction minus the first infrared image The first infrared image after removing the second direction stripes is obtained by taking the original width stripe amplitude at the corresponding position. .

5. A computing device, characterized in that, The device includes: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the infrared image multi-scale stripe noise removal method according to any one of claims 1-3.

6. A computer storage medium, characterized in that, The storage medium stores program instructions, wherein when the program instructions are executed, the device containing the storage medium is controlled to perform the infrared image multi-scale stripe noise removal method according to any one of claims 1-3.

Images