A multi-source noise removal method and system for an infrared imaging system

Abstract

The present application relates to the technical field of image processing, in particular to a multi-source noise removing method and system for an infrared imaging system. The method comprises: acquiring an infrared image collected by the infrared imaging system; constructing a noise mutual interference coupling coefficient according to the brightness distribution of the pixel points in the time neighborhood in the infrared image; combining the brightness change characteristics of the pixel points in the time neighborhood in the infrared image and the noise mutual interference coupling coefficient to determine the noise excitation joint potential of each pixel point in each infrared image; obtaining the potential energy evolution stability index of each region according to the distribution of the noise excitation joint potential in the same region in the local period before each infrared image; obtaining the denoising vicious cycle intensity index according to the similarity of the noise excitation joint potential and the potential energy evolution stability index; and determining the Gaussian kernel space scale parameter by comprehensively considering the potential energy evolution stability index and the denoising vicious cycle intensity index, and filtering the infrared image. The present application improves the denoising effect of the infrared image.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and specifically to a method and system for removing multi-source noise in infrared imaging systems. Background Technology

[0002] Infrared imaging systems are widely used in military reconnaissance, medical diagnosis, and security monitoring. However, their imaging quality is easily affected by multi-source noise, including inherent detector noise, environmental interference, and circuit noise, leading to blurred image details and a decreased signal-to-noise ratio. Therefore, it is necessary to remove complex mixed noise from infrared images.

[0003] Existing methods for multi-source noise removal in infrared imaging systems mainly rely on independent modeling of single-type noise and static filtering under fixed scenes. However, under complex quasi-static or post-registration dynamic imaging conditions, nonlinear coupling exists between multiple sources of noise in the imaging system, such as non-uniform noise, random noise, and fixed-pattern noise. This causes the statistical characteristics of the multi-source noise to drift, resulting in adaptive failure of filtering strategies based on static models. Consequently, the denoised images exhibit texture blurring or residual structured artifacts, limiting the system's accuracy in extracting key information and the reliability of scene perception. Summary of the Invention

[0004] To address the problem of poor denoising performance in existing methods for infrared image denoising, the present invention aims to provide a multi-source noise removal method and system for infrared imaging systems. The specific technical solution adopted is as follows:

[0005] In a first aspect, the present invention provides a method for multi-source noise removal in infrared imaging systems, the method comprising the following steps:

[0006] Acquire multiple consecutive frames of infrared images captured by the infrared imaging system;

[0007] Based on the brightness distribution of pixels in the time neighborhood of infrared images, the noise mutual interference coupling coefficient of each pixel in each frame of infrared images is constructed; combining the brightness change characteristics of pixels in the time neighborhood of infrared images and the noise mutual interference coupling coefficient, the noise excitation joint potential of each pixel in each frame of infrared images is determined.

[0008] Based on the distribution of the joint potential energy of noise excitation in the same region within a local time period before each frame of infrared image, the potential energy evolution stability index of each region in each frame of infrared image is obtained; based on the similarity between the joint potential energy of noise excitation and the potential energy evolution stability index of pixels in each region, the denoising vicious cycle intensity index of each region is obtained.

[0009] By combining the potential energy evolution stability index and the denoising vicious cycle intensity index, the Gaussian kernel spatial scale parameters are determined and used for infrared image filtering.

[0010] Preferably, the step of constructing the noise interference coupling coefficient of each pixel in each frame of the infrared image based on the brightness distribution of pixels in the temporal neighborhood includes:

[0011] The brightness values ​​of the corresponding positions of the candidate pixels in the infrared image within the time neighborhood constitute the brightness sequence of the candidate pixels;

[0012] Obtain the variance of the elements in the brightness sequence;

[0013] The sequence of brightness values ​​at corresponding positions in the next frame of infrared images following each frame of infrared images in the temporal neighborhood of candidate pixels is denoted as the first sequence; the correlation between the brightness sequence of the candidate pixels and the first sequence is calculated.

[0014] Based on the variance of the elements in the brightness sequence and the correlation, the noise interference coupling coefficient of the candidate pixels is obtained.

[0015] The candidate pixel is any pixel in any frame of an infrared image.

[0016] Preferably, obtaining the noise cross-interference coupling coefficient of candidate pixels based on the variance of elements in the brightness sequence and the correlation includes:

[0017] Calculate the first difference between constant 1 and the correlation;

[0018] The product of the first difference and the variance of the elements in the brightness sequence is used as the noise interference coupling coefficient of the candidate pixel.

[0019] Preferably, determining the joint potential energy of noise excitation for each pixel in each frame of the infrared image by combining the brightness variation characteristics of pixels in the temporal neighborhood and the noise cross-interference coupling coefficient includes:

[0020] The brightness sequence is fitted with a straight line to obtain the slope of the fitted line; the drift intensity index of the candidate pixel is obtained based on the slope and the variance of the first difference of the brightness sequence.

[0021] The noise-excited joint potential energy of the candidate pixels is obtained based on the drift intensity index and the noise mutual interference coupling coefficient of the candidate pixels.

[0022] Preferably, obtaining the drift intensity index of candidate pixels based on the variance of the slope and the first-order difference of the brightness sequence includes:

[0023] Obtain the absolute value of the slope; use the product of the absolute value of the slope and the variance of the first-order difference of the brightness sequence as the drift intensity index of the candidate pixel.

[0024] Preferably, the step of obtaining the potential energy evolution stability index of each region in each infrared image frame based on the distribution of the joint potential energy of noise excitation in the same region within a local time period prior to each infrared image frame includes:

[0025] The first average value of the joint potential energy of noise excitation for all pixels in each region is obtained respectively;

[0026] Calculate the sum of the differences of the first average values ​​of the regions corresponding to the regions to be analyzed in two adjacent infrared images within a local time period before the infrared image of the region to be analyzed.

[0027] Calculate the first ratio between the accumulated sum and the first average value corresponding to the region to be analyzed; use the negative correlation mapping result of the first ratio as the potential energy evolution stability index of the region to be analyzed;

[0028] The region to be analyzed is any region in any frame of an infrared image.

[0029] Preferably, the step of obtaining the denoising vicious cycle intensity index of each region based on the similarity between the joint potential energy of noise excitation and the potential energy evolution stability index of pixels in each region includes:

[0030] The noise excitation joint potential energy sequence, which is composed of the noise excitation joint potential energy of each pixel position in the infrared image corresponding to the region to be analyzed within a local time period of the infrared image where the region to be analyzed is located, and the potential energy evolution stability index sequence, which is composed of the potential energy evolution stability index of the corresponding position in the infrared image where the region to be analyzed is located within a local time period of the infrared image, are obtained respectively.

[0031] Calculate the Pearson correlation coefficient between the noise-excited joint potential energy sequence and the potential energy evolution stability index sequence corresponding to each pixel position;

[0032] The ratio of the sum of the absolute values ​​of all Pearson correlation coefficients with negative Pearson correlation coefficients to the number of pixels in the region to be analyzed is used as the intensity index of the denoising vicious cycle in the region to be analyzed.

[0033] Preferably, the comprehensive potential energy evolution stability index and the denoised vicious cycle intensity index are used to determine the Gaussian kernel spatial scale parameters, including:

[0034] Based on the potential energy evolution stability index and the denoising vicious cycle intensity index of the region to be analyzed, the noise contribution dynamic evolution coefficient of the region to be analyzed is obtained.

[0035] Based on the noise contribution dynamic evolution coefficient and the preset Gaussian kernel maximum spatial scale parameter, the Gaussian kernel spatial scale parameter of the region to be analyzed is obtained.

[0036] Preferably, obtaining the noise contribution dynamic evolution coefficient of the region to be analyzed based on the potential energy evolution stability index and the denoising vicious cycle intensity index of the region to be analyzed includes:

[0037] Calculate the second difference between constant 1 and the potential energy evolution stability index of the region to be analyzed, and the first sum of constant 1 and the denoising vicious cycle intensity index of the region to be analyzed, respectively.

[0038] The product of the second difference and the first sum is used as the noise contribution dynamic evolution coefficient of the region to be analyzed.

[0039] In a second aspect, the present invention provides a multi-source noise removal system for infrared imaging systems, the system being used to implement the method of the first aspect, the system comprising:

[0040] The image acquisition module is used to acquire multiple consecutive frames of infrared images collected by the infrared imaging system.

[0041] The first evaluation module is used to construct the noise interference coupling coefficient of each pixel in each frame of infrared image based on the brightness distribution of pixels in the temporal neighborhood; and to determine the noise excitation joint potential of each pixel in each frame of infrared image by combining the brightness change characteristics of pixels in the temporal neighborhood and the noise interference coupling coefficient.

[0042] The second evaluation module is used to obtain the potential energy evolution stability index of each region in each infrared image based on the distribution of the noise excitation joint potential energy in the same region within a local time period before each frame of infrared image; and to obtain the denoising vicious cycle intensity index of each region based on the similarity between the noise excitation joint potential energy of pixels in each region and the potential energy evolution stability index.

[0043] The denoising module is used to determine the Gaussian kernel spatial scale parameters by combining the potential energy evolution stability index and the denoising vicious cycle intensity index, and to perform filtering processing on infrared images.

[0044] The present invention has at least the following beneficial effects:

[0045] This invention constructs a noise interference coupling coefficient based on the brightness distribution and variation characteristics of pixels in the temporal neighborhood of infrared images acquired by an infrared imaging system, thereby determining the joint potential energy of noise excitation. By analyzing the distribution of the joint potential energy of noise excitation in a local time period and its relationship with the potential energy evolution stability index, a denoising vicious cycle intensity index is obtained. Combining the potential energy evolution stability index and the denoising vicious cycle intensity index, the Gaussian kernel spatial scale parameter is adaptively determined and the infrared image is filtered. The method provided by this invention can dynamically assess the spatiotemporal evolution state and vicious cycle risk of noise in the infrared imaging system, and adaptively adjust the filtering intensity to achieve precise processing of strong filtering in high-risk areas and weak filtering in low-risk areas. This avoids texture blurring and detail loss caused by over-filtering in traditional methods to the greatest extent, and also curbs the vicious cycle of noise-artifact-misjudgment, improving the image quality, information extraction accuracy, and scene perception reliability of the infrared imaging system in dynamic and complex scenes. Attached Figure Description

[0046] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart of a multi-source noise removal method for infrared imaging systems provided in an embodiment of the present invention;

[0048] Figure 2 This is a structural block diagram of a multi-source noise removal system for infrared imaging systems provided in an embodiment of the present invention. Detailed Implementation

[0049] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description of a multi-source noise removal method and system for infrared imaging systems based on the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0051] The following detailed description, in conjunction with the accompanying drawings, illustrates a specific scheme for a multi-source noise removal method and system for infrared imaging systems provided by the present invention.

[0052] An embodiment of a multi-source noise removal method for infrared imaging systems:

[0053] The specific scenario addressed in this embodiment is as follows: the infrared images of quasi-static scenes (such as security monitoring background extraction and staring imaging) acquired by the infrared imaging system contain complex noise. This embodiment will adaptively determine the Gaussian kernel spatial scale parameters of the sub-region by combining the acquired infrared image features, so as to achieve adaptive filtering processing of the infrared image.

[0054] This embodiment proposes a multi-source noise removal method for infrared imaging systems, such as... Figure 1 As shown, a multi-source noise removal method for an infrared imaging system according to this embodiment includes the following steps:

[0055] Step S1: Acquire multiple consecutive infrared images captured by the infrared imaging system.

[0056] First, through the image acquisition interface of the infrared imaging system, multiple frames of raw infrared images are continuously acquired per operating condition at the system's nominal frame rate. The frame rate can be selected from three levels: 30fps, 60fps, and 120fps, depending on the dynamic characteristics of the scene. 30fps is selected for static scenes, and 60fps-120fps is selected for fast dynamic scenes. The resolution of a single infrared image is 1280×1024 pixels; the number of infrared images is 500 frames. In specific applications, the implementer can set the frame rate according to the specific situation.

[0057] Through the above methods, multiple consecutive frames of infrared images were obtained.

[0058] Step S2: Based on the brightness distribution of pixels in the time neighborhood of the infrared image, construct the noise mutual interference coupling coefficient of each pixel in each frame of the infrared image; combine the brightness change characteristics of pixels in the time neighborhood of the infrared image and the noise mutual interference coupling coefficient to determine the noise excitation joint potential of each pixel in each frame of the infrared image.

[0059] In infrared imaging systems, fixed pattern noise is caused by the inherent response differences of detector pixels. It manifests as a disturbance of stable temporal characteristics with a fixed spatial position, forming the static basis of noise. Random noise is caused by effects such as electron thermal motion and photon shot, which manifests as irregular temporal fluctuations and dynamic disturbances on the static basis.

[0060] The following embodiment uses a single pixel in an infrared image frame as an example for illustration. Other pixels in other infrared images can be processed using the method provided in this embodiment.

[0061] Specifically, any pixel in any frame of infrared image is recorded as a candidate pixel.

[0062] The brightness value of each pixel is obtained. The brightness values ​​of candidate pixels at corresponding positions within the infrared images in the temporal neighborhood are arranged in chronological order to obtain a brightness sequence of candidate pixels. The infrared images in the temporal neighborhood of a candidate pixel are obtained by acquiring multiple consecutive frames of infrared images centered on the infrared image containing the candidate pixel, and these acquired images are used as the infrared images in the temporal neighborhood of the candidate pixel. It should be noted that the infrared images in the temporal neighborhood do not include the infrared image containing the candidate pixel. In this embodiment, the number of infrared images in the temporal neighborhood is 10; in specific applications, the implementer can set this number according to specific circumstances. The corresponding position of a candidate pixel within the infrared images in the temporal neighborhood is the same position within the infrared images in the temporal neighborhood of the infrared image containing the candidate pixel.

[0063] Obtain the variance of elements in the brightness sequence of candidate pixels. Arrange the brightness values ​​of the corresponding positions in the next infrared image after each frame of the infrared image in the temporal neighborhood of the candidate pixels in chronological order, and denote this obtained sequence as the first sequence. Calculate the correlation between the brightness sequence of candidate pixels and the first sequence. In this embodiment, the Pearson correlation coefficient is used to characterize the correlation; that is, the Pearson correlation coefficient between the brightness sequence of candidate pixels and the first sequence is taken as the correlation between the brightness sequence of candidate pixels and the first sequence. The correlation is used to quantify the stability of the fixed pattern noise, and its value range is... The closer its value is to 1, the stronger the consistency of pixel brightness between adjacent frames, the higher the proportion of fixed pattern noise, and the less it is contaminated by random noise; the smaller its value, the more regular the temporal fluctuation of pixel brightness, mainly dominated by random noise.

[0064] Furthermore, the noise interference coupling coefficient of the candidate pixels is obtained based on the variance and correlation of the elements in the brightness sequence of the candidate pixels.

[0065] As a specific example, the difference obtained by subtracting the correlation from the constant 1 is recorded as the first difference. The first difference is used to quantify the tolerance scale of fixed pattern noise to random noise. The smaller the correlation, the larger the first difference, the weaker the stability of the fixed pattern noise, and the easier it is for the disturbance of random noise to break through the limitation of the fixed pattern noise, forming a stronger coupling interference. Therefore, the product of the first difference and the variance of the elements in the brightness sequence of the candidate pixel is used as the noise mutual interference coupling coefficient of the candidate pixel. The larger the noise mutual interference coupling coefficient, the more significant the coupling effect between the random noise and the fixed pattern noise of the candidate pixel, and the higher the denoising priority.

[0066] Considering that non-uniform noise originates from the inconsistent response of detector pixels to the same radiation, and that this response drifts with time and temperature, manifesting as slow changes in pixel brightness and inter-frame fluctuations, it injects time-varying energy into the multi-source noise coupling system, causing its noise statistical characteristics to drift.

[0067] Based on the above characteristics, the least squares method is used to perform linear fitting on the brightness sequence of candidate pixels to obtain a fitted line. The horizontal axis of the fitted line represents the acquisition time, and the vertical axis represents the brightness value. The slope of the fitted line is obtained. The brightness value of a pixel is its grayscale value in the infrared grayscale image. The methods for linear fitting and obtaining the slope of the fitted line are existing technologies and will not be elaborated further here. The variance of the first-order difference of the brightness sequence of candidate pixels is obtained. This variance reflects the degree of rapid random fluctuation in pixel brightness between frames, quantifying the instability during the drift process. The larger the variance, the more severe the inter-frame pixel brightness jitter accompanying the drift process, and the more serious the damage to image quality caused by non-uniform noise. The absolute value of the slope of the fitted line is obtained. This absolute value quantifies the rate of pixel drift over time. The larger the rate, the more significant the time-varying characteristics of non-uniform noise. The product of the absolute value of the slope of the fitted line and the variance of the first-order difference of the brightness sequence of candidate pixels is used as the drift intensity index of the candidate pixel. The impact of non-uniform noise on the noise coupling system mainly depends on the synergistic effect of drift rate and drift stability. When the drift rate is fast and fluctuates violently, it provides a strong driving force for the noise coupling system, leading to the malignant evolution of image noise. Based on the above characteristics, the noise excitation joint potential energy of the candidate pixel is obtained according to the drift intensity index and the noise mutual interference coupling coefficient of the candidate pixel.

[0068] As a specific implementation, the noise-excited joint potential energy of candidate pixels can be determined as follows: Specifically, the sum of the drift intensity index and the noise mutual interference coupling coefficient of the candidate pixels is calculated, and the normalized result of this sum is taken as the noise-excited joint potential energy of the candidate pixels. A larger noise-excited joint potential energy indicates that the candidate pixels are experiencing rapid and unstable response drift, and are hotspot pixels in the multi-source noise coupling evolution, requiring close attention in the subsequent denoising process. In this embodiment, the data normalization uses the maximum-minimum normalization method, which is an existing technology and will not be elaborated further here.

[0069] Using the above method, the noise excitation joint potential of each pixel in each frame of infrared image can be obtained.

[0070] Step S3: Based on the distribution of the noise excitation joint potential energy in the same region within the local time period before each frame of infrared image, obtain the potential energy evolution stability index of each region in each frame of infrared image; based on the similarity between the noise excitation joint potential energy of pixels in each region and the potential energy evolution stability index, obtain the denoising vicious cycle intensity index of each region.

[0071] Since the mutual interference coupling of random-fixed pattern noise constitutes the static basis of noise, while the drift intensity index of non-uniform noise provides a time-varying driving force for this basis, the synergistic effect of the two determines the noise excitation intensity of the pixel: the stronger the coupling effect of the static noise basis and the greater the time-varying driving force, the higher the excitation energy of the noise and the more serious the damage to image quality.

[0072] Each frame of infrared image is divided into multiple regions of equal size, with each region in different infrared images having a one-to-one correspondence. The size of the sub-regions in the infrared image is set by the implementer according to specific circumstances; for example, each frame of infrared image can be divided into 25 regions.

[0073] The average value of the noise excitation joint potential energy of all pixels in each region is calculated and recorded as the first average value; each region in each frame of infrared image has a corresponding first average value.

[0074] The following explanation uses a region in a frame of infrared image as an example. Other regions in the same frame of infrared image or regions in other infrared images can be processed using the method provided in this embodiment.

[0075] Specifically, any region in any frame of infrared image is designated as the region to be analyzed. The absolute value of the difference between the first average values ​​corresponding to the regions corresponding to the region to be analyzed in two adjacent frames of infrared image within a local time period preceding the infrared image containing the region to be analyzed is calculated. This absolute value is used as the difference between the first average values ​​corresponding to the regions corresponding to the region to be analyzed in two adjacent frames of infrared image. The sum of the differences between the first average values ​​corresponding to the regions corresponding to the region to be analyzed in all adjacent frames of infrared image within a local time period preceding the infrared image containing the region to be analyzed is calculated. The ratio between this sum and the first average value corresponding to the region to be analyzed is designated as the first ratio. The negative correlation mapping result of the first ratio is used as the potential energy evolution stability index of the region to be analyzed. In this embodiment, the exponential function value with the natural constant as the base and the negative first ratio as the exponent is used as the potential energy evolution stability index of the region to be analyzed, thus achieving a negative correlation mapping of the first ratio. The smaller the potential energy evolution stability index, the more unstable the noise evolution path of the region to be analyzed, and the more likely it is to be in a high-risk state. In this embodiment, the local time period is 5 seconds. In specific applications, the implementer can set this value according to specific circumstances. It should be noted that when calculating the first ratio, if the denominator of the calculation formula is 0, a zero-prevention parameter is added to the denominator before calculation to prevent the denominator from being 0. In this embodiment, the zero-prevention parameter is... In specific applications, implementers can configure it according to the specific circumstances.

[0076] In the actual imaging process of infrared imaging systems, there is a vicious cycle of noise enhancement, artifact generation, misjudgment of noise, and over-filtering. That is, noise enhancement causes the denoising algorithm to produce artifacts, and the artifacts are misjudged as noise, which drives the algorithm to perform strong filtering operations, thus destroying image details.

[0077] Based on the above characteristics, the noise-excited joint potential energy of each pixel in the infrared image corresponding to the region to be analyzed within a local time period of the infrared image of the region to be analyzed is obtained. These noise-excited joint potential energies are arranged in temporal order to obtain a noise-excited joint potential energy sequence. The potential energy evolution stability index of the corresponding position in the infrared image of the region to be analyzed within a local time period of the infrared image of the region to be analyzed is obtained. These potential energy evolution stability indices are arranged in temporal order to obtain a potential energy evolution stability index sequence.

[0078] The core manifestation of the vicious cycle is that as the noise-excited potential energy increases, the system stability deteriorates simultaneously. Therefore, the Pearson correlation coefficient between the noise-excited joint potential energy sequence and the potential energy evolution stability index sequence corresponding to each pixel position is calculated. The range of the Pearson correlation coefficient is... A negative Pearson correlation coefficient indicates a negative correlation between noise potential energy and stability index; that is, the higher the potential energy, the worse the stability, consistent with the characteristics of a vicious cycle. Therefore, the ratio of the sum of the absolute values ​​of all negative Pearson correlation coefficients to the number of pixels in the region to be analyzed is used as the denoising vicious cycle intensity index of the region to be analyzed. The sum of the absolute values ​​of all negative Pearson correlation coefficients reflects the total intensity of a significant negative correlation. The larger the denoising vicious cycle intensity index of the region to be analyzed, the more stable and strong the negative correlation between noise potential energy and evolutionary stability is, indicating that the system has fallen into a high-risk-unstable vicious cycle state, with an extremely high risk of artifact generation and detail loss.

[0079] Using the above method, the denoising vicious cycle intensity index of each region in each frame of infrared image can be obtained.

[0080] Step S4: Combine the potential energy evolution stability index and the denoising vicious cycle intensity index to determine the Gaussian kernel spatial scale parameters, and then perform filtering processing on the infrared image.

[0081] In this embodiment, the potential energy evolution stability index and the denoising vicious cycle intensity index are obtained in the above steps. Next, the Gaussian kernel spatial scale parameters of each region will be adaptively determined by combining the potential energy evolution stability index and the denoising vicious cycle intensity index to achieve filtering processing of the infrared image.

[0082] Specifically, the difference between constant 1 and the potential energy evolution stability index of the region to be analyzed is calculated and recorded as the second difference; the sum of constant 1 and the denoising vicious cycle intensity index of the region to be analyzed is recorded as the first sum; the product of the second difference and the first sum is used as the noise contribution dynamic evolution coefficient of the region to be analyzed. In this embodiment, a specific calculation formula for the noise contribution dynamic evolution coefficient is given, and the noise contribution dynamic evolution coefficient of the region to be analyzed can be expressed as:

[0083]

[0084] in, This represents the dynamic evolution coefficient of noise contribution in the region to be analyzed. This represents the potential energy evolution stability index of the region to be analyzed. This represents the intensity index of the denoising vicious cycle in the region to be analyzed.

[0085] This represents the second difference, used to convert the stability index into the instability coefficient; The first sum represents the vicious cycle intensity converted into a risk amplification factor. The larger the noise contribution dynamic evolution coefficient of the region to be analyzed, the more it indicates that the noise in the region to be analyzed is not only on a highly unstable dynamic evolution path, but this instability is also being continuously amplified by the vicious cycle mechanism. The greater the risk of evolving from the noise accumulation state to the artifact generation and detail loss state, the stronger the adaptive denoising processing needs to be applied.

[0086] Next, based on the dynamic evolution coefficient of noise risk in the region to be analyzed, a multi-scale Gaussian filtering adaptive strategy will be constructed to achieve accurate denoising by strong filtering in high-risk areas and weak filtering in low-risk areas.

[0087] Specifically, the Gaussian kernel spatial scale parameter of the region to be analyzed is obtained based on the noise contribution dynamic evolution coefficient of the region to be analyzed and the preset maximum spatial scale parameter of the Gaussian kernel. As a specific example, the product of the noise contribution dynamic evolution coefficient of the region to be analyzed and the preset maximum spatial scale parameter of the Gaussian kernel is used as the Gaussian kernel spatial scale parameter of the region to be analyzed. In this embodiment, the preset maximum spatial scale parameter of the Gaussian kernel is 2.0.

[0088] Using the above method, the Gaussian kernel spatial scale parameters for each region can be obtained. Based on the Gaussian kernel spatial scale parameters of all regions, corresponding Gaussian convolution kernels are generated. The sum of the weights of the convolution kernel elements corresponding to all regions in a frame of infrared image is 1. By performing convolution operation between the convolution kernel and the pixels in the corresponding region of the infrared image, the filtered image can be obtained, thus completing the adaptive filtering and denoising processing of the infrared image. Among them, Gaussian filtering is an existing technology and will not be elaborated on further here.

[0089] Thus, by using the method provided in this embodiment, adaptive filtering processing of infrared images acquired by the infrared imaging system has been achieved.

[0090] This embodiment constructs a noise interference coupling coefficient based on the brightness distribution and variation characteristics of pixels in the temporal neighborhood of infrared images acquired by an infrared imaging system, thereby determining the joint potential energy of noise excitation. By analyzing the distribution of the joint potential energy of noise excitation in a local time period and its relationship with the potential energy evolution stability index, a denoising vicious cycle intensity index is obtained. Combining the potential energy evolution stability index and the denoising vicious cycle intensity index, the Gaussian kernel spatial scale parameter is adaptively determined and the infrared image is filtered. The method provided in this embodiment can dynamically evaluate the spatiotemporal evolution state and vicious cycle risk of noise in response to the nonlinear coupling characteristics of multi-source noise in infrared imaging systems, and adaptively adjust the filtering intensity to achieve precise processing of strong filtering in high-risk areas and weak filtering in low-risk areas. This avoids texture blurring and detail loss caused by over-filtering in traditional methods to the greatest extent, and also curbs the vicious cycle of noise-artifact-misjudgment, improving the image quality, information extraction accuracy, and scene perception reliability of infrared imaging systems in dynamic and complex scenes.

[0091] An embodiment of a multi-source noise removal system for infrared imaging systems:

[0092] See Figure 2 The diagram illustrates a structural block diagram of a multi-source noise removal system for an infrared imaging system according to an embodiment of the present invention. The system may include an image acquisition module, a first evaluation module, a second evaluation module, and a noise reduction module.

[0093] The image acquisition module is used to acquire multiple consecutive frames of infrared images collected by the infrared imaging system.

[0094] The first evaluation module is used to construct the noise interference coupling coefficient of each pixel in each frame of infrared image based on the brightness distribution of pixels in the temporal neighborhood; and to determine the noise excitation joint potential of each pixel in each frame of infrared image by combining the brightness change characteristics of pixels in the temporal neighborhood and the noise interference coupling coefficient.

[0095] The second evaluation module is used to obtain the potential energy evolution stability index of each region in each infrared image based on the distribution of the noise excitation joint potential energy in the same region within a local time period before each frame of infrared image; and to obtain the denoising vicious cycle intensity index of each region based on the similarity between the noise excitation joint potential energy of pixels in each region and the potential energy evolution stability index.

[0096] The denoising module is used to determine the Gaussian kernel spatial scale parameters by combining the potential energy evolution stability index and the denoising vicious cycle intensity index, and to perform filtering processing on infrared images.

[0097] It should be understood that Figure 2The block diagram and modules of the multi-source noise removal system for infrared imaging systems shown can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by appropriate instructions, such as a microprocessor or dedicated hardware. Those skilled in the art will understand that the methods and apparatus described above can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The apparatus and modules described in this specification can be implemented not only with hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., but also with software, for example, executed by various types of processors, or with a combination of the aforementioned hardware circuits and software (e.g., firmware).

[0098] For more details about the above modules, please refer to other parts of this manual; they will not be repeated here.

[0099] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for multi-source noise removal in infrared imaging systems, characterized in that, The method includes the following steps: Acquire multiple consecutive frames of infrared images captured by the infrared imaging system; Based on the brightness distribution of pixels in the time neighborhood of infrared images, the noise mutual interference coupling coefficient of each pixel in each frame of infrared images is constructed; combining the brightness change characteristics of pixels in the time neighborhood of infrared images and the noise mutual interference coupling coefficient, the noise excitation joint potential of each pixel in each frame of infrared images is determined. Based on the distribution of the joint potential energy of noise excitation in the same region within a local time period before each frame of infrared image, the potential energy evolution stability index of each region in each frame of infrared image is obtained; based on the similarity between the joint potential energy of noise excitation and the potential energy evolution stability index of pixels in each region, the denoising vicious cycle intensity index of each region is obtained. By combining the potential energy evolution stability index and the denoising vicious cycle intensity index, the Gaussian kernel spatial scale parameters are determined and used for infrared image filtering. The construction of the noise interference coupling coefficient for each pixel in each frame of infrared image includes: The brightness values ​​of the corresponding positions of the candidate pixels in the infrared image within the time neighborhood constitute the brightness sequence of the candidate pixels; Obtain the variance of the elements in the brightness sequence; The sequence of brightness values ​​at corresponding positions in the next frame of infrared images following each frame of infrared images in the temporal neighborhood of candidate pixels is denoted as the first sequence; the correlation between the brightness sequence of the candidate pixels and the first sequence is calculated. Based on the variance of the elements in the brightness sequence and the correlation, the noise interference coupling coefficient of the candidate pixels is obtained. The candidate pixel is any pixel in any frame of infrared image; The step of obtaining the noise cross-interference coupling coefficient of candidate pixels based on the variance of elements in the brightness sequence and the correlation includes: Calculate the first difference between constant 1 and the correlation; The product of the first difference and the variance of the elements in the brightness sequence is used as the noise interference coupling coefficient of the candidate pixel. The determination of the noise excitation joint potential of each pixel in each frame of the infrared image by combining the brightness variation characteristics of pixels in the time neighborhood and the noise mutual interference coupling coefficient includes: The brightness sequence is fitted with a straight line to obtain the slope of the fitted line; the drift intensity index of the candidate pixel is obtained based on the slope and the variance of the first difference of the brightness sequence. Based on the drift intensity index and the noise mutual interference coupling coefficient of the candidate pixels, the noise excitation joint potential energy of the candidate pixels is obtained. The step of obtaining the drift intensity index of candidate pixels based on the slope and the variance of the first-order difference of the brightness sequence includes: Obtain the absolute value of the slope; use the product of the absolute value of the slope and the variance of the first difference of the brightness sequence as the drift intensity index of the candidate pixel. The method of obtaining the potential energy evolution stability index of each region in each infrared image frame based on the distribution of the joint potential energy of noise excitation in the same region within a local time period preceding each frame of infrared image includes: The first average value of the noise excitation joint potential energy of all pixels in each region is obtained respectively; Calculate the sum of the differences of the first average values ​​of the regions corresponding to the regions to be analyzed in two adjacent infrared images within a local time period before the infrared image of the region to be analyzed. Calculate the first ratio between the accumulated sum and the first average value corresponding to the region to be analyzed; use the negative correlation mapping result of the first ratio as the potential energy evolution stability index of the region to be analyzed; The region to be analyzed is any region in any frame of infrared image. The method for obtaining the denoising vicious cycle intensity index of each region based on the similarity between the joint potential energy of noise excitation and the stability index of potential energy evolution of pixels in each region includes: The noise excitation joint potential energy sequence, which is composed of the noise excitation joint potential energy of each pixel position in the infrared image corresponding to the region to be analyzed within a local time period of the infrared image where the region to be analyzed is located, and the potential energy evolution stability index sequence, which is composed of the potential energy evolution stability index of the corresponding position in the infrared image where the region to be analyzed is located within a local time period of the infrared image, are obtained respectively. Calculate the Pearson correlation coefficient between the noise-excited joint potential energy sequence and the potential energy evolution stability index sequence corresponding to each pixel position; The ratio of the sum of the absolute values ​​of all Pearson correlation coefficients with negative Pearson correlation coefficients to the number of pixels in the region to be analyzed is used as the denoising vicious cycle intensity index of the region to be analyzed. The combined potential energy evolution stability index and the denoised vicious cycle intensity index are used to determine the Gaussian kernel spatial scale parameters, including: Based on the potential energy evolution stability index and the denoising vicious cycle intensity index of the region to be analyzed, the noise contribution dynamic evolution coefficient of the region to be analyzed is obtained. Based on the noise contribution dynamic evolution coefficient and the preset Gaussian kernel maximum spatial scale parameter, the Gaussian kernel spatial scale parameter of the region to be analyzed is obtained; The process of obtaining the noise contribution dynamic evolution coefficient of the region under analysis based on the potential energy evolution stability index and the denoising vicious cycle intensity index of the region under analysis includes: Calculate the second difference between constant 1 and the potential energy evolution stability index of the region to be analyzed, and the first sum of constant 1 and the denoising vicious cycle intensity index of the region to be analyzed, respectively. The product of the second difference and the first sum is used as the noise contribution dynamic evolution coefficient of the region to be analyzed.

2. A multi-source noise removal system for an infrared imaging system, the system being used to perform the method of claim 1, characterized in that, The system includes: The image acquisition module is used to acquire multiple consecutive frames of infrared images collected by the infrared imaging system. The first evaluation module is used to construct the noise interference coupling coefficient of each pixel in each frame of infrared image based on the brightness distribution of pixels in the temporal neighborhood; and to determine the noise excitation joint potential of each pixel in each frame of infrared image by combining the brightness change characteristics of pixels in the temporal neighborhood and the noise interference coupling coefficient. The second evaluation module is used to obtain the potential energy evolution stability index of each region in each infrared image based on the distribution of the noise excitation joint potential energy in the same region within a local time period before each frame of infrared image; and to obtain the denoising vicious cycle intensity index of each region based on the similarity between the noise excitation joint potential energy of pixels in each region and the potential energy evolution stability index. The denoising module is used to determine the Gaussian kernel spatial scale parameters by combining the potential energy evolution stability index and the denoising vicious cycle intensity index, and to perform filtering processing on infrared images.

Images