CN108460745A - A kind of image de-noising method based on non-local mean filtering - Google Patents
A kind of image de-noising method based on non-local mean filtering Download PDFInfo
- Publication number
- CN108460745A CN108460745A CN201810272837.XA CN201810272837A CN108460745A CN 108460745 A CN108460745 A CN 108460745A CN 201810272837 A CN201810272837 A CN 201810272837A CN 108460745 A CN108460745 A CN 108460745A
- Authority
- CN
- China
- Prior art keywords
- natural image
- image
- noisy
- pixel
- noisy natural
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000001914 filtration Methods 0.000 title claims abstract description 20
- 238000012545 processing Methods 0.000 claims abstract description 32
- 239000000284 extract Substances 0.000 claims abstract description 5
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 230000003287 optical effect Effects 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 230000000007 visual effect Effects 0.000 claims description 3
- 229910017435 S2 In Inorganic materials 0.000 claims 1
- 239000011159 matrix material Substances 0.000 claims 1
- 238000003384 imaging method Methods 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 11
- 238000005516 engineering process Methods 0.000 description 7
- 230000005540 biological transmission Effects 0.000 description 4
- 238000007781 pre-processing Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000005672 electromagnetic field Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000003631 expected effect Effects 0.000 description 1
- 230000010358 mechanical oscillation Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000000518 rheometry Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
Abstract
The invention discloses a kind of image de-noising methods based on non-local mean filtering, include the following steps：Image signal acquisition unit obtains and exports noisy natural image signal；Processing unit is pre-processed to obtain noisy natural image data to the noisy natural image signal；Original non-local mean filtering is carried out to noisy natural image data, obtains a filter result figure and method noise pattern；Region belonging to each pixel in method noise pattern is different, extracts image residue information therein, obtains the residual, information figure extracted from method noise pattern；It sums to a filter result figure and residual, information figure, obtains denoising reference chart；The gray value for replacing all pixels point in noisy natural image with the estimated value for all pixels point being calculated, obtains denoising image.The present invention improves signal noise ratio (snr) of image level, is simply easily achieved, will not substantially increase the area of imaging sensor.
Description
Technical field
The present invention relates to a kind of image de-noising methods based on non-local mean filtering.
Background technology
In the modern life, with the extensive use of computer networking technology, the fast development of Computer Multimedia Technology
With the foundation of wideband information network, information takes on more and more important role in people’s lives, study and work.Voice and
Image is the main media that the mankind transmit information, and according to statistics, in the information that the mankind receive, auditory information accounts for 20%, vision letter
Breath account for 60%, wherein the most important and direct information is exactly image information, it is contained much information with it, and transmission speed is fast, effect away from
From the series of advantages such as remote, becomes the mankind and obtain the important sources of information and the important means using information.And in piece image
Including intuitive and information content be word, sound it is incomparable.However, image obtain and transmission process in, no
Ground be can avoid by interference outwardly and inwardly, be usually added into many noises.The factor of picture noise is caused generally to have：It is external
Interference enters external noise caused by internal system, electronic original part or sensor internal load particle through electromagnetic wave or power supply string
Random motion be formed by internal noise, the electromagnetic field variation caused by the mechanical oscillation of certain components or electricity inside electric appliance
Noise caused by rheology, the interference of transmission channel in image transmitting process and the noise and the human factor that are formed and cause
Noise.
The presence of noise has damaged the quality of image, and image is made to become to obscure very much, has seriously affected the vision effect of image
Fruit, or even the feature of image is masked, thus directly affect the processing of pictures subsequent.Therefore, it is necessary to digital picture into
Row pretreatment work carries out denoising to image, ensures that image reaches people's expected effect when in use.Image preprocessing is main
Including：Image recovery, characteristics of image enhancing, image denoising etc., image denoising is a kind of important technology in image preprocessing.It is logical
Picture quality can be effectively improved by crossing Image Denoising Technology, increased signal-to-noise ratio, preferably embodied the letter entrained by original image
Breath, image denoising are that subsequent Digital Image Processing is had laid a good foundation as a kind of important preprocessing means.
In image processing process, denoising is carried out to image and repairs pretreatment operation seeming very necessary, it can be reached
To the purpose for improving picture quality, assisting subsequently carrying out higher level processing.Denoising Algorithm more common at present, although going
Possess good effect in terms of except noise, but is easily damaged the notable structure of image, and more general reparation algorithm, matching
Rate is low, easy tos produce texture superposition phenomenon, it is not high to result in picture quality.A. the non-local mean filter that Baudes et al. is proposed
Wave (Non-local means, NLM) algorithm, it has self-similarity using natural image, pending image is made to achieve denoising
The good result of aspect.Non-local mean filtering algorithm has been widely applied to many digital image processing fields at present, such as
The denoising and reparation of image and video, textures synthesis, the fields such as super-resolution.
The algorithm idea of mean filter is replaced with the average value of several neighborhood territory pixel gray values centered on pending point
The value for changing currently pending pixel, to obtain the image after denoising.This method can remove partial noise, but denoising effect is inadequate
It is ideal.Non-local mean filters the gray value in more each pixel in image overall region, utilizes the non local self similarity of image
Property, according to the similitude of image weight is distributed to each pixel.The algorithm can preferably promote denoising effect, and effectively
Retain image detail and edge.
Image is easy in acquisition and its transmission process due to much noise introducing and by pollution, along with artificial
Or other natural causes, so that information is lost or is damaged, it will the serious vision for reducing picture quality and influencing image
Effect, or even interfere and misled the normal identification of the mankind.Although including less noise or having the image of tiny breakage still
With the ability for showing its original image structure, but in terms of some scientific researches and medical treatment, such as analysis space probation image,
Weather nephogram, map explorer satellite image, SAR images, traffic monitoring image etc., interference and the certain image informations of noise
Loss will be brought with serious catastrophic consequence, and it is very must to remove the mixed noise of image and image processing work thus
It wants.This allows for Image Denoising Technology as the research direction of hot spot in current Digital Image Processing and computer vision.
Though non-local mean is filtered by years development, and achieves many very important achievements in research.But it is improved
The potentiality for promoting denoising performance are still very big, and still there is some problems to be solved.Therefore, for the above present situation, compel
A kind of image de-noising method filtered based on non-local mean will be developed by being essential, to overcome the shortcomings of in currently practical application.
Invention content
The purpose of the present invention is to provide a kind of image de-noising methods based on non-local mean filtering, to solve the above-mentioned back of the body
The problem of being proposed in scape technology.
To achieve the above object, the present invention provides the following technical solutions：
A kind of image de-noising method based on non-local mean filtering, includes the following steps：
S1, image signal acquisition unit obtain and export noisy natural image signal；
S2, processing unit receive the noisy natural image signal and are pre-processed to obtain to the noisy natural image signal noisy
Natural image data；
S3 carries out original non-local mean filtering to noisy natural image data, obtains a filter result figure and method noise
Figure；
S4 judges the region belonging to its each pixel, i.e., to current in the noisy natural image data to method noise pattern
Pixel carries out the calculating of SOBEL operators to obtain the SOBEL operators of the current pixel point, and to the SOBEL of the current pixel point
Operator is corrected, and judges whether the current pixel point needs to carry out denoising according to the SOBEL operators after correction, if so,
S5 is entered step, if it is not, the processing unit judges whether next pixel in the noisy natural image data needs
It makes an uproar processing；
S5, region belonging to each pixel in method noise pattern is different, extracts image residue information therein, i.e., for
Belong to some pixel of the method noise pattern of smooth region, carries out 3 × 3 mean filters, the method for belonging to details area
Some pixel of noise pattern finds out 9 most adjacent pixels from 9 × 9 neighborhoods, is made with the average value of this 9 pixels
For the gray value of the pixel, the residual, information figure extracted from method noise pattern is obtained；
S6 sums to a filter result figure and residual, information figure, obtains denoising reference chart；
S7 is acquired new weight w ref using the weights formula of non-local mean in denoising reference chart, utilizes new weights
Wref carries out non-local mean filtering in noisy natural image, obtains the estimated value of each pixel；
S8 replaces the gray value of all pixels point in noisy natural image with the estimated value for all pixels point being calculated, obtains
To denoising image.
S9 realizes that whether is verification algorithm to algorithm by Visual studio2012 softwares and the libraries opencv into line code
Scheduled result can be reached.
Further, in step sl, which includes photosensitive pixel array and noisy natural image
Signal sensing element；The photosensitive pixel array converts optical signals to the noisy natural image signal；And the noisy natural image
Signal sensing element controls the photosensitive pixel array and the optical signal is converted to the noisy natural image signal, and it is noisy to export this
Natural image signal.
Further, in step s 2, which includes analogy signal processing unit, AD conversion unit, number
Signal processing unit and output unit；Carrying out pretreated step includes：The analogy signal processing unit receives the noisy nature
The noisy natural image signal of picture signal sensing element output, and the noisy natural image signal after processing is exported,
The noisy natural image signal that the analogy signal processing unit exports is converted to the digitized figure by the AD conversion unit
As data.
Compared with prior art, the beneficial effects of the invention are as follows：The image de-noising method based on non-local mean filtering,
The details (such as edge) and flat site (the smaller pixel of SOBEL operators of image can be relatively accurately distinguished by SOBEL operators
Point), at details, is carried out by denoising, i.e., carries out the same of denoising to image for flat site without denoising
When can retain image detail as much as possible, it is horizontal to improve signal noise ratio (snr) of image, is simply easily achieved, will not substantially increase image
The area of sensor.New image residue method for extracting signal proposed by the present invention significantly more efficient can be extracted and be made an uproar in removal
The image residue information lost while sound, is conducive to the raising of denoising effect.The present invention compared with non-local mean method,
The new weights being calculated in denoising reference chart are more accurate, can preferably reach denoising effect.
Specific implementation mode
The technical scheme in the embodiments of the invention will be clearly and completely described below, it is clear that described implementation
Example is only a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, this field is common
The every other embodiment that technical staff is obtained without making creative work belongs to the model that the present invention protects
It encloses.
In the embodiment of the present invention, a kind of image de-noising method based on non-local mean filtering includes the following steps：
S1, image signal acquisition unit obtain and export noisy natural image signal；
S2, processing unit receive the noisy natural image signal and are pre-processed to obtain to the noisy natural image signal noisy
Natural image data；
S3 carries out original non-local mean filtering to noisy natural image data, obtains a filter result figure and method noise
Figure；
S4 judges the region belonging to its each pixel, i.e., to current in the noisy natural image data to method noise pattern
Pixel carries out the calculating of SOBEL operators to obtain the SOBEL operators of the current pixel point, and to the SOBEL of the current pixel point
Operator is corrected, and judges whether the current pixel point needs to carry out denoising according to the SOBEL operators after correction, if so,
S5 is entered step, if it is not, the processing unit judges whether next pixel in the noisy natural image data needs
It makes an uproar processing；
S5, region belonging to each pixel in method noise pattern is different, extracts image residue information therein, i.e., for
Belong to some pixel of the method noise pattern of smooth region, carries out 3 × 3 mean filters, the method for belonging to details area
Some pixel of noise pattern finds out 9 most adjacent pixels from 9 × 9 neighborhoods, is made with the average value of this 9 pixels
For the gray value of the pixel, the residual, information figure extracted from method noise pattern is obtained；
S6 sums to a filter result figure and residual, information figure, obtains denoising reference chart；
S7 is acquired new weight w ref using the weights formula of non-local mean in denoising reference chart, utilizes new weights
Wref carries out non-local mean filtering in noisy natural image, obtains the estimated value of each pixel；
S8 replaces the gray value of all pixels point in noisy natural image with the estimated value for all pixels point being calculated, obtains
To denoising image.
S9 realizes that whether is verification algorithm to algorithm by Visual studio2012 softwares and the libraries opencv into line code
Scheduled result can be reached.
Further, in step sl, which includes photosensitive pixel array and noisy natural image
Signal sensing element；The photosensitive pixel array converts optical signals to the noisy natural image signal；And the noisy natural image
Signal sensing element controls the photosensitive pixel array and the optical signal is converted to the noisy natural image signal, and it is noisy to export this
Natural image signal.
Further, in step s 2, which includes analogy signal processing unit, AD conversion unit, number
Signal processing unit and output unit；Carrying out pretreated step includes：The analogy signal processing unit receives the noisy nature
The noisy natural image signal of picture signal sensing element output, and the noisy natural image signal after processing is exported,
The noisy natural image signal that the analogy signal processing unit exports is converted to the digitized figure by the AD conversion unit
As data.
The image de-noising method based on non-local mean filtering, image can be relatively accurately distinguished by SOBEL operators
Details (such as edge) and flat site (the smaller pixel of SOBEL operators), it is right without denoising at details
In flat site, denoising is carried out, i.e., can retain image detail while carrying out denoising to image as much as possible, improves figure
As signal noise ratio level, simply it is easily achieved, will not substantially increases the area of imaging sensor.New image proposed by the present invention
Residue signal extracting method significantly more efficient can extract the image residue information lost while removing noise, be conducive to
The raising of denoising effect.The present invention is compared with non-local mean method, and the new weights being calculated in denoising reference chart are more
It is accurate to add, and can preferably reach denoising effect.
The above are merely the preferred embodiment of the present invention, it is noted that for those skilled in the art, not
Under the premise of being detached from present inventive concept, several modifications and improvements can also be made, these should also be considered as the protection model of the present invention
It encloses, these all do not interfere with the effect and patent practicability that the present invention is implemented.
Claims (4)
1. a kind of image de-noising method based on non-local mean filtering, which is characterized in that include the following steps：
S1, image signal acquisition unit obtain and export noisy natural image signal；
S2, processing unit receive the noisy natural image signal and are pre-processed to obtain to the noisy natural image signal noisy
Natural image data；
S3 carries out original non-local mean filtering to noisy natural image data, obtains a filter result figure and method noise
Figure；
S4 judges the region belonging to its each pixel, i.e., to current in the noisy natural image data to method noise pattern
Pixel carries out the calculating of SOBEL operators to obtain the SOBEL operators of the current pixel point, and to the SOBEL of the current pixel point
Operator is corrected, and judges whether the current pixel point needs to carry out denoising according to the SOBEL operators after correction, if so,
S5 is entered step, if it is not, the processing unit judges whether next pixel in the noisy natural image data needs
It makes an uproar processing；
S5, region belonging to each pixel in method noise pattern is different, extracts image residue information therein, i.e., for
Belong to some pixel of the method noise pattern of smooth region, carries out 3 × 3 mean filters, the method for belonging to details area
Some pixel of noise pattern finds out 9 most adjacent pixels from 9 × 9 neighborhoods, is made with the average value of this 9 pixels
For the gray value of the pixel, the residual, information figure extracted from method noise pattern is obtained；
S6 sums to a filter result figure and residual, information figure, obtains denoising reference chart；
S7 is acquired new weight w ref using the weights formula of non-local mean in denoising reference chart, utilizes new weights
Wref carries out non-local mean filtering in noisy natural image, obtains the estimated value of each pixel；
S8 replaces the gray value of all pixels point in noisy natural image with the estimated value for all pixels point being calculated, obtains
To denoising image.
2.S9 realizes algorithm into line code verification algorithm whether can by Visual studio2012 softwares and the libraries opencv
Reach scheduled result.
3. the image de-noising method according to claim 1 based on non-local mean filtering, which is characterized in that in step S1
In, which includes photosensitive pixel array and noisy natural image signal sensing element；The light-sensitive image primitive matrix
Row convert optical signals to the noisy natural image signal；And the noisy natural image signal sensing element controls the photosensitive pixel
The optical signal is converted to the noisy natural image signal by array, and exports the noisy natural image signal.
4. the image de-noising method according to claim 1 based on non-local mean filtering, which is characterized in that in step S2
In, which includes analogy signal processing unit, AD conversion unit, digital signal processing unit and output unit；Into
The pretreated step of row includes：The analogy signal processing unit receive noisy natural image signal sensing element output this contain
It makes an uproar natural image signal, and exports the noisy natural image signal after processing, the AD conversion unit is by the analog signal
The noisy natural image signal of processing unit output is converted to the digitized image data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810272837.XA CN108460745A (en) | 2018-03-29 | 2018-03-29 | A kind of image de-noising method based on non-local mean filtering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810272837.XA CN108460745A (en) | 2018-03-29 | 2018-03-29 | A kind of image de-noising method based on non-local mean filtering |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108460745A true CN108460745A (en) | 2018-08-28 |
Family
ID=63238208
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810272837.XA Pending CN108460745A (en) | 2018-03-29 | 2018-03-29 | A kind of image de-noising method based on non-local mean filtering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108460745A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102930508A (en) * | 2012-08-30 | 2013-02-13 | 西安电子科技大学 | Image residual signal based non-local mean value image de-noising method |
| CN103606132A (en) * | 2013-10-31 | 2014-02-26 | 西安电子科技大学 | Multiframe digital image denoising method based on space domain and time domain combination filtering |
| CN106469436A (en) * | 2015-08-17 | 2017-03-01 | 比亚迪股份有限公司 | Image denoising system and image de-noising method |
-
2018
- 2018-03-29 CN CN201810272837.XA patent/CN108460745A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102930508A (en) * | 2012-08-30 | 2013-02-13 | 西安电子科技大学 | Image residual signal based non-local mean value image de-noising method |
| CN103606132A (en) * | 2013-10-31 | 2014-02-26 | 西安电子科技大学 | Multiframe digital image denoising method based on space domain and time domain combination filtering |
| CN106469436A (en) * | 2015-08-17 | 2017-03-01 | 比亚迪股份有限公司 | Image denoising system and image de-noising method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xiao et al. | Making of night vision: Object detection under low-illumination | |
| CN107895376A (en) | Based on the solar panel recognition methods for improving Canny operators and contour area threshold value | |
| CN110072051A (en) | Image processing method and device based on multiple image | |
| CN102750683B (en) | Filtering method for sea surface stripe noise and sea surface stripe cloud in moderate resolution imaging spectroradiometer (MODIS) remote sensing image | |
| CN101710385B (en) | High-performance two-dimensional bar code decoding method | |
| CN110310238A (en) | A kind of removing rain based on single image method based on the compression rewards and punishments neural network for reusing raw information | |
| CN104484652A (en) | Method for fingerprint recognition | |
| CN108833756A (en) | A kind of outdoor energy-saving monitoring camera | |
| CN114821343A (en) | Mangrove remote sensing rapid and accurate extraction method based on cloud platform | |
| CN106303296A (en) | A kind of image mosaic emerging system | |
| CN109688382A (en) | A kind of underwater picture processing system of underwater robot | |
| CN111583113A (en) | Infrared image super-resolution reconstruction method based on generation countermeasure network | |
| CN108460745A (en) | A kind of image de-noising method based on non-local mean filtering | |
| CN107341809A (en) | A kind of coastline towards multi-source Remote Sensing Images automatically extracts and bias correction method | |
| CN109118450B (en) | Low-quality image enhancement method under sand weather condition | |
| CN110321378A (en) | A kind of mobile monitor image identification system and method | |
| CN113592750A (en) | Infrared enhancement method based on gradient histogram | |
| CN112834526A (en) | Optical fiber end face defect detection device and method for visual Internet of things | |
| CN110956090A (en) | Face feature extraction method based on positive and negative illumination invariant unit fusion | |
| CN116128766A (en) | Improved Retinex-Net-based infrared image enhancement method for power equipment | |
| CN102082956A (en) | Method for enhancing sensitivity of image sensor | |
| CN105472336B (en) | A kind of low power image compression module | |
| CN114140366A (en) | Infrared image and visible light image fusion method, system, equipment and terminal | |
| CN116468615A (en) | Image processing method, device, electronic equipment and storage medium | |
| CN113160788A (en) | Reverse-phase noise reduction platform based on waveform analysis |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180828 |
|
| WD01 | Invention patent application deemed withdrawn after publication |