EBAPS device image adaptive digital noise reduction method and device and storage medium

Abstract

The invention relates to the technical field of electron bombardment active pixel sensor preparation, and particularly discloses an EBAPS device image adaptive digital noise reduction method and device and a storage medium, and the method comprises the steps: setting initial parameters which comprise a sub-region image matrix dimension parameter Ns, an abnormal point threshold proportionality coefficient alpha, a sub-region image matrix dimension parameter Ns and an abnormal point threshold proportionality coefficient alpha; a sub-region image filtering matrix dimension parameter Nt, wherein Ntlt; ns; and according to the defined initial parameter Ns, performing sub-region segmentation on the acquired digital image to enable the matrix dimension of the sub-region image to be Ns * Ns, and implementing an adaptive denoising algorithm in the sub-region image obtained after segmentation for processing. The invention relates to an EBASP digital device manufacturing method, and aims to solve the problem that in the prior art, in the manufacturing process of an EBASP digital device, the requirements for the vacuum degree and the processing technology of the device are high, and due to the limitation of objective conditions such as technologies and technologies, different noises are inevitably introduced in all working stages of the device, so that the quality of an output image is affected.

Description

Technical Field

[0001] The present application relates to the technical field of electron bombardment active pixel sensor preparation, and particularly relates to an image adaptive digital noise reduction method, device and storage medium for an EBAPS device. Background Art

[0002] An electron bombardment active pixel sensor (EBAPS) belongs to a new type of digital low-light device. This device can convert a two-dimensional spatial image under low-illumination and weak-light conditions into a video signal convenient for human eyes to observe. It has prominent features such as high gain, low power consumption, simple structure, and digital display. It is the latest progress in the field of low-light night vision technology and also a new direction for future technological development.

[0003] During the preparation of EBASP digital devices, high requirements are placed on the vacuum degree and processing technology of the devices. Limited by objective conditions such as technology and process, different noises are inevitably introduced at each stage of the device operation, such as thermal noise, gain noise, shot noise, etc., thus affecting the quality of the output image. Summary of the Invention

[0004] The purpose of the present application is to provide an image adaptive digital noise reduction method, device and storage medium for an EBAPS device, so as to solve the problem in the prior art that during the preparation of EBASP digital devices, high requirements are placed on the vacuum degree and processing technology of the devices. Limited by objective conditions such as technology and process, different noises are inevitably introduced at each stage of the device operation, thus affecting the quality of the output image.

[0005] To achieve the above purpose, an embodiment of the present application provides an image adaptive digital noise reduction method for an EBAPS device, including the following steps:

[0006] Set initial parameters, where the initial parameters include the sub-region image matrix dimension parameter N s , the outlier threshold ratio coefficient α, and the sub-region image filtering matrix dimension parameter N t , where N t <N s ;

[0007] According to the defined initial parameter N s , perform sub-region segmentation on the collected digital image, so that the dimension of the sub-region image matrix is N s ×N s , and perform processing using an adaptive denoising algorithm on the sub-region image obtained after segmentation.

[0008] Optionally, an adaptive denoising algorithm is implemented on the sub-region images obtained after segmentation, specifically including:

[0009] In each of the sub-region images, calculate the first average value and the first standard deviation of the gray values of the pixel points in the sub-region image respectively;

[0010] According to the initialization parameter α, define the points within the numerical range of α times the first standard deviation centered on the first average value as abnormal points, and perform zero assignment preprocessing;

[0011] Within the range of the sub-region image, recalculate the non-abnormal points to obtain the corresponding second average value and second standard deviation;

[0012] According to the recalculated second average value and second standard deviation, according to the initialization parameter N t , generate a corresponding filtering matrix with a dimension of N t ×N t ;

[0013] Use the generated N t ×N t filtering matrix to perform filtering denoising on the sub-region image of N s ×N s . For non-abnormal points, the corresponding updated value is obtained by synthesizing the original value of the point and the values in the neighboring regions according to the filtering matrix. For abnormal points, the corresponding updated value is obtained by synthesizing the values in the neighboring regions according to the filtering matrix.

[0014] Optionally, the sub-region segmentation of the acquired digital image specifically includes:

[0015] The matrix corresponding to the single-frame acquired image is A, and the corresponding number of rows and columns is N R and N C . According to the preset parameter N for the sub-region image dimension s , divide the acquired image A into N' R rows, N' C columns, a total of N' R ×N' C sub-region images B (m,n) (0 ≤ m ≤ N' R -1, 0 ≤ n ≤ N' C -1):

[0016]

[0017] where the symbol represents the largest integer not exceeding the number .

[0018] Optionally, in each of the sub-region images, calculating a first average value and a first standard deviation of the gray values of the pixel points in the sub-region image specifically includes:

[0019] From the acquired image A and its corresponding element a (k,l) (0 ≤ k ≤ N R -1, 0 ≤ l ≤ N C -1), calculating to obtain the sub-region image B (m,n) The corresponding element in the matrix

[0020]

[0021] According to the sub-region image matrix B (m,n) , calculating the first average value in the sub-region And the corresponding first standard deviation

[0022]

[0023] Optionally, according to the initialization parameter α, defining the points within the numerical range centered on the first average value and α times the first standard deviation as abnormal points, and performing zero-assignment preprocessing, specifically including:

[0024] Performing preprocessing on the abnormal points in the sub-region image matrix B (m,n) . According to the preset parameter α of the abnormal points, if the value of the corresponding pixel point Is not within the range of the first average value of the sub-region and α times the first standard deviation Then it is defined as an abnormal point and zero-assignment preprocessing is performed. The preprocessed sub-region image matrix is denoted as B ( ' m,n) , and the corresponding element is denoted as

[0025]

[0026] Optionally, within the range of the sub-region image, recalculating the non-abnormal points to obtain the corresponding second average value and second standard deviation, specifically including:

[0027] Recalculating the corresponding second average value for the preprocessed sub-region image matrix B ( ' m,n) And the corresponding second standard deviation Where the parameter N ' represents the number of non-zero elements in the sub-region image matrix B s ' ( ' m,n) In, the element Not zero:

[0028]

[0029] Optionally, according to the recalculated second average value and second standard deviation, and according to the initialization parameter N t , a corresponding filtering matrix is generated, specifically including:

[0030] According to the sub-region image filtering matrix dimension parameter N t , the sub-region image B ( ' m,n) Filter matrix T (m,n) is used to filter the elements in B ( ' m,n) . The corresponding filtering matrix T The element in (m,n) is Expression:

[0031]

[0032] Optionally, the generated N t ×N t filtering matrix is used to filter and denoise the N s ×N s sub-region image, specifically including:

[0033] Using the generated filtering matrix T (m,n) , the sub-region image matrix B ( ' m,n) is filtered to generate the sub-region image C (m,n) after the filter, and its corresponding element is

[0034]

[0035] To achieve the above object, the present application also provides an EBAPS device image adaptive digital noise reduction device, including: a memory; and

[0036] a processor connected to the memory, and the processor is configured to execute the steps of the method as described above.

[0037] To achieve the above object, the present application also provides a computer storage medium, on which a computer program is stored, and when the computer program is executed by a machine, the steps of the method as described above are implemented.

[0038] The embodiments of the present application have the following advantages:

[0039] Through the above method, it is possible to eliminate the influence of different noise distributions in different regions of a single-frame image on the image quality, and to alleviate the influence of the difference in noise distribution between different image frames on the image quality, thus having strong robustness. In this application, by automatically calculating the single-frame filtering kernel matrix, the influence of the difference in noise distribution between single frames on the image quality is reduced; by automatically calculating the multi-frame filtering kernel matrix, the influence of the difference in noise distribution between multi-frames on the image quality is reduced; this application can achieve a relatively fast processing speed when processing according to single frames, while taking into account the difference in noise distribution of single frames and the difference in noise distribution between multi-frames, and has good performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings described below are only exemplary, and for those of ordinary skill in the art, without creative efforts, other implementation drawings can be obtained according to the provided drawings.

[0041] Figure 1 It is a flowchart of an image adaptive digital noise reduction method for an EBAPS device provided by at least one embodiment of the present application;

[0042] Figure 2 It is a filter time-domain coefficient diagram of an image adaptive digital noise reduction method for an EBAPS device provided by at least one embodiment of the present application;

[0043] Figure 3 It is a module block diagram of an image adaptive digital noise reduction device for an EBAPS device provided by at least one embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] The following specific embodiments illustrate the embodiments of the present application. Those skilled in the art can easily understand the other advantages and effects of the present application from the content disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of them. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0045] It should be noted that in the claims and the description of the present application, the steps can be executed substantially in parallel or in the reverse order under appropriate circumstances, depending on the functions involved.

[0046] In addition, the technical features involved in different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0047] At present, EBAPS image denoising algorithms are generally divided into two categories. The first category of methods is based on the global noise statistical characteristics of images. Methods such as wavelet transform and deep learning are used to perform noise characteristic statistics, and corresponding adaptive thresholds are calculated according to the statistical characteristics. This type of method ignores the difference between the global noise distribution and the local noise distribution of images. As the difference in noise distribution increases, the denoising performance of this type of method will be greatly reduced. The second category of methods uses local denoising methods. This type of method requires approximate prior data on the local distribution of image noise, selects an appropriate filtering window, and then calculates the processing threshold based on this filtering window. In addition to requiring local prior data, this type of method ignores the randomness of the noise introduced by the images collected at different times during the preparation of EBAPS devices, that is, there are also differences in the local noise distribution in the same area of the image at different times. This will also greatly affect the performance of such denoising algorithms.

[0048] If the EBAPS image denoising algorithm is classified according to the number of images required for data processing, it is roughly divided into two different methods: single-frame image data processing and multi-frame image data processing. By analyzing the existing EBAPS algorithms, the following general rule can be obtained. The denoising effect of the algorithm using multi-frame image data processing is generally stronger than that using single-frame image data processing. In other words, this type of algorithm needs to collect multiple frames of data and use multiple frames of data for comprehensive analysis. On the one hand, collecting multiple frames of images takes a longer time than single-frame images and consumes a longer data time, affecting real-time performance. On the other hand, the storage space occupied by multiple frames of data is larger, and higher hardware configuration is also required. The improvement of the denoising effect comes at the cost of time and storage space.

[0049] In 2021, according to the intensity of the subband signal and the noise intensity, Reference [1] automatically calculated an adaptive threshold designed for noise processing. This value is proportional to the noise variance and inversely proportional to the subband signal intensity, protecting the original signal as much as possible while removing noise. In 2022, Reference [2] used the idea of wavelet transform to process image noise. This method assumes that the wavelet coefficients of the noise are mainly concentrated in the first-level detail layer, and accordingly designed a threshold that varies with the decomposition level to improve the noise filtering effect. In 2022, Reference [3] proposed a quadratic harmonic noise reduction method based on the composite algorithm of variational mode decomposition and wavelet threshold function, reducing the noise reduction threshold by increasing the decomposition level to better suppress noise and achieve a better filtering effect. In 2024, Reference [4] used the single-frame images collected, utilized the dark pixel structure characteristics of EBAPS, analyzed the noise characteristics of EBAPS through the images, estimated the corresponding noise intensity, and improved the traditional wavelet threshold noise reduction according to the estimated noise intensity to achieve adaptive variable threshold processing of the noise. In 2024, Reference [5] applied the deep learning method to EBAPS image noise reduction. According to the characteristics of the EBAPS noise image, a self-supervised two-stage denoising CNN algorithm for the mixed noise of EBAPS images was proposed, and self-supervised learning was achieved by combining U-Net and BSN and drawing on the iterative training idea of IDR.

[0050] The adaptive threshold or deep learning methods adopted in References [1]-[5] are all based on the analysis and statistics of the global noise of EBASP images, and it is required that during the preparation process of EBASP digital devices, the noise distribution characteristics do not change in order to obtain better noise reduction and filtering effects. However, in the actual preparation process, the conditions required by the above adaptive methods cannot be fully met. On the one hand, for the images obtained by EBASP digital devices, the noise is not evenly distributed across the entire image, and the filtering metrics obtained based on the statistical characteristics of the global noise will result in unsatisfactory image denoising due to the difference between the global noise distribution characteristics and the local noise distribution of the image. On the other hand, due to the differences in production processes and production equipment at different times, the statistical characteristics of the introduced noise distribution will also vary during the preparation of the same or different models of EBASP digital devices. For example, thermal noise and gain noise with relatively large statistical characteristics are introduced, which will also greatly reduce the effect of image denoising using the above algorithms. For this reason, CN118134799A discloses an FPGA-based EBAPS low-light image denoising method and system. By continuously acquiring 4 frames of images and determining the shape and size of the local filtering window for each image frame according to the distribution of the image sequence in the time direction, the filtering window is divided into finite shapes such as right angles, quasi-triangles, and semi-triangles for separate processing, and good results have been achieved. However, this method also has the following deficiencies: (1) The shape of the filtering window selected by this method needs to roughly know the distribution of the local noise of the image in order to achieve better filtering effects. In fact, it is similar to References [1]-[5]. During the mass production of devices, it is impossible to accurately know in advance the noise distribution characteristics of each digital device, which makes it extremely difficult to select different filter windows and difficult to guarantee the effect; (2) This algorithm needs to continuously acquire 4 frames of images for joint processing, which will result in too long processing time and increased data volume to be processed, reducing the production efficiency under other unchanged conditions.

[0051] The above references:

[0052] [1] Fang Bin, Chen Jiayi. Wavelet Threshold Denoising Algorithm for Removing Impulse Noise [J]. Laser & Optoelectronics Progress, 2021, 58(22).

[0053] [2]Sun K Z, Lu Y C, Huang L S, et al. Wavelet denoising method based on improved threshold function[C]∥2022 IEEE 10th Joint International Information Technology and Artificial Intelligence Conference(ITAIC), June 17-19, 2022, Chongqing, China. New York: IEEE Press, 2022: 1402-1406。

[0054] [3]Zhang Ruilin, Tu Xinghua. Variational mode decomposition and wavelet threshold function denoising of second harmonic waves[J]. Acta Optica Sinica, 2022, 42(2).

[0055] [4]Liu Xuan, Li Bingzhen, Li Li, et al. Adaptive wavelet threshold denoising of EBAPS images based on pixel dark noise estimation[J]. Acta Optica Sinica, 2024, 44(16).

[0056] [5]Li Bingzhen, Liu Xuan, Zhao Zixiang, et al. Self-supervised two-stage denoising algorithm for EBAPS images based on blind spot network[J]. Acta Optica Sinica, 2024, 44(22).

[0057] Due to the limitations of objective conditions such as technology and process, during the preparation of EBAPS digital devices, it is inevitable to introduce ineliminable noise in the imaging process. Currently, this type of noise can only be suppressed by digital processing methods for the noise contained in the acquired images. This application is proposed to address the deficiencies of the image denoising algorithm during the production of EBAPS devices, improve the robustness of the denoising algorithm, adaptively suppress the noise according to the characteristics of the image, and improve the image quality. The overall concept of this application includes:

[0058] (1) First, divide the image into regions, preprocess the abnormal points in the region, then analyze the characteristics of the image in the region, and adaptively calculate the corresponding filter matrix kernel. After this step of processing, single-frame data processing can be carried out, compensating for the impact of the difference in noise distribution in the image region on the image quality, and automatically updating the filter kernel matrix as the acquired image frames are different, alleviating the impact of the noise distribution in different frames on the image quality, and automatically calculating the matrix corresponding to the filter kernel according to the regional statistical characteristics without the need for prior data on the noise distribution, so as to improve the deficiencies existing in the current algorithm and effectively enhance the EBAPS image denoising effect.

[0059] (2) Before performing local filtering, it is necessary to perform multiple preprocessings and parameter calculations on the data in this area. The required execution time is slightly longer than that of the current single-frame image processing algorithm, but the algorithm performance is improved significantly, with better comprehensive performance.

[0060] An embodiment of the present application provides an EBAPS device image adaptive digital noise reduction method, referring to Figure 1 , Figure 1 is a flowchart of an EBAPS device image adaptive digital noise reduction method provided in at least one embodiment of the present application. It should be understood that the method may further include additional boxes not shown and / or the boxes shown may be omitted, and the scope of the present application is not limited in this regard. The method includes:

[0061] First, set the initial parameters of the algorithm. The initial parameters include the sub-region image matrix dimension parameter N s , the outlier threshold ratio coefficient α, and the sub-region image filtering matrix dimension parameter N t , where N t <N s ;

[0062] Then, according to the defined initial parameter N s , segment the acquired digital image into sub-regions such that the dimension of the sub-region image matrix is N s ×N s , and perform adaptive denoising algorithm processing on the sub-region images obtained after segmentation.

[0063] In some embodiments, the specific data processing process of the foregoing method includes the following steps:

[0064] (11) In each of the sub-region images, calculate the first average value and the first standard deviation of the gray values of the pixel points in the sub-region image respectively;

[0065] (12) According to the initialization parameter α, define the points within the numerical range centered on the first average value and α times the first standard deviation as outlier points, and perform zero assignment preprocessing;

[0066] (13) Within the range of the sub-region image, recalculate the non-outlier points to obtain the corresponding second average value and second standard deviation;

[0067] (14) According to the recalculated second average value and second standard deviation, generate a corresponding filtering matrix according to the initialization parameter N t , and its dimension is N t ×N t ;

[0068] (15) Use the generated N t ×Nt Filtering matrix, for N s ×N s sub-region images for filtering and denoising. For non-outlier points, the corresponding updated value is obtained by synthesizing the original value of the point and the values in the neighboring region according to the filtering matrix. For outlier points, zero-preprocessing is performed in step (2), so the updated value corresponding to the outlier is obtained by synthesizing the values in its neighboring region according to the filtering matrix;

[0069] (16) Other sub-region images are processed according to steps (11)-(15).

[0070] In some embodiments, for convenience of expression, symbols are defined as shown in Table 1.

[0071] Table 1

[0072]

[0073] The specific implementation is as follows:

[0074] First, set the initial parameters of the algorithm, the sub-region image matrix dimension parameter N s , the outlier threshold ratio coefficient α, the sub-region image filtering matrix dimension parameter N t , where N t <N s , N s and N t are odd numbers; if the matrix corresponding to the single-frame acquired image is A, and the corresponding number of rows and columns are N R and N C , then according to the preset parameter N s of the sub-region image dimension, the acquired image A can be divided into N' R rows, N' C columns, a total of N' R ×N' C sub-region images B (m,n) (0 ≤ m ≤ N' R -1, 0 ≤ n ≤ N' C -1), as shown in Equation (1), where the symbol represents the largest integer not exceeding the number .

[0075]

[0076] In the sub-region image B (m,n) (0 ≤ m ≤ N' R -1, 0 ≤ n ≤ N' C -1), calculate the characteristic parameters in the sub-region image according to the following steps, and implement the adaptive denoising algorithm:

[0077] (21) From the acquired image A and its corresponding element a (k,l) (0 ≤ k ≤ N R -1, 0 ≤ l ≤ N C -1), the sub-region image B is calculated (m,n) The corresponding element in the matrix As shown in Equation (2),

[0078]

[0079] According to the sub-region image matrix B (m,n) , calculate the first average value in this region and the corresponding first standard deviation As shown in Equation (3),

[0080]

[0081] (22) Preprocess the abnormal points in the sub-region image matrix B (m,n) . According to the preset parameter α of the abnormal points, if the value of the corresponding pixel point is not within the range of the first average value of this sub-region and α times the first standard deviation then it is defined as an abnormal point and preprocessed by setting it to zero. As shown in Equation (4), the preprocessed sub-region image matrix is denoted as B ( ' m,n) , and the corresponding element is denoted as

[0082]

[0083] (23) Recalculate the corresponding second average value for the preprocessed sub-region image matrix B ( ' m,n) and the corresponding second standard deviation As shown in Equation (5), where the parameter N ' represents the number of non-zero elements in the sub-region image matrix B s ' ( ' m,n) in is the number of non-zero elements.

[0084]

[0085] (24) According to the dimension parameter N of the sub-region image filtering matrix t , generate the filter matrix T for the sub-region image B ( ' m,n) , and filter the elements (m,n) in B ( ' m,n) in . The corresponding filter matrix T (m,n) in the elements The expression is as shown in Equation (6), and the corresponding filter image is as Figure 2 shown.

[0086]

[0087] (25) Using the generated filter matrix T (m,n) , filter the sub-region image matrix B ( ' m,n) to generate the sub-region image C after filtering, and its corresponding element is (m,n) as shown in Equation (7). The expression is as shown in Equation (7).

[0088]

[0089] (26) For any sub-region image B (m,n) (0 ≤ m ≤ N' R -1, 0 ≤ n ≤ N' C -1), repeat the above steps (21)-(25), and then the algorithm for adaptive digital noise reduction of the EBAPS device image proposed in this patent is completed.

[0090] To verify the effect of the algorithm proposed in this application, performance comparisons are made in terms of peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and algorithm execution time. The analysis results are shown in Table 2, where the execution time is represented in relative time form. Under the condition of a certain hardware configuration, the time required for the method proposed in this application to execute is taken as unit 1. In the performance comparison, the parameters adopted in this application are: N s = 301, N t = 25, α = 1.5.

[0091] Table 2:

[0092] Method Peak signal-to-noise ratio Structural similarity Execution time Reference [1] 38.1681 0.9745 0.9691 Reference [2] 37.8224 0.9696 0.9677 Reference [3] 38.0200 0.9740 0.9733 Reference [4] 37.8911 0.9829 0.9732 Reference [5] 38.0607 0.9832 0.9764 Reference patent 42.8901 0.9954 1.4124 This application 44.1023 0.9981 1

[0093] It can be seen from the performance comparisons of various methods that the method proposed in this application has a significant improvement in the performance indicators of peak signal-to-noise ratio and structural similarity compared with other reference methods. In terms of execution time, References [1]-[5] adopted a single-frame processing algorithm, and the time required is less than that of the method proposed in this application, about 97% of the time of this application, and the processing speed is slightly improved. The method in the reference patent adopted a multi-frame data processing method, so compared with the single-frame data processing method, the required time is significantly increased. Through comprehensive comparison of various methods, the method proposed in this application has obvious comprehensive advantages, thus verifying the rationality and applicability of this method, and can effectively enhance the noise reduction effect of the EBAPS device image and improve the image quality.

[0094] In summary, the present application adopts a sub-region adaptive digital noise reduction algorithm to perform anomaly processing on abnormal pixel points in a sub-region. Within this sub-region, characteristic analysis and parameter extraction are carried out based on the characteristics of the image, and the corresponding filtering matrix kernel is calculated adaptively, which can effectively alleviate the deficiencies existing in the current algorithms, namely: (1) calculating the corresponding adaptive threshold according to the noise characteristics of the entire image, while ignoring the difference between the global noise distribution and the local noise distribution of the image, and as the difference in noise distribution increases, the image denoising performance is greatly reduced; (2) adopting a local denoising method, but it requires approximate prior data on the local distribution of image noise, and with the difference in the images collected at different times, large deviations occur in the local prior data, seriously affecting the image quality. The present application can perform single-frame data processing to make up for the impact of the difference in regional noise distribution on the image quality, and can also automatically update the filtering kernel matrix as the image frames collected are different, alleviating the impact of the noise distribution of different frames on the image quality, and automatically calculating the matrix corresponding to the filtering kernel according to the regional statistical characteristics without the need for prior data on the noise distribution, having strong robustness and adaptability.

[0095] The present application reduces the impact of the difference in noise distribution between single frames on the image quality by automatically calculating the single-frame filtering kernel matrix; reduces the impact of the difference in noise distribution between multiple frames on the image quality by automatically calculating the multi-frame filtering kernel matrix; the algorithm of the present application is processed frame by frame, having a relatively fast processing speed; while taking into account the difference in single-frame noise distribution and the difference in multi-frame noise distribution, having good performance. Therefore, from the perspective of processing speed (efficiency) and processing performance, the method of the present application has good comprehensive performance, rather than the existing methods that unilaterally pursue one of the indicators.

[0096] The present application uses a single-frame image for data processing. Before performing local filtering, it is necessary to perform multiple pre-processings and parameter calculations on the data in this region. The required execution time is slightly longer than that of the current single-frame image processing algorithm, but the overall performance of the algorithm is greatly improved, having better comprehensive performance.

[0097] Figure 3 It is a module block diagram of an EBAPS device image adaptive digital noise reduction device provided by at least one embodiment of the present application. The device includes:

[0098] A memory 101; and a processor 102 connected to the memory 101, the processor 102 being configured to: set initial parameters, the initial parameters including the sub-region image matrix dimension parameter N s , the anomaly point threshold ratio coefficient α, the sub-region image filtering matrix dimension parameter N t , where N t <N s ;

[0099] According to the defined initial parameter N s , the acquired digital image is subjected to sub-region segmentation so that the dimension of the sub-region image matrix is N s ×N s , and an adaptive denoising algorithm is implemented for processing in the sub-region image obtained after segmentation.

[0100] In some embodiments, the processor 102 is further configured to: the implementation of the adaptive denoising algorithm for processing in the sub-region image obtained after segmentation specifically includes:

[0101] In each of the sub-region images, calculate the first average value and the first standard deviation of the gray values of the pixel points in the sub-region image respectively;

[0102] According to the initialization parameter α, the points located within the numerical range of α times the first standard deviation centered on the first average value are defined as abnormal points and are preprocessed by setting them to zero;

[0103] Within the range of the sub-region image, the non-abnormal points are recalculated to obtain the corresponding second average value and second standard deviation;

[0104] According to the recalculated second average value and second standard deviation, according to the initialization parameter N t , generate a corresponding filtering matrix with a dimension of N t ×N t ;

[0105] Using the generated N t ×N t filtering matrix, perform filtering denoising on the sub-region image of N s ×N s . For non-abnormal points, the corresponding updated value is obtained by synthesizing the original value of the point and the values in the neighboring regions according to the filtering matrix. For abnormal points, the corresponding updated value is obtained by synthesizing the values in the neighboring regions according to the filtering matrix.

[0106] In some embodiments, the processor 102 is further configured to: the sub-region segmentation of the acquired digital image specifically includes:

[0107] The matrix corresponding to the single-frame acquired image is A, and the corresponding number of rows and columns is N R and N C , according to the preset parameter N of the sub-region image dimension s , divide the acquired image A into N' R rows, N' C columns, for a total of N' R ×N' C sub-region images B (m,n)(0 ≤ m ≤ N' R -1, 0 ≤ n ≤ N' C -1):

[0108]

[0109] where the symbol denotes the largest integer not exceeding the number .

[0110] In some embodiments, the processor 102 is further configured to: in each of the sub-region images, calculate the first average value and the first standard deviation of the gray values of the pixel points in the sub-region image, specifically including:

[0111] From the acquired image A and its corresponding element a (k,l) (0 ≤ k ≤ N R -1, 0 ≤ l ≤ N C -1), calculate the corresponding element in the sub-region image B (m,n) matrix

[0112]

[0113] According to the sub-region image matrix B (m,n) , calculate the first average value and the corresponding first standard deviation

[0114]

[0115] In some embodiments, the processor 102 is further configured to: according to the initialization parameter α, define the points within the numerical range centered on the first average value and α times the first standard deviation as abnormal points, and perform zero-assignment preprocessing, specifically including:

[0116] Perform preprocessing on the abnormal points in the sub-region image matrix B (m,n) . According to the abnormal point preset parameter α, if the value of the corresponding pixel point is not within the range of the first average value of the sub-region and α times the first standard deviation , then it is defined as an abnormal point and zero-assignment preprocessing is performed. The preprocessed sub-region image matrix is denoted as B ( ' m,n) , and the corresponding element is denoted as

[0117]

[0118] In some embodiments, the processor 102 is further configured to: within the sub-region image range, recalculate the non-abnormal points to obtain the corresponding second average value and second standard deviation, specifically including:

[0119] Recalculate the corresponding second bottle mean value for the preprocessed sub-region image matrix B ( ' m,n) Recalculate the corresponding second bottle mean value And the corresponding second standard deviation Where the parameter N s ' represents the number of non-zero elements in the sub-region image matrix B ( ' m,n) In, the element That is not zero:

[0120]

[0121] In some embodiments, the processor 102 is further configured to: according to the recalculated second average value and second standard deviation, and according to the initialization parameter N t , generate the corresponding filtering matrix, specifically including:

[0122] Generate the filter matrix T for the sub-region image B according to the sub-region image filtering matrix dimension parameter N t , and filter the elements in B ( ' m,n) The corresponding filtering matrix T (m,n) , for B ( ' m,n) In the element Perform filtering, and the elements in the corresponding filtering matrix T (m,n) Expression: Expression:

[0123]

[0124] In some embodiments, the processor 102 is further configured to: use the generated N t ×N t Filtering matrix to perform filtering and denoising on the sub-region image of N s ×N s , specifically including:

[0125] Use the generated filtering matrix T (m,n) , to filter the sub-region image matrix B ( ' m,n) To generate the sub-region image C after filtering, and its corresponding elements are (m,n) , its corresponding elements are

[0126]

[0127] For the specific implementation method, please refer to the foregoing method embodiments, which will not be elaborated herein.

[0128] This application can be a method, an apparatus, a system, and / or a computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions thereon for performing various aspects of this application.

[0129] A computer-readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disc (DVD), a memory stick, a floppy disk, a mechanically encoded device such as a punched card or raised structures in grooves having instructions stored thereon, and any suitable combination of the foregoing. The computer-readable storage medium used herein is not construed as an instantaneous signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated through a waveguide or other transmission medium (e.g., an optical pulse through an optical fiber cable), or an electrical signal transmitted through a wire.

[0130] The computer-readable program instructions described herein can be downloaded from the computer-readable storage medium to various computing / processing devices, or downloaded to an external computer or external storage device through a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network may include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0131] The computer program instructions for performing the operations of the present application may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine - related instructions, microcode, firmware instructions, state - setting data, or source code or object code written in any combination of one or more programming languages, including object - oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as the "C" language or similar programming languages. The computer - readable program instructions may be executed entirely on the user's computer, partially on the user's computer, executed as a stand - alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider). In some embodiments, by using the state information of the computer - readable program instructions to customize an electronic circuit, such as a programmable logic circuit, a field - programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit can execute the computer - readable program instructions to implement various aspects of the present application.

[0132] Aspects of the present application are described herein with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer - readable program instructions.

[0133] These computer - readable program instructions can be provided to a processing unit of a general - purpose computer, a special - purpose computer, or other programmable data - processing apparatus to produce a machine such that when these instructions are executed by the processing unit of the computer or other programmable data - processing apparatus, a device is produced that implements the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer - readable program instructions can also be stored in a computer - readable storage medium, which causes a computer, a programmable data - processing apparatus, and / or other devices to operate in a specific manner. Thus, the computer - readable medium storing the instructions includes a manufacture that includes instructions for implementing various aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0134] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices, causing a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other devices to generate a computer-implemented process, so that the instructions executed on the computer, other programmable data processing apparatus, or other devices implement the functions / actions specified in one or more boxes of the flowchart and / or block diagram.

[0135] The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram may represent a module, a segment of a program, or a part of an instruction, which contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions noted in the boxes may occur in a different order than noted in the figures. For example, two consecutive boxes may actually be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and combinations of boxes in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0136] Note that unless otherwise directly stated, all features disclosed in this specification (including any appended claims, abstract, and drawings) may be replaced by alternative features serving the same, equivalent, or similar purpose. Therefore, unless otherwise clearly stated, each feature disclosed is only an example of a group of equivalent or similar features. When used, further, preferably, furthermore, and more preferably are simply the starting points for elaborating another embodiment based on the foregoing embodiments. The content following the further, preferably, furthermore, or more preferably, in combination with the foregoing embodiments, constitutes the complete composition of another embodiment. Combinations can be arbitrarily made among several further, preferably, furthermore, or more preferably settings following the same embodiment to form another embodiment.

[0137] Although the present application has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it on the basis of the present application, which will be obvious to those skilled in the art. Therefore, these modifications or improvements made without departing from the spirit of the present application fall within the scope of protection required by the present application.

Claims

1. An EBAPS device image adaptive digital noise reduction method, characterized in that: The following steps are involved: Set initial parameters, including sub-region image matrix dimension parameter N s , outlier threshold ratio coefficient α, sub-region image filter matrix dimension parameter N t , where N t <N s ; According to the defined initial parameter N s , the collected digital image is divided into sub-regions, so that the dimension of the sub-region image matrix is ​​N s ×N s , an adaptive denoising algorithm is implemented in the sub-region image obtained after segmentation.

2. The EBAPS device image adaptive digital noise reduction method according to claim 1, characterized in that: The step of implementing an adaptive denoising algorithm in the sub-region image obtained after segmentation for processing specifically includes: In each of the sub-region images, respectively calculating a first average value and a first standard deviation of the grayscale values ​​of the pixels of the sub-region image; According to the initialization parameter α, points within a numerical range of α times the first standard deviation centered on the first average value are defined as abnormal points, and are pre-processed by assigning zeros to them; Within the sub-region image range, non-abnormal points are recalculated to obtain corresponding second average values ​​and second standard deviations; According to the recalculated second mean and second standard deviation, according to the initialization parameter N t , generate the corresponding filter matrix, whose dimension is N t ×N t ; Using the generated N t ×N t Filter matrix, for N s ×N s The sub-region image is filtered and denoised. For non-abnormal points, the corresponding updated value is synthesized by the original value of the point and the value of the adjacent region according to the filter matrix. For abnormal points, the corresponding updated value is synthesized by the value of its adjacent region according to the filter matrix.

3. The EBAPS device image adaptive digital noise reduction method according to claim 2, characterized in that: The sub-region segmentation of the acquired digital image specifically includes: The matrix corresponding to a single frame of captured image is A, and the number of rows and columns is N R and N C , according to the sub-region image dimension preset parameter N s , divide the collected image A into N' R Line, N' C Column, total N' R ×N' C Sub-region image B (m,n) (0≤m≤N' R -1,0≤n≤N' C -1): The symbol Indicates no more than the number The maximum integer.

4. The EBAPS device image adaptive digital noise reduction method according to claim 3, characterized in that: The step of calculating the first average value and the first standard deviation of the grayscale values ​​of the pixels of each sub-region image respectively includes: By collecting the image A and its corresponding element a (k,l) (0≤k≤N R -1,0≤l≤N C -1), and the sub-region image B is calculated (m,n) Corresponding elements in the matrix According to the sub-region image matrix B (m,n) , calculate the first mean value in this sub-region and the corresponding first standard deviation 5. The EBAPS device image adaptive digital noise reduction method according to claim 4, characterized in that: The method of defining points within a numerical range of α times the first standard deviation centered on the first average value as abnormal points according to the initialization parameter α and performing zeroing preprocessing specifically includes: Sub-region image matrix B (m,n) The outliers are preprocessed, and the parameter α is preset according to the outliers. If the corresponding pixel The value is not within the range of the first mean value and α times the first standard deviation of the sub-region It is defined as an outlier point and preprocessed by assigning zeros. The sub-region image matrix after preprocessing is recorded as B′ (m,n) , the corresponding elements are denoted as 6. The EBAPS device image adaptive digital noise reduction method according to claim 5, characterized in that: The method of recalculating the non-abnormal points within the sub-region image range to obtain the corresponding second mean value and second standard deviation specifically includes: The preprocessed sub-region image matrix B′ (m,n) Recalculate the corresponding second bottle mean and the corresponding second standard deviation The parameter N′ s Represents the sub-region image matrix B′ (m,n) middle, yuan Non-zero number:

7. The EBAPS device image adaptive digital noise reduction method according to claim 6, characterized in that: The second mean value and the second standard deviation are recalculated according to the initialization parameter N t , generate the corresponding filter matrix, including: According to the sub-region image filter matrix dimension parameter N t , generate the sub-region image B′ (m,n) Filter matrix T (m,n) , for B′ (m,n) Zhongyuan Filtering is performed, corresponding to the filter matrix T (m,n) Medium Element expression:

8. The EBAPS device image adaptive digital noise reduction method according to claim 7, characterized in that: The utilization of the generated N t ×N t Filter matrix, for N s ×N s The sub-region image is filtered and denoised, including: Using the generated filter matrix T (m,n) , for the sub-region image matrix B′ (m,n) Filter and generate the sub-region image C after the filter (m,n) , whose corresponding elements are 9. An EBAPS device image adaptive digital noise reduction device, characterized in that: include: Memory; as well as A processor connected to the memory, the processor being configured to perform the steps of the method according to any one of claims 1 to 8.

10. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a machine, the steps of the method according to any one of claims 1 to 8 are implemented.

Images