Image noise reduction method and device under extremely low illumination

Abstract

The invention relates to the field of image noise reduction, in particular to an image noise reduction method and device under extremely low illumination, and the method comprises the following steps: S1, carrying out the illumination enhancement of a monitoring image environment through a light-emitting device, and carrying out the shooting of a monitoring image in at least four directions, performing weighted summation on pixels in at least four overlapped image blocks in the monitoring image to obtain a first target image; s2, background noise of the monitoring image is removed by using a Gaussian blur filter, contrast enhancement is carried out, the overall contrast of the monitoring image is improved by using a contrast equalization method, and a second target image is obtained; s3, performing edge estimation according to the first target image to obtain an edge estimation result, and performing filtering processing on the second target image according to the edge estimation result to obtain the strong light noise reduction image; and S4, processing the missing region of the target image by adopting an attention module so as to realize classification of the target image. The method has the advantages that the image classification accuracy can be improved, and the error rate of image classification is reduced.

Description

[Technical field]

[0001] The present invention relates to the field of image noise reduction, and in particular to a method and a device for image noise reduction under extremely low illumination. [Background technology]

[0002] In a dimly lit environment, such as when a car is driving at night, there is a lot of noise information in the captured image. Therefore, in order to improve the quality of the captured image, noise reduction processing may be performed on the captured image.

[0003] When the ambient light intensity is weak, the captured image will have large noise, low contrast, and wide blurred areas. Effective information is likely to be submerged, resulting in inaccurate recognition, which has a great impact on image classification and increases the error rate of image classification. [Summary of the invention]

[0004] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide a method and device for image denoising under extremely low illumination, which can improve the accuracy of image classification and reduce the error rate of image classification.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0006] A method for image noise reduction under extremely low illumination comprises the following steps:

[0007] S1: enhancing the illumination of the monitoring image environment by means of a light-emitting device, photographing the monitoring image from at least four directions, and performing weighted summation on pixels in at least four overlapping image blocks in the monitoring image to obtain a first target image;

[0008] S2: Use a Gaussian blur filter to remove background noise from the monitoring image, perform contrast enhancement, and use a contrast equalization method to improve the overall contrast of the monitoring image to obtain a second target image;

[0009] S3: performing edge estimation according to the first target image to obtain an edge estimation result, and performing filtering processing on the second target image according to the edge estimation result to obtain a strong light noise reduction image;

[0010] S4: The attention module is used to process the missing areas of the target image, thereby achieving the classification of the target image.

[0011] Preferably, S1 specifically includes:

[0012] The image blocks in the monitoring image are weighted averaged one by one to obtain an image containing overlapping blocks, including: obtaining a weight value of each image block; calculating the weight value according to a calculation formula, and calculating the image blocks in the image containing overlapping blocks.

[0013] Among them, a i is the i-th image block in the image containing overlapping blocks, To monitor the i-th first image block in the image, is the nth monitoring image with Matching image patches, for The weight value of .

[0014] Preferably, obtaining the weight value of the pixel in the overlapping monitoring image includes: calculating the weight of the pixel in the image block

[0015] where q A (x, y) is the weight value of the pixel in the overlapping image block, b s is the width of the image block.

[0016] Preferably, S2 specifically includes:

[0017] Using Gaussian blur filter to remove background noise is as follows: based on the weighted average of adjacent pixel values ​​in Gaussian blur, the value of each pixel is reset to a new value. The formula used is:

[0018]

[0019] Among them, T(x,y) is a two-dimensional Gaussian function, γ is the standard deviation, and x and y are pixel values.

[0020] Preferably, S3 specifically includes:

[0021] Determine the spatial filtering intensity of each pixel in the first target image signal;

[0022] Extracting local information from the first target image signal to obtain first local information, and determining a joint spatial filtering strength corresponding to each pixel point according to the spatial filtering strength and the first local information;

[0023] The second target image is spatially filtered according to the edge estimation result to obtain a strong light denoised image.

[0024] Preferably, S4 specifically includes:

[0025] S41: The attention module is used to automatically identify the blurred area of ​​the rough strong light denoising image and provide guidance information for the blurred area features;

[0026] S42: The attention-driven module further extracts and refines the fuzzy region feature features from shallow to high layers based on the guidance information.

[0027] A device for reducing image noise under extremely low illumination, the device adopts the above-mentioned method for reducing image noise under extremely low illumination, and the device comprises a shooting module, which is used to enhance the illumination of a monitoring image environment through a light-emitting device, shoot the monitoring image in at least four directions, and perform weighted summation on pixels in at least four overlapping image blocks in the monitoring image to obtain a first target image;

[0028] A denoising module is used to remove background noise from the monitoring image using a Gaussian blur filter, and to perform contrast enhancement, and to use a contrast equalization method to improve the overall contrast of the monitoring image to obtain a second target image;

[0029] An edge estimation module is used to perform edge estimation based on the first target image to obtain an edge estimation result, and filter the second target image based on the edge estimation result to obtain a strong light noise reduction image;

[0030] The attention module is used to process the missing areas of the target image to achieve the classification of the target image.

[0031] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned image noise reduction method are implemented.

[0032] A processor is used to run a program, wherein the image noise reduction method is executed when the program is run.

[0033] A terminal device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned image noise reduction method are implemented.

[0034] The beneficial effects of the present invention are as follows:

[0035] 1. Enhance the illumination of the monitoring image environment through the light-emitting device, shoot the monitoring image in at least four directions, perform weighted summation on the pixels in at least four overlapping image blocks in the monitoring image, and obtain the first target image; for example, a car is driving at night, and a car is coming towards it. The two headlights of the car are the light-emitting devices, which illuminate the car with strong light to improve the illumination. Nowadays, the camera device of the car is generally installed in the direction of the front-view mirror. Since the car will not come directly towards the car, it will generally drive to the left or right of the car. Therefore, the camera device illuminates the car in four directions of up, down, left and right to keep a backup. Of course, the four directions are not directly above, directly below, directly left and directly right, and can also be offset to the left or right from the upper line by 1 degree to 45 degrees.

[0036] 2. Use a Gaussian blur filter to remove background noise from the monitoring image, that is, blur the background in the image, such as trees and roads, and enhance the contrast of the car. Use the contrast equalization method to improve the overall contrast of the monitoring image to obtain the second target image. The contrast equalization method uses high-contrast complementary colors, namely red and green, yellow and purple, blue and orange, to cover the entire car with red, and the trees and roads with green. The use of high contrast can make the background removal more efficient and reduce the deletion of erroneous pixels for the main objects.

[0037] 3. Perform edge estimation based on the first target image to obtain edge estimation results, filter the second target image based on the edge estimation results to obtain a strong light noise reduction image, and filter the pixels connecting roads, trees and cars to further improve the edge pixel clarity and recognition accuracy.

[0038] 4. The attention module is used to process the missing areas of the target image, thereby realizing the classification of the target image. The edge pixels of the missing areas in the main body of the car are learned and acquired, such as color for feature guidance, to improve the overall integrity of the object and facilitate classification and recognition.

[0039] These features and advantages of the present invention will be disclosed in detail in the following specific embodiments and drawings.

Brief Description of the Drawings

[0040] The present invention will be further described below in conjunction with the accompanying drawings:

[0041] Figure 1 This is a flow chart of the first embodiment of the present invention. [Specific implementation method]

[0042] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0043] The following first describes the concepts involved in this application in conjunction with the accompanying drawings. It should be noted that the following description of each concept is only to make the content of this application easier to understand and does not limit the scope of protection of this application.

[0044] Embodiment 1:

[0045] A method for image noise reduction under extremely low illumination, such as Figure 1 As shown, the following steps are included:

[0046] S1: Enhance the illumination of the monitoring image environment through the light-emitting device, shoot the monitoring image in at least four directions, perform weighted summation on the pixels in at least four overlapping image blocks in the monitoring image, and obtain the first target image; for example, a car is driving at night, and a car is coming head-on. The two headlights of the car are the light-emitting devices, which illuminate the car with strong light to improve the illumination. Nowadays, the camera device of the car is generally installed in the direction of the front-view mirror. Since the car will not drive directly towards the car, it will generally drive to the left or right of the car. Therefore, the camera device illuminates the car in four directions of up, down, left and right to keep a backup. Of course, the four directions are not directly above, directly below, directly left and directly right, and can also be offset to the left or right from the upper line by 1 degree to 45 degrees.

[0047] Furthermore, S1 specifically includes:

[0048] The image blocks in the monitoring image are weighted averaged one by one to obtain an image containing overlapping blocks, including: obtaining a weight value of each image block; calculating the weight value according to a calculation formula, and calculating the image blocks in the image containing overlapping blocks.

[0049] Among them, a i is the i-th image block in the image containing overlapping blocks, To monitor the i-th first image block in the image, is the nth monitoring image with Matching image patches, for The weight value of .

[0050] Obtaining the weight value of the pixel in the overlapping monitoring image includes: calculating the weight of the pixel in the image block

[0051] where q A (x, y) is the weight value of the pixel in the overlapping image block, b s is the width of the image block. Pixels with high weight values ​​are retained and pixels with low weight values ​​are removed to reduce the impact of blurred pixels on object recognition, enhance the clarity of the photo, and improve the accuracy of object recognition.

[0052] S2: Use a Gaussian blur filter to remove background noise from the monitoring image, that is, blur the background in the image, such as trees and roads, and enhance the contrast of the car. Use the contrast equalization method to improve the overall contrast of the monitoring image to obtain the second target image. The contrast equalization method uses high-contrast complementary colors, namely red and green, yellow and purple, blue and orange, to cover the entire car with red, and the entire tree and road with green. Using high contrast can make the background removal more efficient and reduce the deletion of erroneous pixels of the main object;

[0053] Furthermore, S2 specifically includes:

[0054] Using Gaussian blur filter to remove background noise is as follows: based on the weighted average of adjacent pixel values ​​in Gaussian blur, the value of each pixel is reset to a new value. The formula used is:

[0055]

[0056] Among them, T(x,y) is a two-dimensional Gaussian function, γ is the standard deviation, and x and y are pixel values.

[0057] S3: performing edge estimation according to the first target image to obtain an edge estimation result, performing filtering processing on the second target image according to the edge estimation result to obtain a strong light noise reduction image, and performing filtering processing on the pixels connecting the road, trees and cars to further improve the edge pixel clarity and recognition accuracy;

[0058] S3 specifically includes:

[0059] Determine the spatial filtering intensity of each pixel in the first target image signal;

[0060] Extract local information from the first target image signal to obtain first local information, perform preliminary edge estimation on the first target image in S1, and then determine the joint spatial filtering strength corresponding to each pixel point according to the spatial filtering strength and the first local information;

[0061] According to the edge estimation result, the second target image in S2 covered with contrast color is spatially filtered to obtain a strong light noise reduction image, thereby improving the accuracy of edge pixel recognition.

[0062] S4: The attention module is used to process the missing areas of the target image, thereby achieving the classification of the target image. The edge pixels of the missing areas in the main body of the car are learned and acquired, such as color for feature guidance, to improve the overall integrity of the object and facilitate classification and recognition.

[0063] Furthermore, S4 specifically includes,

[0064] S41: The attention module is used to automatically identify the blurred area of ​​the rough strong light denoising image and provide guidance information for the blurred area features;

[0065] S42: The attention-driven module further extracts and refines the features of the blurred area from shallow to high layers based on the guidance information to improve the overall integrity of the object.

[0066] A device for reducing image noise under extremely low illumination, the device adopts the above-mentioned method for reducing image noise under extremely low illumination, and the device comprises a shooting module, which is used to enhance the illumination of a monitoring image environment through a light-emitting device, shoot the monitoring image in at least four directions, and perform weighted summation on pixels in at least four overlapping image blocks in the monitoring image to obtain a first target image;

[0067] A denoising module is used to remove background noise from the monitoring image using a Gaussian blur filter, and to perform contrast enhancement, and to use a contrast equalization method to improve the overall contrast of the monitoring image to obtain a second target image;

[0068] An edge estimation module is used to perform edge estimation based on the first target image to obtain an edge estimation result, and filter the second target image based on the edge estimation result to obtain a strong light noise reduction image;

[0069] The attention module is used to process the missing areas of the target image to achieve the classification of the target image.

[0070] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned image noise reduction method are implemented.

[0071] A processor is used to run a program, wherein the image noise reduction method is executed when the program is run.

[0072] A terminal device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned image noise reduction method are implemented.

[0073] Those of ordinary skill in the art will appreciate that the units (or modules, steps, etc., the same below) of the examples described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0074] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored, etc.

[0075] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0076] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which 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 perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-0nly Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc., which can store program code.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A method for image noise reduction under extremely low illumination, characterized in that: The following steps are involved: S1: enhancing the illumination of the monitoring image environment by means of a light-emitting device, photographing the monitoring image from at least four directions, and performing weighted summation on pixels in at least four overlapping image blocks in the monitoring image to obtain a first target image; S2: using a Gaussian blur filter to remove background noise from the monitoring image, and performing contrast enhancement, and using a contrast equalization method to improve the overall contrast of the monitoring image to obtain a second target image; S3: performing edge estimation according to the first target image to obtain an edge estimation result, and performing filtering processing on the second target image according to the edge estimation result to obtain the strong light noise reduction image; S4: Using an attention module to process the missing area of ​​the target image, thereby achieving classification of the target image.

2. The method for image noise reduction under extremely low illumination as claimed in claim 1, characterized in that: The S1 specifically includes: The image blocks in the monitoring image are weighted averaged one by one to obtain an image containing overlapping blocks, including: obtaining a weight value of each image block; calculating the weight value according to a calculation formula, and calculating the weight value of the image blocks in the image containing overlapping blocks. Among them, a i is the i-th image block in the image containing overlapping blocks, is the i-th first image block in the monitoring image, is the nth monitoring image with Matching image patches, for The weight value of .

3. The method for image noise reduction under extremely low illumination as claimed in claim 2, characterized in that: Obtaining the weight value of the pixel in the overlapping monitoring image includes: calculating the weight of the pixel in the image block where q A (x, y) is the weight value of the pixel in the overlapping image block, b s is the width of the image block.

4. The method for image noise reduction under extremely low illumination as claimed in claim 1, characterized in that: The S2 specifically includes: Using Gaussian blur filter to remove background noise is as follows: based on the weighted average of adjacent pixel values ​​in Gaussian blur, the value of each pixel is reset to a new value. The formula used is: Among them, T(x,y) is a two-dimensional Gaussian function, γ is the standard deviation, and x and y are pixel values.

5. The method for image noise reduction under extremely low illumination as claimed in claim 1, characterized in that: The S3 specifically includes: Determining the spatial filtering intensity of each pixel in the first target image signal; Extracting local information from the first target image signal to obtain first local information, and determining a joint spatial filtering strength corresponding to each pixel point according to the spatial filtering strength and the first local information; The second target image is spatially filtered according to the edge estimation result to obtain the high-light denoised image.

6. The method for image noise reduction under extremely low illumination as claimed in claim 1, characterized in that: The S4 specifically includes: S41: The attention module is used to automatically identify the blur area of ​​the rough strong light noise reduction image and provide guidance information for the blur area features; S42: The attention driving module further extracts and refines the fuzzy area feature features from shallow to high layers based on the guidance information.

7. A device for image noise reduction under extremely low illumination, the device adopts the method for image noise reduction under extremely low illumination according to any one of claims 1 to 6, characterized in that: The device comprises A shooting module, used to enhance the illumination of the monitoring image environment through a light-emitting device, shoot the monitoring image in at least four directions, and perform weighted summation on pixels in at least four overlapping image blocks in the monitoring image to obtain a first target image; A denoising module, used to remove background noise from the monitoring image using a Gaussian blur filter, and to perform contrast enhancement, and to improve the overall contrast of the monitoring image using a contrast equalization method, so as to obtain a second target image; An edge estimation module, configured to perform edge estimation according to the first target image to obtain an edge estimation result, and perform filtering processing on the second target image according to the edge estimation result to obtain the strong light noise reduction image; The attention module is used to process the missing area of ​​the target image, so as to achieve classification of the target image.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the image noise reduction method according to any one of claims 1 to 6 are implemented.

9. A processor, characterized in that: The processor is used to run a program, wherein the image noise reduction method according to any one of claims 1 to 6 is executed when the program is run.

10. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the image noise reduction method according to any one of claims 1 to 6 are implemented.

Images