CN113487496B - Image denoising method, system and device based on pixel type inference - Google Patents
Image denoising method, system and device based on pixel type inference Download PDFInfo
- Publication number
- CN113487496B CN113487496B CN202110618845.7A CN202110618845A CN113487496B CN 113487496 B CN113487496 B CN 113487496B CN 202110618845 A CN202110618845 A CN 202110618845A CN 113487496 B CN113487496 B CN 113487496B
- Authority
- CN
- China
- Prior art keywords
- pixel
- gradient
- image
- type
- texture
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000012545 processing Methods 0.000 claims abstract description 21
- 238000004364 calculation method Methods 0.000 claims abstract description 11
- 230000011218 segmentation Effects 0.000 claims abstract description 4
- 238000001914 filtration Methods 0.000 claims description 21
- 238000010586 diagram Methods 0.000 claims description 10
- 238000001514 detection method Methods 0.000 claims description 9
- 238000000354 decomposition reaction Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000004891 communication Methods 0.000 claims description 3
- 230000007123 defense Effects 0.000 claims description 3
- 238000003672 processing method Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 5
- 238000004422 calculation algorithm Methods 0.000 description 7
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
Abstract
The invention is suitable for the technical field of image data processing, and provides an image denoising method, an image denoising system and an image denoising device based on pixel type inference, wherein the image denoising method comprises gradient information calculation, texture edge-gradient direction segmentation, pixel type inference and pixel type division; the image processing system comprises a calculation module, a classification module, a denoising module and a reconstruction module; the system device includes a memory and a processor. Therefore, the invention can better keep texture details while improving the denoising effect.
Description
Technical Field
The present invention relates to the field of image data processing technologies, and in particular, to an image denoising method, system and device based on pixel type inference.
Background
Due to the influence of the environment, acquisition equipment and other factors, the image is inevitably polluted by noise in the acquisition, compression and transmission processes, so that the image information is distorted and lost. Image noise can adversely affect the processing tasks of subsequent images (e.g., image segmentation, object recognition, style migration, etc.), and therefore image denoising plays an important role in modern image processing systems.
The purpose of image denoising is to recover a sharp image from a noisy image, which is a classical inverse problem in computer vision. The existing image processing algorithms are many, and common ones include median filtering, gaussian filtering, ROF algorithm, BM3D algorithm, sparse representation algorithm, machine learning, deep learning algorithm and the like, but the algorithms have some defects such as calculation efficiency problem, characteristic loss problem and the like while reducing noise level to a certain extent.
Most of the noise of the image belongs to Gaussian noise, gaussian filtering is linear smoothing filtering, is suitable for eliminating Gaussian noise, but cannot well retain image texture characteristics in Gaussian filtering, so that some scholars propose an adaptive Gaussian filtering algorithm which retains image details to a certain extent, and meanwhile, the computing efficiency and the noise reduction effect are reduced.
In summary, it is clear that the prior art has inconvenience and defects in practical use, so that improvement is needed.
Disclosure of Invention
In view of the above drawbacks, the present invention aims to provide an image denoising method, system and device based on pixel type inference, which can improve denoising effect and better preserve texture details.
The invention provides an image denoising method based on pixel type inference, which comprises the following steps:
step one gradient information calculation
The gradient values of the image in the horizontal direction and the vertical direction are calculated by means of a Sobel operator, and the gradient intensity and the gradient direction of the image are calculated by means of the gradient values in the four directions of horizontal direction, vertical direction, 45 degrees and-45 degrees.
Step two texture edge-gradient direction segmentation
The texture edge and the gradient direction of the image are divided into four directions of horizontal, vertical, 45 degrees and-45 degrees according to the gradient defense line, and an inference basis is provided for the pixel type of the image.
Step three pixel type inference
And deducing the pixel type of the image by adopting the methods of extremum deduction, double-threshold detection and connection method judgment.
Step four sub-pixel type image denoising
And adopting a corresponding image denoising scheme according to the inferred pixel type, and outputting a final image.
According to the image denoising method based on pixel type inference, in the first step, the Sobel operator of 3*3 is selected to determine the gradient strength of the image.
According to the image denoising method based on pixel type inference, in the second step, the pixel types of the image comprise three types, namely, noise pixel determination, texture pixel determination and common pixel determination.
According to the image denoising method based on pixel type inference, in the third step, the pixel type inference flow of the image comprises the following steps:
s1: initializing gradient strength double thresholds TH and TL, and initializing and determining a noise pixel type judgment coefficient P;
s2: according to the gradient direction decomposition diagram, calculating whether f (x, y) is greater than the maximum value of the adjacent pixels along the gradient direction or less than the minimum value of the adjacent pixels along the gradient direction, if so, continuing to execute the operation S3, otherwise, executing the operation S4;
s3: comparing the maximum or minimum of 8 pixels around f (x, y), if the following formula is satisfied, representing that the determined noise type determination coefficient P is satisfied, such pixel type is determined to be the determined noise type, otherwise such pixel is determined to be the determined texture pixel type;
s4: calculating whether G (x, y) is greater than TH, if so, determining to be a certain texture pixel, and if not, executing S5;
step 5: calculating whether G (x, y) is larger than TL, if so, judging as a common pixel type, and if not, executing S6;
s6: a determination is made as to whether a texel is within the (x, y) neighborhood 8 pixel, if so, to determine the texel, and if not, to a normal pixel type.
According to the image denoising method based on pixel type inference, in the fourth step, the processing method of each type of pixels comprises the following steps: the pixels are of a certain texture pixel type, and Gaussian filtering processing is adopted; the pixels are of the common pixel type, a common pixel mean value of a neighborhood window is adopted to replace a determined texture pixel and a determined noise pixel in the neighborhood window, and then Gaussian filtering processing is carried out; the pixel type is a defined texel, and no filtering process is performed.
According to the image denoising method based on pixel type inference of the present invention, the present invention also provides an image processing system, comprising: the computing module is mainly used for computing the gradient value and the gradient direction of the missing pixels; the classification module is mainly used for deducing each pixel type; the denoising module is mainly used for denoising each pixel of the noise image according to the pixel type; and the reconstruction module is mainly used for reconstructing the processed image according to all the denoised pixel values.
According to the system of the invention, the operations executed by the computing module specifically comprise: solving Sobel operators in four directions of horizontal, vertical, 45 degrees and minus 45 degrees; calculating gradient values of horizontal, vertical, 45 degrees and minus 45 degrees; the gradient strength of the pixel is obtained by utilizing gradient values in all directions; obtaining the gradient direction of the pixel; forming a gradient intensity and gradient direction matrix of each pixel.
According to the system of the invention, the operations executed by the classification module specifically comprise: forming a texture direction decomposition diagram according to the gradient strength and the gradient direction; determining gradient strength double thresholds TH and TL and determining a noise pixel type judgment coefficient P; identifying and judging three pixel types by utilizing extremum deduction, double-threshold detection and communication method judgment; a two-dimensional matrix map of pixel types is formed.
According to the system of the invention, the operations executed by the denoising module specifically comprise: determining a Gaussian filter kernel according to the image information; filling related pixels according to pixel types to form a neighborhood window to be filtered; and calculating the filtered pixel value according to the type denoising rule.
According to the system of the present invention, there is also provided a processing apparatus for an image processing system, the apparatus comprising a memory and a processor, the memory being capable of storing a photograph to be denoised, a denoised photograph and a plurality of pieces of program code; the processor is capable of operating the denoising method and the system.
The technical scheme of the invention has the following beneficial effects: increasing gradient intensity performance of noise points and texture points by introducing 45 DEG, -45 DEG gradient information; dividing the texture direction and the gradient direction into four directions of horizontal, vertical, 45 degrees and-45 degrees by means of a texture direction and gradient direction decomposition diagram; the gradient information, the texture direction and the gradient direction are combined, and the inference of three pixel types of determining noise pixels, determining texture pixels and common pixels is realized by means of extremum inference, double-threshold detection and a connection method judgment method. Through the Gaussian filtering method for the determined noise pixels, the filtered output of the determined texture pixels and the Gaussian filtering method for the common pixel neighborhood mean value replacement (the mean value is used for determining the Gaussian pixels and the determined texture pixels), the image blurring is reduced, the image texture details are kept to the greatest extent, and the calculation processing efficiency is improved to a certain extent. To sum up, this patent is through above-mentioned technical scheme, when improving the denoising effect, better reservation texture detail has very strong application spreading value.
Drawings
FIG. 1 is a flow chart of an image denoising method according to the present invention;
FIG. 2 is an exploded view of the texture direction and gradient direction of the present invention;
FIG. 3 is a flow chart of pixel type inference in accordance with the present invention;
FIG. 4 is a flow chart of three types of pixel denoising according to the present invention;
FIG. 5 is a mean-shift determined texel and determined noise pixel Gaussian filter calculation of the present invention;
fig. 6 is a block diagram of an image denoising system according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings in the embodiments. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
As shown in fig. 1, which is a main flowchart of an embodiment of an image denoising method based on pixel type inference, the image denoising method comprises the following steps:
step one gradient information calculation
The gradient values in the horizontal and vertical directions are calculated by means of the Sobel operator, and the gradient intensity and the gradient direction are calculated by means of the gradient values in the four directions of horizontal, vertical, 45 DEG and-45 deg.
The texture of the image can be in different directions, so that the gradient strength is determined by adopting Sobel operators in four directions 3*3 of horizontal, vertical, 45 degrees and-45 degrees, and the Sobel operators in all directions are shown in the following formulas.
The gradient values in the four directions of horizontal, vertical, 45 ° and-45 ° are shown in the following formula:
G x (x,y)＝f(x,y)*Sobel x (x,y)
G y (x,y)＝f(x,y)*Sobel y (x,y)
G 45° (x,y)＝f(x,y)*Sobel 45° (x,y)
G -45° (x,y)＝f(x,y)*Sobel -45° (x,y)
wherein G is x (x,y)、G y (x,y)、G 45° (x,y)、G -45° (x, y) are gradient values in four directions of horizontal, vertical, 45 DEG and-45 DEG respectively; f (x, y) represents the pixel (x, y) gray value.
Gradient intensity of pixel (x, y) as shown in the following formula:
the gradient direction is shown in the formula:
step two texture edge-gradient direction segmentation
The texture edge and the gradient direction are divided into four directions of horizontal, vertical, 45 degrees and-45 degrees according to the gradient defense line, and an inference basis is provided for the pixel type.
For comparison purposes, the present invention divides the texture direction into 4 directions, namely horizontal, vertical, 45 ° and-45 °, and the gradient direction is also divided into four directions (perpendicular to the texture direction). See fig. 2 for an exploded view of the texture direction, gradient direction.
Step three pixel type inference
And deducing the pixel type by adopting methods of extremum deduction, double-threshold detection and connection method judgment.
The invention comprises three pixel types, namely, determining noise pixels, determining texture pixels and common pixels, wherein the common pixel type represents low gradient value and is difficult to judge as a pixel which is not polluted by noise or a pixel type with noise. The three pixel types are inferred by comprehensively utilizing extremum inference, double-threshold detection and connection method judgment, and an inference flow chart is shown in fig. 3, wherein the inference flow comprises the following steps:
s1: the dual thresholds TH, TL for the gradient intensity (TH stands for the set high gradient threshold, TL stands for the set low gradient threshold) are initialized, and the noise pixel type judgment coefficient P is initialized and determined.
S2: according to the gradient direction decomposition diagram, calculating whether f (x, y) is greater than the maximum value of the adjacent pixels along the gradient direction or less than the minimum value of the adjacent pixels along the gradient direction, if so, continuing to execute the operation S3, otherwise, executing the operation S4.
S3: comparing the maxima or minima of 8 pixels around f (x, y), if the following formula is satisfied, represents that the determined noise type decision coefficient P is satisfied, such pixel type is decided as the determined noise type, otherwise such pixel is decided as the determined texel type.
S4: whether G (x, y) is greater than TH is calculated, if so, a decision is made to determine a texel, and if not, S5 is performed.
Step 5: whether G (x, y) is greater than TL is calculated, if so, it is determined to be a normal pixel type, and if not, S6 is performed.
S6: a determination is made as to whether a texel is within the (x, y) neighborhood 8 pixel, if so, to determine the texel, and if not, to a normal pixel type.
Step four-pel type Gaussian filter denoising
For the inferred three pixel types, a corresponding gaussian filtering scheme is employed and the final filtered image is output.
According to the inferred flow, determining that the noise pixel type belongs to a high-intensity noise type, and is possibly located in a texture area or a non-texture area, wherein the noise pixel type is directly processed by Gaussian filtering; the common pixel type comprises pixels which are not polluted by noise and pixels polluted by noise, wherein the determined texture pixels and the determined noise pixels in the neighborhood window have little contribution degree to Gaussian filtering of the pixels and even cause interference, so that the average value of the common pixels in the neighborhood window is adopted to replace the determined texture pixels and the determined noise pixels, and then filtering processing is carried out; the texels are determined and the details are left unfiltered.
Three types of pixel denoising flow charts are shown in fig. 4. Mean replacement the determination of texels and the determination of noise pixels gaussian filtering calculations are seen in fig. 5. Window (2k+1) window Gaussian template (the value of k is 1, 3*3 window is selected), and the calculation formula of each element value in the template is as follows:
the invention also provides a working system based on the image denoising method, referring to fig. 6, the system at least comprises:
1. a calculating module for calculating the horizontal, vertical, 45 DEG, -45 DEG gradient value of each pixel, and calculating the gradient intensity and gradient direction of each pixel according to the gradient value.
In one embodiment, the horizontal, vertical, 45 °, -45 ° gradient values for each pixel are taken according to the Sobel operator; the gradient intensity of each pixel is calculated by adopting a square root method of the square sum of gradient values of horizontal, vertical, 45 degrees and-45 degrees; the gradient direction is obtained by adopting a horizontal gradient value and a vertical gradient value.
2. And the classification module is used for determining noise pixels, determining texture pixels and deducing three pixel types of common pixels.
In one embodiment, a texture (gradient) direction exploded view is formed from the gradient strength and gradient direction; forming a texture (gradient) direction exploded view according to the gradient strength and the gradient direction; identifying and judging three pixel types by utilizing extremum deduction, double-threshold detection and communication method judgment; a two-dimensional matrix map of pixel types is formed.
3. And the denoising module is used for denoising each pixel of the noise image according to the pixel type.
In one embodiment, a Gaussian filter kernel is initialized based on image information; filling related pixels according to pixel types to form a neighborhood window to be filtered; and calculating the filtered pixel value according to the classification processing rule.
4. And the reconstruction module is used for reconstructing the processed image according to all the denoised pixel values.
Further, in one embodiment of a processing apparatus of the present invention, the apparatus comprises a memory capable of storing a photograph to be denoised, a post-denoising photograph, and a plurality of program codes, and a processor capable of running the foregoing image denoising method and running the entire foregoing system.
In summary, in the image processing method based on pixel type inference in the invention, gradient intensity representation of noise points and texture points is increased by introducing 45 DEG, -45 DEG gradient information; dividing the texture direction and the gradient direction into four directions of horizontal, vertical, 45 degrees and-45 degrees by means of a texture direction and gradient direction decomposition diagram; the gradient information, the texture direction and the gradient direction are combined, and the inference of three pixel types of determining noise pixels, determining texture pixels and common pixels is realized by means of extremum inference, double-threshold detection and a connection method judgment method. Through the Gaussian filtering method for the determined noise pixels, the filtered output of the determined texture pixels and the Gaussian filtering method for the common pixel neighborhood mean value replacement (the mean value is used for determining the Gaussian pixels and the determined texture pixels), the image blurring is reduced, the image texture details are kept to the greatest extent, and the calculation processing efficiency is improved to a certain extent. In summary, the beneficial effects of the invention are as follows: image denoising method, system and device based on pixel type inference, and texture details are better reserved while denoising effect is improved.
Of course, the present invention is capable of other various embodiments and its several details are capable of modification and variation in light of the present invention, as will be apparent to those skilled in the art, without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (4)
1. An image denoising method based on pixel type inference, comprising the following steps:
step one gradient information calculation
Calculating gradient values of the image in the horizontal direction and the vertical direction by means of a Sobel operator, and calculating gradient strength and gradient direction of the image by means of the gradient values in the four directions of horizontal direction, vertical direction, 45 degrees and-45 degrees;
step two texture edge-gradient direction segmentation
Dividing the texture edge and the gradient direction of the image into four directions of horizontal, vertical, 45 degrees and-45 degrees according to the gradient defense line to form a texture direction and gradient direction decomposition diagram, providing an inference basis for the pixel types of the image, wherein the pixel types of the image comprise three types, and determining noise pixels, texture pixels and common pixels;
step three pixel type inference
The pixel type of the image is inferred by adopting a combination method of extremum inference, double-threshold detection and connection method judgment, and the pixel type inference flow of the image comprises the following steps:
s1: initializing gradient strength double thresholds TH and TL, initializing and determining a noise pixel type judgment coefficient P, wherein TH represents a set gradient high threshold value, and TL represents a set gradient low threshold value;
s2: according to the gradient direction decomposition diagram, calculating whether f (x, y) is greater than the maximum value of the adjacent pixels along the gradient direction or less than the minimum value of the adjacent pixels along the gradient direction, if so, continuing to execute S3 operation, otherwise, executing S4 operation, wherein f (x, y) represents the pixel pointA gray value;
s3: comparing the maximum or minimum of 8 pixels around f (x, y), if the following formula is satisfied, representing that the determined noise type determination coefficient P is satisfied, such pixel type is determined to be the determined noise type, otherwise such pixel is determined to be the determined texture pixel type;
s4: calculating whether G (x, y) is greater than TH, if so, determining to determine the texture pixel, and if not, executing S5, wherein G (x, y) represents the pixel pointGradient strength of (c);
s5: calculating whether G (x, y) is larger than TL, if so, judging as a common pixel type, and if not, executing S6;
s6: calculating whether a certain texture pixel exists in the (x, y) neighborhood 8 pixels, if so, determining the texture pixel, and if not, determining the common pixel type;
denoising method for step four pixel types
And adopting a corresponding denoising method for the inferred pixel type, and outputting a final image, wherein the processing method for each type of pixel comprises the following steps:
the pixels are of a certain noise pixel type, and Gaussian filtering processing is adopted;
the pixels are of the common pixel type, a common pixel mean value of a neighborhood window is adopted to replace a determined texture pixel and a determined noise pixel in the neighborhood window, and then Gaussian filtering processing is carried out;
the pixel type is a defined texel, and no filtering process is performed.
2. The method of claim 1, wherein in the first step, a Sobel operator of 3*3 is selected to determine the gradient strength of the image.
3. An image processing system for use in the image denoising method of claim 1, comprising:
the computing module is mainly used for computing the gradient value and the gradient direction of each pixel, and the operations executed by the computing module specifically comprise:
solving Sobel operators in four directions of horizontal, vertical, 45 degrees and minus 45 degrees;
calculating gradient values of horizontal, vertical, 45 degrees and minus 45 degrees;
the gradient strength of the pixel is obtained by utilizing gradient values in all directions;
obtaining the gradient direction of the pixel;
forming a gradient intensity and gradient direction matrix of each pixel;
the classifying module is mainly used for deducing each pixel type, and the operations executed by the classifying module specifically comprise:
forming a texture direction decomposition diagram according to the gradient strength and the gradient direction;
determining gradient strength double thresholds TH and TL, determining a noise pixel type judgment coefficient P, wherein TH represents a set gradient high threshold value, and TL represents a set gradient low threshold value;
identifying and judging three pixel types by utilizing extremum deduction, double-threshold detection and communication method judgment;
forming a pixel type two-dimensional matrix diagram;
the denoising module is mainly used for denoising each pixel of the noise image according to the pixel type, and the denoising module specifically performs the operations comprising:
determining a Gaussian filter kernel according to the image information;
filling related pixels according to pixel types to form a neighborhood window to be filtered;
calculating a filtered pixel value according to the type denoising rule;
and the reconstruction module is mainly used for reconstructing the processed image according to all the denoised pixel values.
4. A processing apparatus for the image processing system of claim 3, characterized in that:
the apparatus includes a memory and a processor, the memory capable of storing a photograph to be denoised, a denoised photograph and a plurality of program code; the processor is capable of running the denoising method of claim 1 or 2 and the system of claim 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110618845.7A CN113487496B (en) | 2021-06-03 | 2021-06-03 | Image denoising method, system and device based on pixel type inference |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110618845.7A CN113487496B (en) | 2021-06-03 | 2021-06-03 | Image denoising method, system and device based on pixel type inference |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113487496A CN113487496A (en) | 2021-10-08 |
| CN113487496B true CN113487496B (en) | 2023-09-08 |
Family
ID=77934497
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110618845.7A Active CN113487496B (en) | 2021-06-03 | 2021-06-03 | Image denoising method, system and device based on pixel type inference |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113487496B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109064418A (en) * | 2018-07-11 | 2018-12-21 | 成都信息工程大学 | A kind of Images Corrupted by Non-uniform Noise denoising method based on non-local mean |
| CN111179291A (en) * | 2019-12-27 | 2020-05-19 | 凌云光技术集团有限责任公司 | Edge pixel point extraction method and device based on neighborhood relationship |
| CN112150474A (en) * | 2020-10-12 | 2020-12-29 | 山东省科学院海洋仪器仪表研究所 | Underwater bubble image feature segmentation and extraction method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9852353B2 (en) * | 2014-11-12 | 2017-12-26 | Adobe Systems Incorporated | Structure aware image denoising and noise variance estimation |
-
2021
- 2021-06-03 CN CN202110618845.7A patent/CN113487496B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109064418A (en) * | 2018-07-11 | 2018-12-21 | 成都信息工程大学 | A kind of Images Corrupted by Non-uniform Noise denoising method based on non-local mean |
| CN111179291A (en) * | 2019-12-27 | 2020-05-19 | 凌云光技术集团有限责任公司 | Edge pixel point extraction method and device based on neighborhood relationship |
| CN112150474A (en) * | 2020-10-12 | 2020-12-29 | 山东省科学院海洋仪器仪表研究所 | Underwater bubble image feature segmentation and extraction method |
Non-Patent Citations (1)
| Title |
|---|
| 岑红.计算机视觉技术的图像识别与复原研究.《哈尔滨理工大学学报》.2020,142-146. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113487496A (en) | 2021-10-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108921800B (en) | Non-local mean denoising method based on shape self-adaptive search window | |
| CN108765325B (en) | Small unmanned aerial vehicle blurred image restoration method | |
| CN112950508A (en) | Drainage pipeline video data restoration method based on computer vision | |
| CN108510451B (en) | Method for reconstructing license plate based on double-layer convolutional neural network | |
| CN108154519A (en) | Dividing method, device and the storage medium of eye fundus image medium vessels | |
| CN110675340A (en) | Single image defogging method and medium based on improved non-local prior | |
| CN109509163B (en) | FGF-based multi-focus image fusion method and system | |
| CN107798670B (en) | Dark channel prior image defogging method using image guide filter | |
| Zhu et al. | Edge-preserving guided filtering based cost aggregation for stereo matching | |
| CN110349112B (en) | Two-stage image denoising method based on self-adaptive singular value threshold | |
| CN115131351B (en) | Engine oil radiator detection method based on infrared image | |
| CN109214993B (en) | Visual enhancement method for intelligent vehicle in haze weather | |
| CN112150371A (en) | Image noise reduction method, device, equipment and storage medium | |
| CN113793278A (en) | Improved remote sensing image denoising method with minimized weighted nuclear norm and selectively enhanced Laplace operator | |
| Yousaf et al. | Single image dehazing and edge preservation based on the dark channel probability-weighted moments | |
| CN116152115A (en) | Garbage image denoising processing method based on computer vision | |
| CN113421210B (en) | Surface point Yun Chong construction method based on binocular stereoscopic vision | |
| CN112288726B (en) | Method for detecting foreign matters on belt surface of underground belt conveyor | |
| Das et al. | Histogram equalization techniques for contrast enhancement: a review | |
| CN113487496B (en) | Image denoising method, system and device based on pixel type inference | |
| Htet et al. | The edges detection in images using the clustering algorithm | |
| CN114723663A (en) | Preprocessing defense method aiming at target detection and resisting attack | |
| CN115035397A (en) | Underwater moving target identification method and device | |
| CN103810677A (en) | Remote sensing image enhancement method based on super-empirical mode decomposition | |
| Fang et al. | An improved DCP-based image defogging algorithm combined with adaptive fusion strategy |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |