A monitoring image enhancement method and system based on machine vision

Abstract

The present application relates to the technical field of image data processing, and more particularly to a monitoring image enhancement method and system based on machine vision. The method comprises the following steps: taking a pixel point in a monitoring video stream image frame as the center to obtain the local range of the pixel point; obtaining the noise index of the pixel point according to the distance and the gray difference between the pixel point and other pixel points in the corresponding local range, thereby calculating the noise degree of the image frame; obtaining a reference image frame set of the image frame, and performing normalization processing on the product of the noise degree standard deviation of the reference image frame set of the image frame and the noise degree of the image frame, thereby obtaining the actual noise degree of the image frame; calculating the window radius of the image frame; using the window radius of each image frame in the guided filter algorithm to perform denoising, and performing image enhancement based on the denoised image frame, thereby effectively improving the accuracy of monitoring image enhancement.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing, and in particular to a monitoring image enhancement method and system based on machine vision. BACKGROUND

[0002] The monitoring image based on machine vision is a process of extracting useful information from the video or image captured by the monitoring camera using advanced image processing technology and algorithm. These technologies can automatically analyze, identify, track and respond to targets or events in the monitoring scene.

[0003] In order to improve the practicability and accuracy of the monitoring image based on machine vision, the prior art provides many monitoring image processing methods based on machine vision. For example, the patent application file with publication number CN118154635A discloses a monitoring image processing method and system based on machine vision. The application obtains a monitoring image and performs denoising processing; determines a background image and a target image based on the monitoring image to obtain a difference image; performs image segmentation on the monitoring image by combining the region growing method and the difference method to obtain a clear foreground image; and calculates the centroid position coordinates of the monitoring target according to the pixel distribution of the foreground image using the centroid detection algorithm, and performs subsequent processing.

[0004] The above-mentioned prior art mainly solves the technical problem of processing the denoised monitoring image, so that the influence of background interference can be reduced when the background is complex and changeable. However, the monitoring image is often affected by environmental light, camera quality, transmission loss and other factors, resulting in insufficient image clarity. If the image segmentation processing is directly performed without processing, the accuracy of the obtained monitoring detection result will be low.

[0005] Therefore, how to accurately implement the monitoring image enhancement based on machine vision is a problem to be solved by those skilled in the art. SUMMARY

[0006] In order to solve the technical problem of how to accurately implement the monitoring image enhancement based on machine vision, the present application provides a monitoring image enhancement method and system based on machine vision.

[0007] In a first aspect, the present application provides a monitoring image enhancement method based on machine vision, which adopts the following technical solution:

[0008] A monitoring image enhancement method based on machine vision, comprising the steps of:

[0009] The local range of the pixel point is obtained by taking the pixel point in the image frame of the monitoring video stream as the center; the noise index of the pixel point is obtained according to the distance and the gray difference between the pixel point and other pixel points in the corresponding local range; the negative number of the standard deviation of the noise indexes of all pixel points in the image frame is taken as the index of the exponential function with e as the base, to obtain the second index of the image frame; the product of the second index and the average of the noise indexes of all pixel points in the image frame is taken as the noise degree of the image frame; the reference image frame set of the image frame is obtained, and the product of the standard deviation of the noise degree of the reference image frame set of the image frame and the noise degree of the image frame is normalized to obtain the actual noise degree of the image frame; the window radius of the image frame is calculated, and the window radius is positively correlated with the actual noise degree; the window radius of each image frame is used in the guided filtering algorithm to remove noise, and image enhancement is performed based on the denoised image frame.

[0010] When the monitoring image enhancement based on machine vision is implemented, the image quality of the image frame can be improved by the histogram equalization algorithm to enhance the image quality of the image frame, and the subsequent detection is facilitated. In this process, the present application considers that the noise in the image frame will affect the enhancement effect of the monitoring picture; based on this, the present application obtains the gray difference of each pixel point in the local range, and combines the noise similarity degree between the image frames, to finally obtain the actual noise degree of the image frame. Based on the actual noise degree of the image frame, the filtering and denoising of the image frame can be accurately realized, thereby effectively improving the accuracy of the monitoring image enhancement.

[0011] According to the monitoring image enhancement method based on machine vision provided by the present application, the preset local range of the pixel point in the image frame of the monitoring video stream is obtained, and the method further comprises the following steps: collecting the monitoring video stream by the monitoring device and performing gray scale binary processing to obtain the image frame in the monitoring video stream and the time stamp corresponding to the image frame; obtaining the local range of each pixel point in the image frame by the size of the preset local range.

[0012] The present application considers that the monitoring image data is relatively large, and therefore the complexity of the image data can be reduced by performing gray scale binary processing on the collected monitoring image, the calculation and storage resources required for storing and processing the image are reduced, and the running speed of the image processing algorithm is improved.

[0013] According to the monitoring image enhancement method based on machine vision provided by the present application, the noise index of the pixel point is obtained by the following method: the absolute value of the gray difference between the pixel point and other pixel points in the corresponding local range is obtained as the first index, and the reciprocal of the Euclidean distance between the pixel point and other pixel points in the corresponding local range is taken as the weight of the first index for weighted average, to obtain the noise index of the pixel point.

[0014] The present invention provides a precise method for calculating the noise level of an image frame. By analyzing the grayscale differences of pixels in a local range in an image frame and the overall grayscale fluctuations of pixels in the image frame, the noise level of the image frame can be accurately obtained, thus preparing for the subsequent adjustment of the window radius of the image frame.

[0015] According to a machine vision-based surveillance image enhancement method provided by the present invention, the reference image frame set acquisition method of the image frame includes: forming the image frames into a time sequence based on the timestamps of the image frames, and presetting the size and step size of the sliding window; taking any image frame as a target image frame, obtaining the absolute mean of the noise level differences between the target image frame and the image frames in the sliding window where the target image frame is located, and forming the reference image frame set of the target image frame from the image frames in the sliding window corresponding to the minimum value of the absolute mean of all noise level differences.

[0016] The present invention takes into account the prominence of noise points. The higher the similarity in noise levels between an image frame and adjacent frames, the smaller the overall difference, and the less likely the image frame is to contain noise points. Therefore, by obtaining the noise similarity between an image frame and its reference set of frames, the actual noise level of the image frame can be accurately determined.

[0017] According to a machine vision-based surveillance image enhancement method provided by the present invention, a window radius acquisition method of an image frame includes: ; is the window radius of the i-th image frame, is the preset reference window radius, is the actual noise level of the i-th image frame.

[0018] According to a machine vision-based surveillance image enhancement method provided by the present invention, the window radius of each image frame is used for denoising in a guided filtering algorithm, including: obtaining a guiding image of the image frame; and obtaining a filtered output image based on the window radius of the image frame, the linear coefficient in the window, and the guiding image.

[0019] The present invention removes noise in the surveillance image by guiding the filtering algorithm, which can retain the original details of the image as much as possible while removing the noise, so that the surveillance image can be accurately enhanced based on this and the interference of noise on the enhancement effect can be reduced.

[0020] According to a machine vision-based monitoring image enhancement method provided by the present invention, the image enhancement based on the denoised image frame includes: performing image enhancement on the denoised image frame through a histogram equalization algorithm.

[0021] According to the method, the image frames after denoising are subjected to image enhancement, and then the enhanced image frames are detected.

[0022] In a second aspect, the application provides a monitoring image enhancement system based on machine vision, which adopts the following technical scheme:

[0023] The monitoring image enhancement system based on machine vision comprises a processor and a memory, and the memory stores computer program instructions, which, when executed by the processor, implement the monitoring image enhancement method based on machine vision.

[0024] The monitoring image enhancement method based on machine vision is generated into a computer program and stored in the memory to be loaded and executed by the processor, so that a terminal device is manufactured according to the memory and the processor, and use is facilitated.

[0025] The application has the following technical effects:

[0026] Based on the above technical scheme, when the monitoring image enhancement based on machine vision is implemented, the image frames in the monitoring video stream are subjected to quality enhancement through the histogram equalization algorithm, so that the quality of the image frames is improved and subsequent detection is facilitated. In this process, the application considers that the noise in the image frames will affect the enhancement effect of the monitoring image, and based on this, the application obtains the gray level difference of each pixel point in the image frames in a local range, and combines the noise similarity degree between the image frames to finally obtain the actual noise degree of the image frames. Based on the actual noise degree of the image frames, the filtering and denoising of the image frames can be accurately implemented, so that the accuracy of the monitoring image enhancement is effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0027] The above and other objects, features and advantages of the present application will become readily apparent by reference to the following detailed description, when taken in conjunction with the accompanying drawings, in which several embodiments of the present application are shown. In the drawings, like reference numerals refer to like elements throughout.

[0028] Figure 1 A flowchart of a monitoring image enhancement method based on machine vision provided by the application is shown. DETAILED DESCRIPTION

[0029] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below, obviously, the described embodiments are some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work belong to the scope of protection of the present application.

[0030] It should be understood that when the claims, the specification and the drawings of the present application use the terms "first", "second" and the like, they are only used to distinguish different objects, and are not used to describe a specific sequence. The terms "include" and "contain" used in the specification and claims of the present application indicate the presence of the described features, whole, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, whole, steps, operations, elements, components and / or sets thereof.

[0031] The monitoring image based on machine vision is a process of extracting useful information from the video or image captured by the monitoring camera using advanced image processing techniques and algorithms. These techniques can automatically analyze, identify, track and respond to targets or events in the monitoring scene.

[0032] However, the monitoring image is often affected by environmental light, camera quality, transmission loss and other factors, resulting in insufficient image clarity. If the monitoring image is not processed directly, the accuracy of the monitoring detection result obtained will be low.

[0033] The histogram equalization algorithm is a commonly used image enhancement algorithm, which can increase the overall contrast of the image by redistributing the gray levels of the image pixels, so that the subtle differences in the image are more obvious, thereby improving the visual quality of the image.

[0034] Based on this, the embodiment of the present application discloses a monitoring image enhancement method based on machine vision, please refer to Figure 1 shown, Figure 1 A flowchart of a monitoring image enhancement method based on machine vision provided by the embodiment of the present application is shown in the figure. The method processes the image frames in the monitoring video stream by the histogram equalization algorithm, which can improve the overall quality of the image frames, so that the subsequent detection of the monitoring video can be accurately realized based on it. The specific steps include the following steps.

[0035] S1: Obtain the pixel points in the image frames of the monitoring video stream, and obtain the local range of the pixel points with the pixel points as the center.

[0036] The size of the local range of the pixel points can be set to 7x7, and the size of the local range can be set according to actual needs, which is not limited in the embodiment of the present application.

[0037] It can be understood that the local range of the pixel point is constructed with the pixel point as the center, and thus the obtained local range is composed of the pixel point itself and other pixel points except the pixel point.

[0038] For example, in the embodiment of the application, the preset local range of the pixel point in the monitoring video stream image frame is obtained with the pixel point as the center, and before that, it further includes: collecting a monitoring video stream through a monitoring device and performing gray scale binarization processing to obtain image frames in the monitoring video stream and time stamps corresponding to the image frames; and obtaining the local range of each pixel point in the image frames according to the size of the preset local range.

[0039] Specifically, when the monitoring device collects the monitoring video stream, the frame rate of the monitoring device can be set to 30fps, and the collection duration can be 3 seconds, and the number of image frames collected by the monitoring device is 90.

[0040] It should be noted that the frame rate and the collection duration can be set according to actual needs, and the embodiment of the application does not make too many limitations here.

[0041] It should be noted that the monitoring device will be disturbed by the surrounding environment during the process of shooting the monitoring video, resulting in the existence of noise points in the monitoring image obtained by shooting. The existence of noise points may interfere with the effect of histogram equalization, so that the enhanced image frame still contains noise, and even may enhance the noise in the opposite direction, reducing the effect and accuracy of image frame enhancement. Since the image frames in the monitoring video stream contain more details, a filtering algorithm is needed that can preserve image details while denoising.

[0042] The guided filtering algorithm is a filtering algorithm that can preserve edges. The algorithm filters the input image through a guide image, so that the final output image is generally similar to the input image, but the texture part is similar to the guide image. This filtering method can preserve edge and texture information in the image while smoothing out noise.

[0043] When filtering, the window radius needs to be appropriately increased in the area with more noise points to enhance the filtering effect; and the window radius needs to be appropriately reduced in the area with fewer noise points to avoid excessive smoothing, which can more effectively remove noise while maintaining the clarity of the image.

[0044] The selection of the window radius determines the filtering effect of the image frame. Based on this, the embodiment of the application first obtains the window radius of the image frame by calculating the noise degree of the image frame, and then enhances the image frame after denoising.

[0045] S2: Obtain the noise index of the pixel point according to the distance and gray scale difference between the pixel point and other pixel points in the corresponding local range, and obtain the noise degree of the image frame based on the noise index of each pixel point in the image frame.

[0046] It should be noted that the dust interference and other factors of the surrounding environment of the monitoring device can cause the noise points in the monitoring image obtained by shooting, and the gray scale of the noise points in the local range deviates from the original change characteristics, so that the difference between the gray scale of the noise points and the gray scale of the surrounding pixels is large. In addition, the gray scale of the edge position pixel in the image frame can be greatly different from the gray scale of the non-edge position, and therefore the deviation degree cannot be represented only by the gray scale difference between the pixel and other pixels in the local range. The edge position pixel in the image frame is a continuous pixel, and the noise point is usually isolated and irrelevant to the change of the surrounding pixels.

[0047] Based on this, in the embodiment of the application, the deviation degree between the pixel and other data points in the local range of the pixel is obtained by obtaining the gray scale difference between the pixel and other pixels in the local range of the pixel and the distance between the pixels, and the greater the gray scale deviation degree, the higher the possibility of the pixel being a noise point.

[0048] For example, in the embodiment of the application, when the noise index of the pixel in the image frame is obtained, the gray scale difference absolute value between the pixel and other pixels in the corresponding local range can be obtained as the first index, and the reciprocal of the Euclidean distance between the pixel and other pixels in the corresponding local range is weighted and averaged as the weight of the first index to obtain the noise index of the pixel.

[0049] It can be understood that the other pixels in the local range corresponding to the current pixel are all the remaining pixels in the local range except the current pixel.

[0050] For example, in the embodiment of the application, the noise index of the pixel is determined, and the specific process can be referred to the following relationship:

[0051] ;

[0052] the noise index of the i-th pixel, the noise index of the i-th pixel, the size of the local range of the i-th pixel, the other pixels in the local range of the i-th pixel except the i-th pixel, the gray scale difference absolute value between the i-th pixel and the j-th other pixel in the local range, the Euclidean distance between the i-th pixel and the j-th other pixel in the local range.

[0053] ​​​​​​​In the above formula, is the first index of the first pixel point and the first other pixel point in the local range, the greater the first index, the greater the gray difference between the first pixel point and the first other pixel point in the local range, the greater the deviation degree of the first pixel point, and the higher the possibility of the first pixel point being a noise point. is the weight of the first index, that is, the reciprocal of the Euclidean distance between the first pixel point and the first other pixel point in the local range, the smaller the Euclidean distance between the first pixel point and the first other pixel point in the local range, the greater the weight of the first other pixel point on determining the deviation degree of the first pixel point.

[0054]

[0055]

[0056] Based on the above formula, the noise index of each pixel point in the image frame can be obtained, and the noise degree of the image frame can be obtained according to the noise index of each pixel point in the image frame, that is, the following step is continued to be executed.

[0057] It needs to be further explained that the guided filter algorithm will focus on obtaining the number of noise points in the image frame when filtering. The greater the average of the noise index of each pixel point in the image frame, the higher the noise degree of the image frame. However, the high noise degree of the image frame can be caused by individual outstanding pixel points or the high average of the overall noise index of the pixel points. The noise degree of the image frame obtained directly based on the average of the noise index of each pixel point in the image frame cannot determine the source of the noise, and thus cannot determine the number of noise points in the image frame.

[0058] ​​​​​​​​​​​​​​​​​​​If the noise degree of the image frame is caused by multiple noise points, and each noise point is distributed in each area in the image frame, the noise degree of the image frame is higher, in order to improve the denoising effect, the window radius needs to be increased, but in fact, each noise point is distributed in each area in the image frame, and increasing the window radius will cause more normal pixels to be included in the filtering range, increasing the risk of false filtering of normal pixels, and reducing the image detail.

[0059] Therefore, in the embodiment of the present application, the standard deviation of the noise indicators of all the pixel points in the image frame is taken as the exponent of the exponential function with e as the base, to obtain a second indicator of the image frame; and the product of the second indicator and the average of the noise indicators of all the pixel points in the image frame is taken as the noise degree of the image frame.

[0060] For example, in the embodiment of the present application, when obtaining the noise degree of the image frame, the negative of the standard deviation of the noise indicators of all the pixel points in the image frame is taken as the exponent of the exponential function with e as the base, to obtain a second indicator of the image frame; and the product of the second indicator and the average of the noise indicators of all the pixel points in the image frame is taken as the noise degree of the image frame.

[0061] For example, the noise degree of the image frame is determined, and the specific method can be referred to the following relationship:

[0062] ;

[0063] the noise degree of the image frame, the average of the noise indicators of all the pixel points in the image frame, the standard deviation of the noise indicators of all the pixel points in the image frame, the exponential function with e as the base.

[0064] Based on the above steps, the noise degree of each image frame can be accurately obtained, and based on the noise of each image frame, the window radius when each image frame is guided filtering can be obtained, that is, the following steps are continued to be executed.

[0065] S3: obtaining a reference image frame set of the image frame, and performing normalization processing on the product of the standard deviation of the noise degree of the reference image frame set of the image frame and the noise degree of the image frame, to obtain the actual noise degree of the image frame.

[0066] It should be noted that the noise degree of each pixel point can be obtained based on the analysis of the pixel gray distribution in each image frame in the above steps. The noise points in the monitoring video stream appear randomly, that is, they are distributed discretely in each image frame. Therefore, the actual noise degree of the image frame is also reflected in the noise degree difference with its adjacent image frames. If the noise degree difference is larger, it means that the possibility of the image frame being disturbed by random noise points is larger, and the corresponding actual noise degree is also larger. If the noise degree difference is smaller, the possibility of the image frame being disturbed by random noise points is smaller.

[0067] Based on this, an embodiment of the present invention obtains a section of adjacent image frames of an image frame as a reference image frame set, and corrects the noise level of the image frame based on the similarity of the noise levels between the image frame and the reference image frame set to obtain the actual noise level of the image frame.

[0068] For example, in an embodiment of the present invention, obtaining a reference image frame set of an image frame includes the following two possible implementation methods:

[0069] In one possible implementation, the image frames can be organized into a time sequence based on the timestamps of the image frames, and the size and step size of the sliding window can be preset; any image frame is taken as the target image frame, and the absolute mean of the noise level differences between the target image frame and the image frames in the sliding window where the target image frame is located is obtained, and the image frames in the sliding window corresponding to the minimum value of the absolute mean of all noise level differences are used to form a reference image frame set for the target image frame.

[0070] Among them, the size of the sliding window can be set to 11, the step size of the sliding window can be set to 1, and the size of the sliding window is the number of image frames contained in the sliding window; the size and step size of the sliding window can be set according to actual needs, and the embodiment of the present invention does not impose too many restrictions here.

[0071] In another possible implementation, when determining the reference image frame set of the i-th image frame, the size of the reference image frame set may be preset. ; Get the i-th image frame and the i-th image frame respectively image frames and The absolute value of the noise level difference between the image frames is used, and the image frame corresponding to the smaller absolute value of the noise level difference is used as the first adjacent image frame of the i-th image frame; if the image frame corresponding to the smaller absolute value of the noise level difference is the i-th image frame image frames, then we can continue to obtain the i-th image frame and the image frames and The absolute value of the noise level difference between the image frames is obtained, and the image frame corresponding to the smaller absolute value of the noise level difference is used as the second adjacent image frame of the i-th image frame, and this cycle is repeated until the image frame is obtained. The adjacent image frames of the i-th image frame are adjacent image frames and the i-th image frame together constitute a reference image frame set of the i-th image frame.

[0072] The size of the reference image frame set may be set to 11; the size of the reference image frame set is the number of image frames included in the reference image frame set, and may be specifically set according to actual needs.

[0073] After obtaining the reference image frame set of the image frame based on any of the above possible implementations, the noise level of the image frame can be corrected based on the similarity of the noise levels between the image frame and the reference image frame set to obtain the actual noise level.

[0074] For example, in an embodiment of the present invention, the actual noise level of an image frame is determined, and specific reference may be made to the following relationship:

[0075] ;

[0076] is the actual noise level of the i-th image frame, is the noise level of the i-th image frame, is the standard deviation of the noise level of the reference image frame set of the i-th image frame, is the hyperbolic tangent function.

[0077] Among them, the hyperbolic tangent function is used for normalization processing.

[0078] In the above formula, the larger the standard deviation of the noise level of the reference image frame set of the i-th image frame, the lower the similarity of the noise level between the i-th image frame and its reference image frame set, the higher the noise performance level of the i-th image frame, and the higher the corresponding actual noise level.

[0079] After obtaining the actual noise level of each image frame based on the above steps, the window radius of each image frame can be adjusted based on the actual noise level of each image frame, thereby accurately implementing guided filtering.

[0080] S4: Calculate the window radius of the image frame; use the window radius of each image frame to perform denoising in the guided filtering algorithm, and perform image enhancement based on the denoised image frame.

[0081] It should be noted that the higher the actual noise level of the image frame, the higher the possibility of noise. The filtering effect can be enhanced by increasing the window radius. The higher the actual noise level, the higher the degree of enhancement. The window radius is positively correlated with the actual noise level.

[0082] For example, in the embodiment of the present invention, the window radius of the image frame may be determined by referring to the following relationship:

[0083] ;

[0084] is the window radius of the i-th image frame, is the preset reference window radius, is the actual noise level of the i-th image frame.

[0085] The preset reference window radius can be set as 4, and can be set according to actual needs, and the embodiment of the present application does not make too many limitations here.

[0086] For example, in the embodiment of the present application, the window radius of each image frame is used in the guided filter algorithm for denoising, including: obtaining a guided image of the image frame; obtaining a filter output image according to the window radius of the image frame, linear coefficients in the window, and the guided image.

[0087] The guided image of the image frame can be the image frame itself, that is, the input image. The linear coefficient is a constant in the window radius, which can be calculated by substituting the average value and variance of the pixels in the window radius of the guided image into the formula.

[0088] The specific steps of using the window radius of each image frame in the guided filter algorithm for filtering and denoising can be realized by formula transformation in the guided filter algorithm, and the embodiment of the present application does not make too many limitations here.

[0089] After the image frames in the monitoring video stream are filtered and denoised based on the above-mentioned manner, image enhancement can be performed based on the denoised image frames, so as to reduce the interference of noise on the enhancement effect.

[0090] For example, in the embodiment of the present application, the image enhancement is performed based on the denoised image frames, including: performing image enhancement on the denoised image frames by using a histogram equalization algorithm.

[0091] Specifically, the frequency distribution of each pixel value in the denoised image frame is calculated, a histogram is generated and normalized to obtain the probability density distribution of each gray level; the cumulative distribution function is calculated, and the cumulative distribution function is used as a mapping function to map the gray level of the original image to a new gray level to obtain an enhanced image frame.

[0092] The specific steps of performing image enhancement on the denoised image frames by using the histogram equalization algorithm can be realized by using the prior art, and the embodiment of the present application does not make too many limitations here.

[0093] It can be understood that after the image frames of the monitoring video stream are processed by the image enhancement algorithm, the clarity and detail of the image are significantly improved, and the accuracy of the detection result can be effectively improved.

[0094] For example, in the embodiment of the present application, the image enhancement is performed based on the denoised image frames, and then further includes: detecting the enhanced image frames.

[0095] The specific steps of detecting the enhanced image frames can be realized by using the prior art, and the embodiment of the present application does not make too many limitations here.

[0096] It can be seen that in the embodiment of the application, when the monitoring image based on machine vision is enhanced, the local range of a pixel point in the monitoring video stream image frame can be obtained with the pixel point as the center; the noise index of the pixel point is obtained according to the distance and the gray difference between the pixel point and other pixel points in the corresponding local range; the negative number of the standard deviation of the noise indexes of all pixel points in the image frame is taken as the index of the exponential function with e as the base, to obtain the second index of the image frame; the product of the second index and the average of the noise indexes of all pixel points in the image frame is taken as the noise degree of the image frame; the reference image frame set of the image frame is obtained, and the product of the standard deviation of the noise degrees of the reference image frame set of the image frame and the noise degree of the image frame is normalized to obtain the actual noise degree of the image frame. is the window radius of the i-th image frame, is the preset reference window radius, is the actual noise degree of the i-th image frame; the window radius of each image frame is used for denoising in the guided filtering algorithm, and image enhancement is performed based on the denoised image frame.

[0097] In this way, the image frame in the monitoring video stream is enhanced in quality by the histogram equalization algorithm, which can improve the quality of the image frame and facilitate subsequent detection. In this process, the application considers that the noise in the image frame will affect the enhancement effect of the monitoring picture; based on this, the application obtains the gray difference of each pixel point in the local range of the image frame, and combines the noise similarity degree between the image frames, to finally obtain the actual noise degree of the image frame. Based on the actual noise degree of the image frame, the filtering denoising of the image frame can be accurately realized, thereby effectively improving the accuracy of the monitoring image enhancement.

[0098] The embodiment of the application further discloses a monitoring image enhancement system based on machine vision, which comprises a processor and a memory, and the memory stores computer program instructions.

[0099] The above system further comprises a communication bus and a communication interface and other components familiar to those skilled in the art, the settings and functions of which are known in the art, and thus will not be described here.

[0100] ​In this disclosure, a "storage medium" or "computer readable medium" can be any available medium that can be accessed by a general purpose or special purpose computer system. By way of example, and not limitation, such computer readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store desired program code means in the form of computer-executable instructions or data structures and that can be accessed by a general purpose or special purpose computer system. Combinations of the above should also be included within the scope of the computer readable media.

[0101] While the present application has been illustrated and described in detail in the drawings and foregoing description, such illustration and description is to be considered illustrative or exemplary and not restrictive; the present application is not limited to the disclosed embodiments.

[0102] The above are only preferred embodiments of the present application, not to limit the protection scope of the present application, therefore: all equivalent changes made on the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A monitoring image enhancement method based on machine vision, characterized in that: include: The local range of a pixel point in the monitoring video stream image frame is obtained with the pixel point as the center; the noise index of the pixel point is obtained based on the distance and grayscale difference between the pixel point and other pixels in the corresponding local range; the method for obtaining the noise index of the pixel point includes: The absolute value of the grayscale difference between the pixel point and other pixels in the corresponding local range is obtained as the first indicator, and the reciprocal of the Euclidean distance between the pixel point and other pixels in the corresponding local range is used as the weight of the first indicator for weighted averaging to obtain the noise index of the pixel point; The negative of the standard deviation of the noise index of all pixels in the image frame is used as the exponent of an exponential function with base e to obtain a second index of the image frame; the product of the second index and the mean of the noise index of all pixels in the image frame is used as the noise level of the image frame; Acquire a reference image frame set of the image frame, and normalize the product of the noise level standard deviation of the reference image frame set and the noise level of the image frame to obtain the actual noise level of the image frame; The window radius of the image frame is calculated, and the window radius is positively correlated with the actual noise level; the window radius of each image frame is used in the guided filtering algorithm for denoising, and image enhancement is performed based on the denoised image frame.

2. The method for enhancing surveillance images based on machine vision according to claim 1, characterized in that: The method of obtaining a preset local range of a pixel point in the monitoring video stream image frame as the center also includes: The monitoring video stream is collected by the monitoring equipment and grayscale binarization is performed to obtain the image frames in the monitoring video stream and the timestamps corresponding to the image frames; the local range of each pixel point in the image frame is obtained by presetting the size of the local range.

3. The method for enhancing surveillance images based on machine vision according to claim 2, characterized in that: The method for obtaining a reference image frame set of the image frame includes: The image frames are grouped into a time sequence based on their timestamps, and the size and step size of the sliding window are preset; any image frame is taken as the target image frame, and the absolute mean of the noise level differences between the target image frame and the image frames in the sliding window where the target image frame is located is obtained, and the image frames in the sliding window corresponding to the minimum value of the absolute mean of all noise level differences are used to form a reference image frame set for the target image frame.

4. The method for enhancing surveillance images based on machine vision according to claim 1, wherein: Methods for obtaining the window radius of an image frame include: ; is the window radius of the i-th image frame, is the preset reference window radius, is the actual noise level of the i-th image frame.

5. The method for enhancing surveillance images based on machine vision according to claim 1, wherein: The denoising using the window radius of each image frame in the guided filtering algorithm includes: A guide image of the image frame is obtained; and a filtered output image is obtained according to a window radius of the image frame, a linear coefficient in the window, and the guide image.

6. The method for enhancing surveillance images based on machine vision according to claim 1, characterized in that: The image enhancement based on the denoised image frame includes: The denoised image frames are enhanced using the histogram equalization algorithm.

7. The method for enhancing surveillance images based on machine vision according to claim 1, characterized in that: The image enhancement is performed based on the denoised image frame, and then further includes: Perform detection on the enhanced image frame.

8. A monitoring image enhancement system based on machine vision, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a monitoring image enhancement method based on machine vision according to any one of claims 1 to 7 is implemented.

Images