Smart vision super-photosensitive noise reduction method and system and medium

Abstract

The invention provides a smart vision super-photosensitive noise reduction method and system and a medium. The method comprises the following steps: acquiring an image sensor parameter value through an illumination intensity value corresponding to an image acquisition time point, acquiring a real-time image, then carrying out preprocessing to obtain an optimized image, carrying out processing according to the optimized image to obtain noise category feature data and pixel value statistical data, and carrying out processing according to the pixel value statistical data to obtain noise category feature data. Obtaining an adaptive noise reduction processing method according to the noise category feature data and the noise intensity evaluation index, performing noise reduction processing on the optimized image, obtaining a noise reduction image, obtaining and processing image quality parameter data of the noise reduction image, obtaining a noise reduction actual effect index, and finally performing threshold comparison. Adjusting an adaptive noise reduction processing method according to a threshold comparison result; the self-adaptive noise reduction processing method is obtained through image preprocessing and noise category and noise intensity recognition, and intelligent noise reduction of the intelligent visual super-photosensitive image is achieved.

Description

Technical Field

[0001] The present application relates to the field of visual image processing technology, and more specifically, to an intelligent visual super-sensitive noise reduction method, system and medium. Background Art

[0002] In the construction of smart construction sites, visual monitoring systems play a vital role. However, the construction site environment is complex and changeable, with poor lighting conditions and a large number of interference factors, such as dust and smoke. The images acquired by traditional visual acquisition equipment in this environment often have serious noise problems, which greatly affects the subsequent image analysis, target recognition and other processing processes, leading to misjudgments of safety hazards on the construction site, construction progress and other situations, and cannot meet the needs of smart construction sites for accurate visual information. Some existing noise reduction methods have poor noise reduction effects or excessive noise reduction leading to loss of image details when dealing with the ultra-sensitive requirements in complex environments such as smart construction sites.

[0003] In view of the above problems, effective technical solutions are urgently needed. Summary of the invention

[0004] The purpose of this application is to provide a smart vision super-sensitive noise reduction method, system and medium, which can obtain an adaptive noise reduction processing method through image preprocessing, identification of noise category and noise intensity, and realize intelligent noise reduction of smart vision super-sensitive images.

[0005] This application also provides a smart visual super-sensitivity noise reduction method, including the following steps: Obtaining the light intensity value corresponding to the image acquisition time point, obtaining the image sensor parameter value according to the light intensity value, acquiring the real-time image according to the image sensor parameter value, and preprocessing the real-time image to obtain an optimized image; Processing is performed according to the optimized image to obtain noise category characteristic data and pixel value statistical data, and processing is performed according to the pixel value statistical data to obtain a noise intensity evaluation index; Acquire an adaptive noise reduction processing method according to the noise category feature data and the noise intensity evaluation index, and perform noise reduction processing on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image; Acquire image quality parameter data of the noise reduction image, perform processing according to the image quality parameter data, and obtain a noise reduction effectiveness index; The noise reduction effectiveness index is compared with a preset noise reduction effectiveness evaluation threshold, and the adaptive noise reduction processing method is adjusted according to the threshold comparison result.

[0006] Optionally, in the intelligent visual super-sensitive noise reduction method described in the present application, obtaining the light intensity value corresponding to the image acquisition time point, obtaining the image sensor parameter value according to the light intensity value, acquiring the real-time image according to the image sensor parameter value, and preprocessing the real-time image to obtain the optimized image includes: Obtain the light intensity value corresponding to the image acquisition time point, and query the preset light parameter mapping table according to the light intensity value to obtain the image sensor parameter value, including acquisition frequency, exposure time, aperture size and gain value; Taking photos according to the acquisition frequency, exposure time, aperture size and gain value to obtain multiple frames of real-time images; The multiple frames of real-time images are preprocessed to obtain optimized images.

[0007] Optionally, in the intelligent visual super-sensitive noise reduction method described in the present application, the preprocessing of the multiple frames of real-time images to obtain optimized images includes: Extracting ORB descriptors corresponding to corner points respectively according to the multiple frames of real-time images; The matching feature points are obtained by matching the distances between the ORB descriptors of multiple frames of real-time images; Processing is performed based on the matching feature points to obtain a geometric transformation relationship, and processing is performed based on the geometric transformation relationship to obtain an affine transformation matrix; The multiple frames of real-time images are optimized respectively according to the affine transformation matrix to obtain optimized images.

[0008] Optionally, in the intelligent visual super-sensitive noise reduction method described in the present application, the processing according to the optimized image to obtain noise category feature data and pixel value statistics, and the processing according to the pixel value statistics to obtain the noise intensity evaluation index include: Inputting the optimized image into a preset image noise recognition model for processing to obtain noise category feature data; Obtaining pixel value statistics according to the optimized image, including pixel standard deviation, pixel variance and signal-to-noise ratio; The pixel standard deviation, pixel variance and signal-to-noise ratio are input into a preset noise intensity assessment model for processing to obtain a noise intensity assessment index.

[0009] Optionally, in the intelligent visual super-sensitive noise reduction method described in the present application, the adaptive noise reduction processing method is obtained according to the noise category feature data and the noise intensity evaluation index, and the optimized image is subjected to noise reduction processing according to the adaptive noise reduction processing method to obtain a noise-reduced image, including: Comparing the noise intensity evaluation index with a preset noise intensity standard evaluation index to obtain a relative value of noise intensity; Performing a threshold comparison between the relative value of the noise intensity and a preset noise level assessment threshold, and obtaining a noise level according to the threshold comparison result, including weak noise or strong noise; According to the noise category feature data combined with the noise level, a preset noise reduction method database is queried to obtain an adaptive noise reduction processing method, and noise reduction processing is performed on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image.

[0010] Optionally, in the smart visual super-sensitive noise reduction method described in the present application, the step of obtaining image quality parameter data of the noise reduction image and performing processing according to the image quality parameter data to obtain the noise reduction effectiveness index includes: Obtain image quality parameter data of the denoised image, including peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index; The peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index are input into a preset noise reduction processing effectiveness evaluation model for processing to obtain a noise reduction effectiveness index.

[0011] Optionally, in the intelligent visual super-sensing noise reduction method described in the present application, the noise reduction effectiveness index is compared with a preset noise reduction effectiveness evaluation threshold, and the adaptive noise reduction processing method is adjusted according to the threshold comparison result, including: Processing the noise reduction effectiveness index and the preset noise reduction demand index to obtain a noise reduction effectiveness deviation rate; Comparing the noise reduction effectiveness deviation rate with a preset noise reduction effectiveness deviation rate threshold; If it is less than or equal to the preset noise reduction effectiveness deviation rate threshold, the noise reduction effectiveness is in normal state; If it is greater than a preset noise reduction effectiveness deviation rate threshold, the noise reduction effectiveness is in an abnormal state, and the adaptive noise reduction processing method is adjusted.

[0012] Optionally, the intelligent visual super-sensing noise reduction method described in the present application further includes: Obtaining the gradient direction, gradient amplitude and direction interval of the denoised image, wherein the direction interval includes an edge direction interval and a texture direction interval, and the gradient amplitude includes an edge direction gradient amplitude and a texture direction gradient amplitude; If the gradient direction is within the edge direction interval and the edge direction gradient amplitude is less than a preset edge direction gradient amplitude threshold, querying a preset enhancement coefficient list to obtain an edge direction enhancement coefficient according to a gradient amplitude deviation rate between the edge direction gradient amplitude and the preset edge direction gradient amplitude threshold; Enhance the edge direction gradient amplitude according to the edge direction enhancement coefficient to obtain the edge direction gradient optimization amplitude; If the gradient direction is within the texture direction interval, obtaining the texture pixel value variance and the texture pixel value standard deviation within the texture direction interval; Inputting the texture pixel value variance and the texture pixel value standard deviation into a preset texture direction enhancement evaluation model for processing to obtain a texture direction enhancement coefficient; The texture direction gradient amplitude is enhanced according to the texture direction enhancement coefficient to obtain the texture direction gradient optimization amplitude.

[0013] In a second aspect, the present application provides an intelligent visual super-sensing noise reduction system, the system comprising: a memory and a processor, the memory comprising a program of an intelligent visual super-sensing noise reduction method, and the program of the intelligent visual super-sensing noise reduction method is executed by the processor to implement the following steps: Obtaining the light intensity value corresponding to the image acquisition time point, obtaining the image sensor parameter value according to the light intensity value, acquiring the real-time image according to the image sensor parameter value, and preprocessing the real-time image to obtain an optimized image; Processing is performed according to the optimized image to obtain noise category characteristic data and pixel value statistical data, and processing is performed according to the pixel value statistical data to obtain a noise intensity evaluation index; Acquire an adaptive noise reduction processing method according to the noise category feature data and the noise intensity evaluation index, and perform noise reduction processing on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image; Acquire image quality parameter data of the noise reduction image, perform processing according to the image quality parameter data, and obtain a noise reduction effectiveness index; The noise reduction effectiveness index is compared with a preset noise reduction effectiveness evaluation threshold, and the adaptive noise reduction processing method is adjusted according to the threshold comparison result.

[0014] In the third aspect, the present application also provides a computer-readable storage medium, which stores a smart visual super-sensitive noise reduction method program. When the smart visual super-sensitive noise reduction method program is executed by a processor, the steps of the smart visual super-sensitive noise reduction method as described in any one of the above items are implemented.

[0015] From the above, it can be seen that the intelligent vision super-sensitive noise reduction method, system and medium provided in this application realize intelligent noise reduction of intelligent vision super-sensitive images by obtaining an adaptive noise reduction processing method through image preprocessing, identification of noise categories and noise intensity.

[0016] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or understood by practicing the embodiments of the present application. The purpose and other advantages of the present application can be realized and obtained by the structures specifically pointed out in the written description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 A flow chart of a smart visual super-sensing noise reduction method provided in an embodiment of the present application; Figure 2 A flowchart of obtaining an optimized image by a smart visual super-sensitive noise reduction method provided in an embodiment of the present application; Figure 3 A flow chart of obtaining a noise reduction effectiveness index for an intelligent visual super-sensitive noise reduction method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0020] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0021] Please refer to Figure 1 , Figure 1 This is a flow chart of a smart visual super-sensing noise reduction method in some embodiments of the present application. This smart visual super-sensing noise reduction method is used in terminal devices, such as computers, mobile phone terminals, etc. This smart visual super-sensing noise reduction method includes the following steps: S11, obtaining a light intensity value corresponding to an image acquisition time point, obtaining an image sensor parameter value according to the light intensity value, acquiring a real-time image according to the image sensor parameter value, and preprocessing the real-time image to obtain an optimized image; S12, processing according to the optimized image to obtain noise category feature data and pixel value statistical data, and processing according to the pixel value statistical data to obtain a noise intensity evaluation index; S13, obtaining an adaptive noise reduction processing method according to the noise category feature data and the noise intensity evaluation index, and performing noise reduction processing on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image; S14, acquiring image quality parameter data of the noise reduction image, and performing processing according to the image quality parameter data to obtain a noise reduction effectiveness index; S15, performing a threshold comparison between the noise reduction effectiveness index and a preset noise reduction effectiveness evaluation threshold, and adjusting the adaptive noise reduction processing method according to the threshold comparison result.

[0022] It should be noted that in order to achieve effective noise reduction on the captured image, the light intensity value corresponding to the image capture time point is first obtained. Different light intensity values ​​correspond to different image sensor parameter values, including acquisition frequency, exposure time, aperture size and gain value. Image acquisition is implemented according to the queried image sensor parameter values ​​to obtain multiple frames of real-time images, and preprocessing is performed to obtain an optimized image. Then, the corresponding noise category feature data and pixel value statistical data are identified according to the optimized image, and the pixel value statistical data are processed to obtain a noise intensity evaluation index for expressing the noise intensity of the category of noise. The corresponding adaptive noise reduction processing method is obtained according to the noise category feature data and the noise intensity evaluation index, and the optimized image is subjected to noise reduction processing to obtain a noise reduction image. The image quality parameter data of the noise reduction image is extracted and processed to obtain a noise reduction effectiveness index. Finally, the adaptive noise reduction processing method is adjusted according to the threshold comparison result.

[0023] Please refer to Figure 2 , Figure 2 It is a flow chart of obtaining an optimized image in a smart visual super-sensitive noise reduction method in some embodiments of the present application. According to an embodiment of the present invention, obtaining the light intensity value corresponding to the image acquisition time point, obtaining the image sensor parameter value according to the light intensity value, acquiring a real-time image according to the image sensor parameter value, and preprocessing the real-time image to obtain an optimized image includes: S21, obtaining the light intensity value corresponding to the image acquisition time point, and querying the preset light parameter mapping table according to the light intensity value to obtain the image sensor parameter value, including acquisition frequency, exposure time, aperture size and gain value; S22, taking photos according to the acquisition frequency, exposure time, aperture size and gain value to obtain multiple frames of real-time images; S23, preprocessing the multiple frames of real-time images to obtain optimized images.

[0024] It should be noted that the lighting conditions at different image acquisition time points are different, and the required image sensor parameter values ​​are also different. In order to ensure the quality of the acquired image, the light intensity value corresponding to the image acquisition time point is first obtained, and the preset lighting parameter mapping table is queried according to the light intensity value to obtain the image sensor parameter values ​​including acquisition frequency, exposure time, aperture size and gain value. Among them, the preset lighting parameter mapping table is obtained by querying the preset intelligent vision super-sensitive noise reduction database, and then photos are taken according to the queried acquisition frequency, exposure time, aperture size and gain value to obtain multiple frames of real-time images, and the image is preprocessed to obtain an optimized image for subsequent noise reduction processing.

[0025] According to an embodiment of the present invention, the preprocessing of the multiple frames of real-time images to obtain optimized images includes: Extracting ORB descriptors corresponding to corner points respectively according to the multiple frames of real-time images; The matching feature points are obtained by matching the distances between the ORB descriptors of multiple frames of real-time images; Processing is performed based on the matching feature points to obtain a geometric transformation relationship, and processing is performed based on the geometric transformation relationship to obtain an affine transformation matrix; The multiple frames of real-time images are optimized respectively according to the affine transformation matrix to obtain optimized images.

[0026] It should be noted that, since there may be slight displacement of the device, jitter and other factors during the acquisition process that may cause the image to shift, it is necessary to first align the obtained multiple frames of real-time images, and use the FAST (Features from Accelerated Segment Test) corner detection algorithm to determine whether the point is a corner point. After the corner point is determined, the ORB (Oriented FAST and Rotated BRIEF) descriptor corresponding to the corner point is extracted. For example, for a 31*31 corner point neighborhood, n pairs of pixels are randomly selected, and the grayscale values ​​of each pair of pixels are compared. If the grayscale value of the first pixel is greater than the grayscale value of the second pixel, the corresponding binary bit is set to 1, otherwise it is set to 0. For two frames of images, matching is performed by calculating the distance between the descriptors. If n pairs of matching feature points are found in the two frames of images, those skilled in the art can represent their geometric transformation relationship according to actual needs, obtain an affine transformation matrix according to the geometric transformation relationship, and optimize the multiple frames of real-time images according to the obtained affine transformation matrix, including rotation, scaling, translation or shearing, to obtain optimized images.

[0027] According to an embodiment of the present invention, the step of processing the optimized image to obtain noise category feature data and pixel value statistics, and processing the pixel value statistics to obtain a noise intensity evaluation index includes: Inputting the optimized image into a preset image noise recognition model for processing to obtain noise category feature data; Obtaining pixel value statistics according to the optimized image, including pixel standard deviation, pixel variance and signal-to-noise ratio; The pixel standard deviation, pixel variance and signal-to-noise ratio are input into a preset noise intensity assessment model for processing to obtain a noise intensity assessment index.

[0028] It should be noted that different types and intensities of image noise use different noise reduction processing methods. In order to match the optimal noise reduction processing method, the optimized image is first input into the preset image noise recognition model for processing to obtain noise category feature data, wherein the image noise recognition model is obtained by obtaining a large number of historical sample optimized images and corresponding noise category feature data for training; then the pixel value statistics of the optimized image are obtained, including pixel standard deviation, pixel variance and signal-to-noise ratio, wherein the pixel standard deviation and pixel variance are obtained by calculating the standard deviation and variance of multiple pixel values ​​within the preset range of the optimized image, and the signal-to-noise ratio is used to measure the relative relationship between signal strength and noise strength. The obtained pixel standard deviation, pixel variance and signal-to-noise ratio are input into the preset noise intensity assessment model for processing to obtain the noise intensity assessment index: The noise intensity evaluation index calculation formula in the noise intensity evaluation model is: ; in, is the noise intensity evaluation index, , , are pixel standard deviation, pixel variance and signal-to-noise ratio, respectively. , It is the preset characteristic coefficient (the characteristic coefficient is obtained by querying the preset intelligent vision super-sensitive noise reduction database).

[0029] According to an embodiment of the present invention, the method of acquiring an adaptive noise reduction processing method according to the noise category feature data and the noise intensity evaluation index, and performing noise reduction processing on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image includes: Comparing the noise intensity evaluation index with a preset noise intensity standard evaluation index to obtain a relative value of noise intensity; Performing a threshold comparison between the relative value of the noise intensity and a preset noise level assessment threshold, and obtaining a noise level according to the threshold comparison result, including weak noise or strong noise; According to the noise category feature data combined with the noise level, a preset noise reduction method database is queried to obtain an adaptive noise reduction processing method, and noise reduction processing is performed on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image.

[0030] It should be noted that the obtained noise intensity evaluation index is compared with the preset noise intensity standard evaluation index to obtain a relative value of noise intensity. For example, the obtained noise intensity evaluation index is 70, and the preset noise intensity standard evaluation index is 100, then 70 / 100=0.7 is the relative value of noise intensity; the relative value of noise intensity is compared with the preset noise level assessment threshold. In this embodiment, the noise level assessment threshold is set to (0, 0.75], (0.75, 1], corresponding to weak noise and strong noise, respectively. For example, the obtained relative value of noise intensity is 0.7, which corresponds to weak noise; according to the noise category feature data combined with the noise level, the preset visual super-sensitive noise reduction database is queried to obtain an adaptive noise reduction processing method, and the optimized image is denoised according to the adaptive noise reduction processing method to obtain a noise-reduced image, wherein the preset visual super-sensitive noise reduction database is obtained according to the noise category feature data and the noise level and the corresponding adaptive noise reduction processing method.

[0031] Please refer to Figure 3 , Figure 3 The flowchart of obtaining a noise reduction effectiveness index of a smart visual super-sensing noise reduction method in some embodiments of the present application. According to an embodiment of the present invention, the image quality parameter data of the noise reduction image is obtained, and the image quality parameter data is processed to obtain the noise reduction effectiveness index, including: S31, obtaining image quality parameter data of the noise reduction image, including peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index; S32, inputting the peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index into a preset noise reduction processing effectiveness evaluation model for processing to obtain a noise reduction effectiveness index.

[0032] It should be noted that in order to evaluate the image quality after noise reduction processing, image quality parameter data including peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index are first obtained, wherein the image quality parameter data can be calculated by those skilled in the art as needed. The peak signal-to-noise ratio refers to the logarithm of the mean square error calculated after obtaining the pixel value. The higher the peak signal-to-noise ratio value, the better the noise suppression effect of the noise reduction method. The structural similarity index refers to the local mean, variance and covariance of the image obtained by processing. The structural similarity index value is between -1 and 1. The closer the value is to 1, the better the noise reduction. The more similar the denoised image is to the original image in structure, the better the noise reduction effect and detail retention effect are; the noise suppression ratio is obtained by calculating the noise power in the image. The higher the value of the noise suppression ratio, the better the noise suppression effect of the denoising method is; the edge preservation index is obtained by calculating the similarity between the edge pixels of the original image and the corresponding edge pixels of the denoised image. The higher the edge preservation index value, the better the edge preservation effect is during the denoising process; the obtained peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index are input into the preset denoising processing effectiveness evaluation model for processing to obtain the denoising effectiveness index: The noise reduction effectiveness index calculation formula in the noise reduction processing effectiveness evaluation model is: ; in, is the noise reduction effectiveness index, , , , They are peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index, It is the preset characteristic coefficient (the characteristic coefficient is obtained by querying the preset intelligent vision super-sensitive noise reduction database).

[0033] According to an embodiment of the present invention, performing a threshold comparison between the noise reduction effectiveness index and a preset noise reduction effectiveness evaluation threshold, and adjusting the adaptive noise reduction processing method according to the threshold comparison result, includes: Processing the noise reduction effectiveness index and the preset noise reduction demand index to obtain a noise reduction effectiveness deviation rate; Comparing the noise reduction effectiveness deviation rate with a preset noise reduction effectiveness deviation rate threshold; If it is less than or equal to the preset noise reduction effectiveness deviation rate threshold, the noise reduction effectiveness is in normal state; If it is greater than a preset noise reduction effectiveness deviation rate threshold, the noise reduction effectiveness is in an abnormal state, and the adaptive noise reduction processing method is adjusted.

[0034] It should be noted that the noise reduction effectiveness index obtained is processed with the preset noise reduction demand index to obtain the noise reduction effectiveness deviation rate; The noise reduction effectiveness deviation rate calculation formula is: ; in, is the noise reduction effectiveness deviation rate, , are noise reduction effectiveness index and noise reduction demand index respectively; the noise reduction effectiveness deviation rate is compared with the preset noise reduction effectiveness deviation rate threshold. In this embodiment, the noise reduction effectiveness deviation rate threshold is set to (0, 0.3], (0.3, 1], corresponding to the normal state and the abnormal state respectively. For example, if the obtained noise reduction effectiveness deviation rate is 0.2, the noise reduction effectiveness is in the normal state. If the obtained noise reduction effectiveness deviation rate is 0.4, the noise reduction effectiveness is in the abnormal state, and the adaptive noise reduction processing method is adjusted.

[0035] According to an embodiment of the present invention, it also includes: Obtaining the gradient direction, gradient amplitude and direction interval of the denoised image, wherein the direction interval includes an edge direction interval and a texture direction interval, and the gradient amplitude includes an edge direction gradient amplitude and a texture direction gradient amplitude; If the gradient direction is within the edge direction interval and the edge direction gradient amplitude is less than a preset edge direction gradient amplitude threshold, querying a preset enhancement coefficient list to obtain an edge direction enhancement coefficient according to a gradient amplitude deviation rate between the edge direction gradient amplitude and the preset edge direction gradient amplitude threshold; Enhance the edge direction gradient amplitude according to the edge direction enhancement coefficient to obtain the edge direction gradient optimization amplitude; If the gradient direction is within the texture direction interval, obtaining the texture pixel value variance and the texture pixel value standard deviation within the texture direction interval; Inputting the texture pixel value variance and the texture pixel value standard deviation into a preset texture direction enhancement evaluation model for processing to obtain a texture direction enhancement coefficient; The texture direction gradient amplitude is enhanced according to the texture direction enhancement coefficient to obtain the texture direction gradient optimization amplitude.

[0036] It should be noted that in order to highlight the image details and prevent excessive enhancement from causing noise amplification, the gradient direction and gradient amplitude of the denoised image are first obtained. , its gradient direction , gradient amplitude ,in, is the horizontal gradient image, is a vertical gradient image, and then the number of direction intervals n is determined, where 360° / n is the direction interval; if the gradient direction is within the edge direction interval, and the edge direction gradient amplitude is less than the preset edge direction gradient amplitude threshold, in this embodiment, the edge direction gradient amplitude threshold is set to 10, and if the edge direction gradient amplitude is 8, then That is, the gradient amplitude deviation rate between the edge direction gradient amplitude and the preset edge direction gradient amplitude threshold. According to the gradient amplitude deviation rate, the preset enhancement coefficient list is queried to obtain the edge direction enhancement coefficient, wherein the preset enhancement coefficient list is obtained by querying the preset intelligent visual super-sensitive noise reduction database, and the edge direction gradient amplitude is enhanced according to the obtained edge direction enhancement coefficient. The edge direction gradient amplitude calculation formula is: ; in, Optimize the magnitude of the edge direction gradient, , are the edge direction gradient amplitude and edge direction enhancement coefficient respectively; If the gradient direction is within the texture direction interval, then obtaining the texture pixel value variance and the texture pixel value standard deviation within the texture direction interval, wherein the texture pixel value variance and the texture pixel value standard deviation are obtained by obtaining the pixel values ​​within a preset area and then respectively calculating the variance and the standard deviation, and the texture pixel value variance and the texture pixel value standard deviation are input into a preset texture direction enhancement evaluation model for processing to obtain a texture direction enhancement coefficient; The calculation formula of the texture direction enhancement coefficient is: ; in, is the texture direction enhancement coefficient, , are the texture pixel value variance and texture pixel value standard deviation, respectively. , The texture direction gradient amplitude is enhanced according to the preset characteristic coefficient (the characteristic coefficient is obtained by querying the preset intelligent visual super-sensitive noise reduction database) according to the obtained texture direction enhancement coefficient; The texture direction gradient optimization amplitude calculation formula is: ; in, Optimize the magnitude of the texture directional gradient, , are the texture direction gradient amplitude and the texture direction enhancement coefficient respectively.

[0037] It is worth mentioning that according to an embodiment of the present invention, it also includes: Obtain pixel values ​​and corresponding pixel brightness values ​​within a preset area of ​​the denoised image, process according to the pixel brightness values, obtain local contrast data, and extract maximum brightness values ​​and minimum brightness values ​​according to the pixel brightness values; Compare the local contrast data with a preset local contrast threshold: If it is less than or equal to the preset local contrast threshold, it corresponds to low brightness, and the average brightness value and brightness distribution width value of the pixels in the preset area are obtained; Processing is performed according to the average brightness value and the brightness distribution width value in combination with the pixel brightness value and the minimum brightness value to obtain a pixel brightness adjustment value, and adjusting the low brightness area according to the pixel brightness adjustment value; If it is greater than the preset local contrast threshold, it corresponds to high brightness, and the cumulative distribution function value corresponding to the pixel brightness value is obtained through the preset pixel brightness histogram; Processing is performed according to the cumulative distribution function value, the maximum brightness value and the minimum brightness value to obtain a corresponding equalized pixel value, and the high brightness area is adjusted according to the equalized pixel value.

[0038] It should be noted that, in order to highlight the image details, firstly, the pixel values ​​and the corresponding pixel brightness values ​​in the preset area of ​​the denoised image are obtained, and the pixel brightness values ​​are processed to obtain local contrast data, which is the ratio of the difference between the maximum brightness value and the minimum brightness value of the pixel values ​​in the preset area to the sum of the maximum brightness value and the minimum brightness value. Then, the obtained local contrast data is compared with the preset local contrast threshold. In this embodiment, the preset local contrast threshold is set to (0, 0.3], (0.3, 1], corresponding to low brightness and high brightness, respectively; for example, if the obtained local contrast data is 0.2, it corresponds to low brightness, and the average brightness value and brightness distribution width value of the pixels in the preset area are obtained, wherein the brightness distribution width value refers to the difference between the maximum brightness value and the minimum brightness value, and the pixel brightness adjustment value is obtained by processing according to the average brightness value and the brightness distribution width value; The pixel brightness adjustment value calculation formula is: ; in, is the pixel brightness adjustment value, , , , They are pixel brightness value, minimum brightness value, average brightness value and brightness distribution width value respectively. is the preset characteristic coefficient (the characteristic coefficient is obtained by querying the preset intelligent visual super-sensitive noise reduction database); Finally, the pixel value of the low brightness area is adjusted according to the obtained pixel brightness adjustment value; If the local contrast data obtained is 0.5, it corresponds to high brightness. First, the cumulative distribution function value corresponding to the brightness value is obtained by the preset pixel brightness histogram, where the pixel brightness histogram is obtained by querying the preset intelligent visual super-sensitive noise reduction database, and then processed according to the cumulative distribution function value, the maximum brightness value and the minimum brightness value to obtain the corresponding equalized pixel value; The equalized pixel value calculation formula is: ; in, To equalize pixel values, , , are the cumulative distribution function value, the maximum brightness value, and the minimum brightness value respectively; Finally, the high brightness area is adjusted according to the obtained equalized pixel value.

[0039] The present invention also discloses an intelligent visual super-sensing noise reduction system, including a memory and a processor, wherein the memory includes an intelligent visual super-sensing noise reduction method program, and when the intelligent visual super-sensing noise reduction method program is executed by the processor, the following steps are implemented: Obtaining the light intensity value corresponding to the image acquisition time point, obtaining the image sensor parameter value according to the light intensity value, acquiring the real-time image according to the image sensor parameter value, and preprocessing the real-time image to obtain an optimized image; Processing is performed according to the optimized image to obtain noise category characteristic data and pixel value statistical data, and processing is performed according to the pixel value statistical data to obtain a noise intensity evaluation index; Acquire an adaptive noise reduction processing method according to the noise category feature data and the noise intensity evaluation index, and perform noise reduction processing on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image; Acquire image quality parameter data of the noise reduction image, perform processing according to the image quality parameter data, and obtain a noise reduction effectiveness index; The noise reduction effectiveness index is compared with a preset noise reduction effectiveness evaluation threshold, and the adaptive noise reduction processing method is adjusted according to the threshold comparison result.

[0040] It should be noted that in order to achieve effective noise reduction on the captured image, the light intensity value corresponding to the image capture time point is first obtained. Different light intensity values ​​correspond to different image sensor parameter values, including acquisition frequency, exposure time, aperture size and gain value. Image acquisition is implemented according to the queried image sensor parameter values ​​to obtain multiple frames of real-time images, and preprocessing is performed to obtain an optimized image. Then, the corresponding noise category feature data and pixel value statistical data are identified according to the optimized image, and the pixel value statistical data are processed to obtain a noise intensity evaluation index for expressing the noise intensity of the category of noise. The corresponding adaptive noise reduction processing method is obtained according to the noise category feature data and the noise intensity evaluation index, and the optimized image is subjected to noise reduction processing to obtain a noise reduction image. The image quality parameter data of the noise reduction image is extracted and processed to obtain a noise reduction effectiveness index. Finally, the adaptive noise reduction processing method is adjusted according to the threshold comparison result.

[0041] According to an embodiment of the present invention, obtaining the light intensity value corresponding to the image acquisition time point, obtaining the image sensor parameter value according to the light intensity value, acquiring the real-time image according to the image sensor parameter value, and preprocessing the real-time image to obtain the optimized image includes: Obtain the light intensity value corresponding to the image acquisition time point, and query the preset light parameter mapping table according to the light intensity value to obtain the image sensor parameter value, including acquisition frequency, exposure time, aperture size and gain value; Taking photos according to the acquisition frequency, exposure time, aperture size and gain value to obtain multiple frames of real-time images; The multiple frames of real-time images are preprocessed to obtain optimized images.

[0042] It should be noted that the lighting conditions at different image acquisition time points are different, and the required image sensor parameter values ​​are also different. In order to ensure the quality of the acquired image, the light intensity value corresponding to the image acquisition time point is first obtained, and the preset lighting parameter mapping table is queried according to the light intensity value to obtain the image sensor parameter values ​​including acquisition frequency, exposure time, aperture size and gain value. Among them, the preset lighting parameter mapping table is obtained by querying the preset intelligent vision super-sensitive noise reduction database, and then photos are taken according to the queried acquisition frequency, exposure time, aperture size and gain value to obtain multiple frames of real-time images, and the image is preprocessed to obtain an optimized image for subsequent noise reduction processing.

[0043] According to an embodiment of the present invention, the preprocessing of the multiple frames of real-time images to obtain optimized images includes: Extracting ORB descriptors corresponding to corner points respectively according to the multiple frames of real-time images; The matching feature points are obtained by matching the distances between the ORB descriptors of multiple frames of real-time images; Processing is performed based on the matching feature points to obtain a geometric transformation relationship, and processing is performed based on the geometric transformation relationship to obtain an affine transformation matrix; The multiple frames of real-time images are optimized respectively according to the affine transformation matrix to obtain optimized images.

[0044] It should be noted that, since there may be slight displacement of the device, jitter and other factors during the acquisition process that may cause the image to shift, it is necessary to first align the obtained multiple frames of real-time images, and use the FAST (Features from Accelerated Segment Test) corner detection algorithm to determine whether the point is a corner point. After the corner point is determined, the ORB (Oriented FAST and Rotated BRIEF) descriptor corresponding to the corner point is extracted. For example, for a 31*31 corner point neighborhood, n pairs of pixels are randomly selected, and the grayscale values ​​of each pair of pixels are compared. If the grayscale value of the first pixel is greater than the grayscale value of the second pixel, the corresponding binary bit is set to 1, otherwise it is set to 0. For two frames of images, matching is performed by calculating the distance between the descriptors. If n pairs of matching feature points are found in the two frames of images, those skilled in the art can represent their geometric transformation relationship according to actual needs, obtain an affine transformation matrix according to the geometric transformation relationship, and optimize the multiple frames of real-time images according to the obtained affine transformation matrix, including rotation, scaling, translation or shearing, to obtain optimized images.

[0045] According to an embodiment of the present invention, the step of processing the optimized image to obtain noise category feature data and pixel value statistics, and processing the pixel value statistics to obtain a noise intensity evaluation index includes: Inputting the optimized image into a preset image noise recognition model for processing to obtain noise category feature data; Obtaining pixel value statistics according to the optimized image, including pixel standard deviation, pixel variance and signal-to-noise ratio; The pixel standard deviation, pixel variance and signal-to-noise ratio are input into a preset noise intensity assessment model for processing to obtain a noise intensity assessment index.

[0046] It should be noted that different types and intensities of image noise use different noise reduction processing methods. In order to match the optimal noise reduction processing method, the optimized image is first input into the preset image noise recognition model for processing to obtain noise category feature data, wherein the image noise recognition model is obtained by obtaining a large number of historical sample optimized images and corresponding noise category feature data for training; then the pixel value statistics of the optimized image are obtained, including pixel standard deviation, pixel variance and signal-to-noise ratio, wherein the pixel standard deviation and pixel variance are obtained by calculating the standard deviation and variance of multiple pixel values ​​within the preset range of the optimized image, and the signal-to-noise ratio is used to measure the relative relationship between signal strength and noise strength. The obtained pixel standard deviation, pixel variance and signal-to-noise ratio are input into the preset noise intensity assessment model for processing to obtain the noise intensity assessment index: The noise intensity evaluation index calculation formula in the noise intensity evaluation model is: ; in, is the noise intensity evaluation index, , , are pixel standard deviation, pixel variance and signal-to-noise ratio, respectively. , It is the preset characteristic coefficient (the characteristic coefficient is obtained by querying the preset intelligent vision super-sensitive noise reduction database).

[0047] According to an embodiment of the present invention, the method of acquiring an adaptive noise reduction processing method according to the noise category feature data and the noise intensity evaluation index, and performing noise reduction processing on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image includes: Comparing the noise intensity evaluation index with a preset noise intensity standard evaluation index to obtain a relative value of noise intensity; Performing a threshold comparison between the relative value of the noise intensity and a preset noise level assessment threshold, and obtaining a noise level according to the threshold comparison result, including weak noise or strong noise; According to the noise category feature data combined with the noise level, a preset noise reduction method database is queried to obtain an adaptive noise reduction processing method, and noise reduction processing is performed on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image.

[0048] It should be noted that the obtained noise intensity evaluation index is compared with the preset noise intensity standard evaluation index to obtain a relative value of noise intensity. For example, the obtained noise intensity evaluation index is 70, and the preset noise intensity standard evaluation index is 100, then 70 / 100=0.7 is the relative value of noise intensity; the relative value of noise intensity is compared with the preset noise level assessment threshold. In this embodiment, the noise level assessment threshold is set to (0, 0.75], (0.75, 1], corresponding to weak noise and strong noise, respectively. For example, the obtained relative value of noise intensity is 0.7, which corresponds to weak noise; according to the noise category feature data combined with the noise level, the preset visual super-sensitive noise reduction database is queried to obtain an adaptive noise reduction processing method, and the optimized image is denoised according to the adaptive noise reduction processing method to obtain a noise-reduced image, wherein the preset visual super-sensitive noise reduction database is obtained according to the noise category feature data and the noise level and the corresponding adaptive noise reduction processing method.

[0049] According to an embodiment of the present invention, the step of acquiring image quality parameter data of the denoised image and performing processing according to the image quality parameter data to obtain the denoising effectiveness index includes: Obtain image quality parameter data of the denoised image, including peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index; The peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index are input into a preset noise reduction processing effectiveness evaluation model for processing to obtain a noise reduction effectiveness index.

[0050] It should be noted that in order to evaluate the image quality after noise reduction processing, image quality parameter data including peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index are first obtained, wherein the image quality parameter data can be calculated by those skilled in the art as needed. The peak signal-to-noise ratio refers to the logarithm of the mean square error calculated after obtaining the pixel value. The higher the peak signal-to-noise ratio value, the better the noise suppression effect of the noise reduction method. The structural similarity index refers to the local mean, variance and covariance of the image obtained by processing. The structural similarity index value is between -1 and 1. The closer the value is to 1, the better the noise reduction. The more similar the denoised image is to the original image in structure, the better the noise reduction effect and detail retention effect are; the noise suppression ratio is obtained by calculating the noise power in the image. The higher the value of the noise suppression ratio, the better the noise suppression effect of the denoising method is; the edge preservation index is obtained by calculating the similarity between the edge pixels of the original image and the corresponding edge pixels of the denoised image. The higher the edge preservation index value, the better the edge preservation effect is during the denoising process; the obtained peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index are input into the preset denoising processing effectiveness evaluation model for processing to obtain the denoising effectiveness index: The noise reduction effectiveness index calculation formula in the noise reduction processing effectiveness evaluation model is: ; in, is the noise reduction effectiveness index, , , , They are peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index, It is the preset characteristic coefficient (the characteristic coefficient is obtained by querying the preset intelligent vision super-sensitive noise reduction database).

[0051] According to an embodiment of the present invention, performing a threshold comparison between the noise reduction effectiveness index and a preset noise reduction effectiveness evaluation threshold, and adjusting the adaptive noise reduction processing method according to the threshold comparison result, includes: Processing the noise reduction effectiveness index and the preset noise reduction demand index to obtain a noise reduction effectiveness deviation rate; Performing a threshold comparison between the noise reduction effectiveness deviation rate and a preset noise reduction effectiveness deviation rate threshold; If it is less than or equal to the preset noise reduction effectiveness deviation rate threshold, the noise reduction effectiveness is in normal state; If it is greater than a preset noise reduction effectiveness deviation rate threshold, the noise reduction effectiveness is in an abnormal state, and the adaptive noise reduction processing method is adjusted.

[0052] It should be noted that the noise reduction effectiveness index obtained is processed with the preset noise reduction demand index to obtain the noise reduction effectiveness deviation rate; The noise reduction effectiveness deviation rate calculation formula is: ; in, is the noise reduction effectiveness deviation rate, , are noise reduction effectiveness index and noise reduction demand index respectively; the noise reduction effectiveness deviation rate is compared with the preset noise reduction effectiveness deviation rate threshold. In this embodiment, the noise reduction effectiveness deviation rate threshold is set to (0, 0.3], (0.3, 1], corresponding to the normal state and the abnormal state respectively. For example, if the obtained noise reduction effectiveness deviation rate is 0.2, the noise reduction effectiveness is in the normal state. If the obtained noise reduction effectiveness deviation rate is 0.4, the noise reduction effectiveness is in the abnormal state, and the adaptive noise reduction processing method is adjusted.

[0053] According to an embodiment of the present invention, it also includes: Obtaining the gradient direction, gradient amplitude and direction interval of the denoised image, wherein the direction interval includes an edge direction interval and a texture direction interval, and the gradient amplitude includes an edge direction gradient amplitude and a texture direction gradient amplitude; If the gradient direction is within the edge direction interval and the edge direction gradient amplitude is less than a preset edge direction gradient amplitude threshold, querying a preset enhancement coefficient list to obtain an edge direction enhancement coefficient according to a gradient amplitude deviation rate between the edge direction gradient amplitude and the preset edge direction gradient amplitude threshold; Enhance the edge direction gradient amplitude according to the edge direction enhancement coefficient to obtain the edge direction gradient optimization amplitude; If the gradient direction is within the texture direction interval, obtaining the texture pixel value variance and the texture pixel value standard deviation within the texture direction interval; Inputting the texture pixel value variance and the texture pixel value standard deviation into a preset texture direction enhancement evaluation model for processing to obtain a texture direction enhancement coefficient; The texture direction gradient amplitude is enhanced according to the texture direction enhancement coefficient to obtain the texture direction gradient optimization amplitude.

[0054] It should be noted that in order to highlight the image details and prevent excessive enhancement from causing noise amplification, the gradient direction and gradient amplitude of the denoised image are first obtained. , its gradient direction , gradient amplitude ,in, is the horizontal gradient image, is a vertical gradient image, and then the number of direction intervals n is determined, where 360° / n is the direction interval; if the gradient direction is within the edge direction interval, and the edge direction gradient amplitude is less than the preset edge direction gradient amplitude threshold, in this embodiment, the edge direction gradient amplitude threshold is set to 10, and if the edge direction gradient amplitude is 8, then That is, the gradient amplitude deviation rate between the edge direction gradient amplitude and the preset edge direction gradient amplitude threshold. According to the gradient amplitude deviation rate, the preset enhancement coefficient list is queried to obtain the edge direction enhancement coefficient, wherein the preset enhancement coefficient list is obtained by querying the preset intelligent visual super-sensitive noise reduction database, and the edge direction gradient amplitude is enhanced according to the obtained edge direction enhancement coefficient. The edge direction gradient amplitude calculation formula is: ; in, Optimize the magnitude of the edge direction gradient, , are the edge direction gradient amplitude and edge direction enhancement coefficient respectively; If the gradient direction is within the texture direction interval, then obtaining the texture pixel value variance and the texture pixel value standard deviation within the texture direction interval, wherein the texture pixel value variance and the texture pixel value standard deviation are obtained by obtaining the pixel values ​​within a preset area and then respectively calculating the variance and the standard deviation, and the texture pixel value variance and the texture pixel value standard deviation are input into a preset texture direction enhancement evaluation model for processing to obtain a texture direction enhancement coefficient; The calculation formula of the texture direction enhancement coefficient is: ; in, is the texture direction enhancement coefficient, , are the texture pixel value variance and texture pixel value standard deviation, respectively. , The texture direction gradient amplitude is enhanced according to the preset characteristic coefficient (the characteristic coefficient is obtained by querying the preset intelligent visual super-sensitive noise reduction database) according to the obtained texture direction enhancement coefficient; The texture direction gradient optimization amplitude calculation formula is: ; in, Optimize the magnitude of the texture directional gradient, , are the texture direction gradient amplitude and the texture direction enhancement coefficient respectively.

[0055] It is worth mentioning that according to an embodiment of the present invention, it also includes: Obtain pixel values ​​and corresponding pixel brightness values ​​within a preset area of ​​the denoised image, process according to the pixel brightness values, obtain local contrast data, and extract maximum brightness values ​​and minimum brightness values ​​according to the pixel brightness values; Compare the local contrast data with a preset local contrast threshold: If it is less than or equal to the preset local contrast threshold, it corresponds to low brightness, and the average brightness value and brightness distribution width value of the pixels in the preset area are obtained; Processing is performed according to the average brightness value and the brightness distribution width value in combination with the pixel brightness value and the minimum brightness value to obtain a pixel brightness adjustment value, and adjusting the low brightness area according to the pixel brightness adjustment value; If it is greater than the preset local contrast threshold, it corresponds to high brightness, and the cumulative distribution function value corresponding to the pixel brightness value is obtained through the preset pixel brightness histogram; Processing is performed according to the cumulative distribution function value, the maximum brightness value and the minimum brightness value to obtain a corresponding equalized pixel value, and the high brightness area is adjusted according to the equalized pixel value.

[0056] It should be noted that, in order to highlight the image details, firstly, the pixel values ​​and the corresponding pixel brightness values ​​in the preset area of ​​the denoised image are obtained, and the pixel brightness values ​​are processed to obtain local contrast data, which is the ratio of the difference between the maximum brightness value and the minimum brightness value of the pixel values ​​in the preset area to the sum of the maximum brightness value and the minimum brightness value. Then, the obtained local contrast data is compared with the preset local contrast threshold. In this embodiment, the preset local contrast threshold is set to (0, 0.3], (0.3, 1], corresponding to low brightness and high brightness, respectively; for example, if the obtained local contrast data is 0.2, it corresponds to low brightness, and the average brightness value and brightness distribution width value of the pixels in the preset area are obtained, wherein the brightness distribution width value refers to the difference between the maximum brightness value and the minimum brightness value, and the pixel brightness adjustment value is obtained by processing according to the average brightness value and the brightness distribution width value; The pixel brightness adjustment value calculation formula is: ; in, is the pixel brightness adjustment value, , , , They are pixel brightness value, minimum brightness value, average brightness value and brightness distribution width value respectively. is the preset characteristic coefficient (the characteristic coefficient is obtained by querying the preset intelligent visual super-sensitive noise reduction database); Finally, the pixel value of the low brightness area is adjusted according to the obtained pixel brightness adjustment value; If the local contrast data obtained is 0.5, it corresponds to high brightness. First, the cumulative distribution function value corresponding to the brightness value is obtained by the preset pixel brightness histogram, where the pixel brightness histogram is obtained by querying the preset intelligent visual super-sensitive noise reduction database, and then processed according to the cumulative distribution function value, the maximum brightness value and the minimum brightness value to obtain the corresponding equalized pixel value; The equalized pixel value calculation formula is: ; in, To equalize pixel values, , , are the cumulative distribution function value, the maximum brightness value, and the minimum brightness value respectively; Finally, the high brightness area is adjusted according to the obtained equalized pixel value.

[0057] The third aspect of the present invention provides a readable storage medium, which stores a smart visual super-sensitive noise reduction method program. When the smart visual super-sensitive noise reduction method program is executed by a processor, the steps of the smart visual super-sensitive noise reduction method as described in any one of the above items are implemented.

[0058] The present invention discloses a smart vision super-sensitive noise reduction method, system and medium, which realizes intelligent noise reduction of smart vision super-sensitive images by obtaining an adaptive noise reduction processing method through image preprocessing, identification of noise categories and noise intensity.

[0059] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0060] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0061] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0062] Those skilled in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, the aforementioned program can be stored in a readable storage medium, and when the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, and other media that can store program codes.

[0063] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

Claims

1. A smart visual super-sensing noise reduction method, characterized in that: The following steps are involved: Obtaining the light intensity value corresponding to the image acquisition time point, obtaining the image sensor parameter value according to the light intensity value, acquiring the real-time image according to the image sensor parameter value, and preprocessing the real-time image to obtain an optimized image; Processing is performed according to the optimized image to obtain noise category characteristic data and pixel value statistical data, and processing is performed according to the pixel value statistical data to obtain a noise intensity evaluation index; Acquire an adaptive noise reduction processing method according to the noise category feature data and the noise intensity evaluation index, and perform noise reduction processing on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image; Acquire image quality parameter data of the noise reduction image, perform processing according to the image quality parameter data, and obtain a noise reduction effectiveness index; The noise reduction effectiveness index is compared with a preset noise reduction effectiveness evaluation threshold, and the adaptive noise reduction processing method is adjusted according to the threshold comparison result.

2. The intelligent visual super-sensing noise reduction method according to claim 1 is characterized in that: The step of obtaining a light intensity value corresponding to an image acquisition time point, obtaining an image sensor parameter value according to the light intensity value, acquiring a real-time image according to the image sensor parameter value, and preprocessing the real-time image to obtain an optimized image includes: Obtain the light intensity value corresponding to the image acquisition time point, and query the preset light parameter mapping table according to the light intensity value to obtain the image sensor parameter value, including acquisition frequency, exposure time, aperture size and gain value; Taking photos according to the acquisition frequency, exposure time, aperture size and gain value to obtain multiple frames of real-time images; The multiple frames of real-time images are preprocessed to obtain optimized images.

3. The intelligent visual super-sensing noise reduction method according to claim 2 is characterized in that: The preprocessing of the multiple frames of real-time images to obtain optimized images includes: Extracting ORB descriptors corresponding to corner points respectively according to the multiple frames of real-time images; The matching feature points are obtained by matching the distances between the ORB descriptors of multiple frames of real-time images; Processing is performed based on the matching feature points to obtain a geometric transformation relationship, and processing is performed based on the geometric transformation relationship to obtain an affine transformation matrix; The multiple frames of real-time images are optimized respectively according to the affine transformation matrix to obtain optimized images.

4. The intelligent visual super-sensing noise reduction method according to claim 3 is characterized in that: The step of processing the optimized image to obtain noise category feature data and pixel value statistics, and processing the pixel value statistics to obtain a noise intensity evaluation index comprises: Inputting the optimized image into a preset image noise recognition model for processing to obtain noise category feature data; Obtaining pixel value statistics according to the optimized image, including pixel standard deviation, pixel variance and signal-to-noise ratio; The pixel standard deviation, pixel variance and signal-to-noise ratio are input into a preset noise intensity assessment model for processing to obtain a noise intensity assessment index.

5. The intelligent visual super-sensing noise reduction method according to claim 4 is characterized in that: The step of acquiring an adaptive noise reduction processing method according to the noise category feature data and the noise intensity evaluation index, and performing noise reduction processing on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image includes: Comparing the noise intensity evaluation index with a preset noise intensity standard evaluation index to obtain a relative value of noise intensity; Performing a threshold comparison between the relative value of the noise intensity and a preset noise level assessment threshold, and obtaining a noise level according to the threshold comparison result, including weak noise or strong noise; According to the noise category feature data combined with the noise level, a preset noise reduction method database is queried to obtain an adaptive noise reduction processing method, and noise reduction processing is performed on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image.

6. The intelligent visual super-sensing noise reduction method according to claim 5 is characterized in that: The step of obtaining image quality parameter data of the noise reduction image and performing processing according to the image quality parameter data to obtain a noise reduction effectiveness index includes: Obtain image quality parameter data of the denoised image, including peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index; The peak signal-to-noise ratio, structural similarity index, noise suppression ratio and edge preservation index are input into a preset noise reduction processing effectiveness evaluation model for processing to obtain a noise reduction effectiveness index.

7. The intelligent visual super-sensing noise reduction method according to claim 6 is characterized in that: The step of comparing the noise reduction effectiveness index with a preset noise reduction effectiveness evaluation threshold, and adjusting the adaptive noise reduction processing method according to the threshold comparison result, includes: Processing the noise reduction effectiveness index and the preset noise reduction demand index to obtain a noise reduction effectiveness deviation rate; Performing a threshold comparison between the noise reduction effectiveness deviation rate and a preset noise reduction effectiveness deviation rate threshold; If it is less than or equal to the preset noise reduction effectiveness deviation rate threshold, the noise reduction effectiveness is in normal state; If it is greater than a preset noise reduction effectiveness deviation rate threshold, the noise reduction effectiveness is in an abnormal state, and the adaptive noise reduction processing method is adjusted.

8. The intelligent visual super-sensing noise reduction method according to claim 7 is characterized in that: Also includes: Obtaining the gradient direction, gradient amplitude and direction interval of the denoised image, wherein the direction interval includes an edge direction interval and a texture direction interval, and the gradient amplitude includes an edge direction gradient amplitude and a texture direction gradient amplitude; If the gradient direction is within the edge direction interval and the edge direction gradient amplitude is less than a preset edge direction gradient amplitude threshold, querying a preset enhancement coefficient list to obtain an edge direction enhancement coefficient according to a gradient amplitude deviation rate between the edge direction gradient amplitude and the preset edge direction gradient amplitude threshold; Enhance the edge direction gradient amplitude according to the edge direction enhancement coefficient to obtain the edge direction gradient optimization amplitude; If the gradient direction is within the texture direction interval, obtaining the texture pixel value variance and the texture pixel value standard deviation within the texture direction interval; Inputting the texture pixel value variance and the texture pixel value standard deviation into a preset texture direction enhancement evaluation model for processing to obtain a texture direction enhancement coefficient; The texture direction gradient amplitude is enhanced according to the texture direction enhancement coefficient to obtain the texture direction gradient optimization amplitude.

9. An intelligent visual super-sensitive noise reduction system, characterized in that: The invention comprises a memory and a processor, wherein the memory comprises a program of a smart visual super-sensitive noise reduction method, and when the program of the smart visual super-sensitive noise reduction method is executed by the processor, the following steps are implemented: Obtaining the light intensity value corresponding to the image acquisition time point, obtaining the image sensor parameter value according to the light intensity value, acquiring the real-time image according to the image sensor parameter value, and preprocessing the real-time image to obtain an optimized image; Processing is performed according to the optimized image to obtain noise category characteristic data and pixel value statistical data, and processing is performed according to the pixel value statistical data to obtain a noise intensity evaluation index; Acquire an adaptive noise reduction processing method according to the noise category feature data and the noise intensity evaluation index, and perform noise reduction processing on the optimized image according to the adaptive noise reduction processing method to obtain a noise-reduced image; Acquire image quality parameter data of the noise reduction image, perform processing according to the image quality parameter data, and obtain a noise reduction effectiveness index; The noise reduction effectiveness index is compared with a preset noise reduction effectiveness evaluation threshold, and the adaptive noise reduction processing method is adjusted according to the threshold comparison result.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a smart vision super-sensitive noise reduction method program. When the smart vision super-sensitive noise reduction method program is executed by the processor, the steps of a smart vision super-sensitive noise reduction method as described in any one of claims 1 to 8 are implemented.

Images