CN104346786B - Image denoising algorithm based on Demons algorithm - Google Patents
Image denoising algorithm based on Demons algorithm Download PDFInfo
- Publication number
- CN104346786B CN104346786B CN201410621477.1A CN201410621477A CN104346786B CN 104346786 B CN104346786 B CN 104346786B CN 201410621477 A CN201410621477 A CN 201410621477A CN 104346786 B CN104346786 B CN 104346786B
- Authority
- CN
- China
- Prior art keywords
- algorithm
- image
- denoising
- demons
- gradient
- 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
Abstract
The invention relates to an image denoising algorithm based on a Demons algorithm. The method comprises the following steps: firstly, on the basis of the Demons algorithm, regarding a diffusion process as image registration, and establishing a new Demons denoising algorithm based on image registration, wherein the PM algorithm, with relatively classical denoising performance, of the algorithm is preferable; secondly, in view of that local feature depending on gradient information representation images is insufficient in an image denoising process and second-order differential quantities contain richer information, taking a level set curvature as a driving force factor for controlling an image structure to be introduced in an image registration denoising algorithm, and establishing an image denoising algorithm of gradient and curvature dual driving force, namely a dual driving algorithm; finally, adopting an additive operator splitting algorithm (AOS algorithm) to process the algorithm to obtain an image after denoising. The denoising performance is superior, and the integral structure of the image is kept intact, the image SNR (Signal to Noise Ratio) after denoising is improved by about 15 dB compared with other Demons algorithms and is improved by about 25 dB compared with the PM algorithm, and the definition is also greatly promoted.
Description
Technical field
The present invention relates to image processing field, is that, based on the Image denoising algorithm of partial differential equation, the algorithm is based on
Demons algorithm improvements.
Background technology
Digital picture is the source that many ambits obtain information, but image in gatherer process often because each side
Face reason introduces noise.Therefore, in image procossing and computer realm, image denoising is one of most basic problem.Nearly tens
Year, Partial Differential Equation method starts to be widely used in image procossing, the aspect such as denoising, segmentation, rim detection, enhancing in image
Research all make remarkable progress.
So far, researcher has been proposed for many anisotropy parameter algorithms, wherein most classical is order conduction
Coefficient depends on the Second Order Partial differential algorithm of image gradient, i.e. PM algorithms.With deepening continuously to the technical research, Hen Duoshi
Test result to show, PM algorithm existing defects:First, it cannot correctly distinguish edge and noise sometimes, thus process little yardstick
The noise effects in space are undesirable；Second, the correctness of diffusion coefficient function is verified without theory.Therefore, many scholars couple
Diffusion coefficient is improved in PM algorithms, attempts to set up and more effectively protect edge smoothing filter, and achieve some
Exhibition.But most of modified hydrothermal process is all built upon on classic algorithm, without the defect for fundamentally solving PM algorithms.
The content of the invention
It is contemplated that the image for becoming apparent from is processed out, to be close to original image.Because traditional algorithm is ageing low, answer
Miscellaneous degree is higher, and often occurs excess smoothness and smooth insufficient phenomenon in processing procedure, and this algorithm is calculated in Demons
It is improved on the basis of method, makes image anisotropy denoising process simplification into process of image registration, it is not necessary to considers design diffusion
Function, Grads threshold and the problems such as distinguish edge, reduce algorithm complex and to improve algorithm ageing, therefore simplify
Image denoising problem, efficiently make use of overall structure function to control picture structure.Denoising performance aspect, Y-PSNR and clear
Clear degree is greatly improved.
The present invention is employed the following technical solutions：A kind of Image denoising algorithm based on Demons algorithms, it is characterised in that bag
Include following steps：
Step one, noise image is carried out Gauss filtering, remove larger noise；
Step 2, using Demons algorithm principles, set up based on the Image denoising algorithm of Demons algorithms
Wherein div, ▽ are respectively divergence operator and gradient operator, I I0It is respectively
Initial pictures and plus make an uproar image, | ▽ I | are gradient modulus value；
Step 3, by level set curvatureDemons is incorporated into as correction driven factor
In image denoising model, structure function is set upFinally setting up Dual Drive Denoising Algorithm isWherein | ▽ I |, | ▽ f | are gradient modulus value, α
It is coherent coefficient with β, is determined by curve matching；
Step 4, with half implicit expression additive operator division numerical algorithm discretization is carried out to Dual Drive Denoising Algorithm in step 3
Process, obtain image after denoising, process is as follows：
Jing abbreviations,
WithRepresent I (xi,yj；tn),f(xi,yj；tn), its implied format is
Wherein Δ t be time step, Ax、AyThe corresponding coefficient matrix of difference is respectively carried out to x and y directions,
Solve In+1, fn+1：
Deformation can be obtained
1) work as i=1 ..., during N, calculateWithElement on three diagonal： With And solved using chasing methodObtain
2) j=1 ... is worked as, it is same to calculate during MWithThree diagonal on element, ask
SolutionObtain
3) calculate
4) repeatedly 1)~3), through successive ignition, obtain picture rich in detail.
Above-mentioned Demons algorithms areWherein F is reference picture, and M is floating image,Displacement needed for (x, y) place from from floating image to reference picture deformation, div, ▽ be respectively divergence operator and
Gradient operator.
The beneficial effect that the present invention reaches：1. in terms of the complexity of algorithm, the information content of needs is few, and method is simple, no
The local detail information of analysis of the image emphatically is needed, only need to be considered from the overall structure of image, set up overall structure function to control
Imaged structure, realizes image denoising, makes contaminated image closer to original image；2. at the ageing aspect of algorithm, because
The information content that the present invention needs is few, and the complexity of enforcement is low, so as to reduce the process time of algorithm；3. in the denoising of algorithm
Can aspect, the Y-PSNR and definition of image greatly improve, by noise pollution image Jing after this algorithm process more
Close original image.
Description of the drawings
Fig. 1 is to set up the flow chart based on the Image denoising algorithm of Demons algorithms.
Specific embodiment
In order that the objects, technical solutions and advantages of the present invention become more apparent, it is right below in conjunction with drawings and Examples
The present invention is further elaborated.It should be appreciated that specific embodiment described herein is only to explain the present invention, and
It is not used in the restriction present invention.
As shown in figure 1, noise image is pre-processed, that is, Gauss filtering is carried out, remove larger noise, next sentenced
Whether disconnected picture structure deforms, that is, consider gradient | the ▽ I | of noisy image when doing rim detection, due to by noise jamming so that
Detection to image border, texture is not accurate enough, and the detailed information such as the Edge texture of image itself are destroyed during denoising, and
The structure for making image deforms.
(1) in the case where picture structure does not deform, then whether all pixels of detection image lead directly to, if all
Pixel leads directly to, then obtain and export clearly image；It is straight-through if not all pixels, then return to noise image pretreatment stage
Processed.
(2) picture structure have deformation in the case of, due to image level set curvature in the presence of noise, can
The detailed information such as the Edge texture of effective detection image, therefore the structure control function of level set curvature is set up, for maintaining image
Structure, be allowed to not deform.By level set curvatureIntroduce as correction driven factor
To Demons image denoising modelsIn, set up structure function
Finally setting up Dual Drive Denoising Algorithm is
Next discretization is carried out to Dual Drive Denoising Algorithm with half implicit expression additive operator division numerical algorithm (AOS algorithms)
Process.Jing abbreviations, obtainWith
WithRepresent I (xi,yj；tn),f(xi,
yj；tn), its implied format is respectivelyWherein Δ t is
Time step, Ax、AyThe corresponding coefficient matrix of difference is respectively carried out to x and y directions.
Solve In+1, fn+1：
Deformation can be obtained
1) work as i=1 ..., during N, calculateWithElement on three diagonal： With SolveObtain
2) j=1 ... is worked as, it is same to calculate during MWithThree diagonal on element,
SolveObtain
3) calculate
4) repeatedly 1)~3), through successive ignition, until obtaining picture rich in detail.
Whether all pixels lead directly in last detection image, are to export picture rich in detail, otherwise return noise image pre-
Process step.
Not only denoising performance is superior for the present invention, and the overall structure of image is remained intact, the signal noise ratio (snr) of image after denoising
Compared with Demons denoisings, other algorithms improve 15dB or so, and compared with PM algorithms 25dB or so is improve, and definition is also significantly lifted.
It is more than the better embodiment of the present invention, but protection scope of the present invention not limited to this.It is any to be familiar with this area
Technical staff disclosed herein technical scope in, the conversion expected without creative work or replacement all should be covered
Within protection scope of the present invention.Therefore the protection domain that protection scope of the present invention should be limited by claim is defined.
Claims (2)
1. a kind of image de-noising method based on Demons algorithms, it is characterised in that comprise the following steps：
Step one, noise image is carried out Gauss filtering, remove larger noise；
Step 2, using Demons algorithm principles, set up based on the Image denoising algorithm of Demons algorithmsWherein div,Respectively divergence operator and gradient operator, I, I0Respectively be plus make an uproar image and
Initial pictures,For gradient modulus value；
Step 3, by level set curvatureDemons images are incorporated into as correction driven factor
In denoising model, structure function is set upFinally setting up Dual Drive Denoising Algorithm isWhereinFor gradient modulus value, α and β
It is coherent coefficient；
Step 4, with half implicit expression additive operator division numerical algorithm sliding-model control is carried out to algorithm in step 3, obtain denoising
Image afterwards, process is as follows：
Jing abbreviations,
WithSubstitute I (xi,yj；tn),f(xi,yj；tn), its implied format is
Wherein, Δ t is time step, Al,A′lRefer respectively to regard to In、fnAlong x, the coefficient matrix general name in y-axis direction.；
Solve In+1, fn+1：
Deformation can be obtained
1) i=1 is worked as, during N, calculateWithElement on three diagonal： With
And solved using chasing method Obtain
2) j=1 is worked as, it is same to calculate during MWithThree diagonal on element, ask
SolutionObtain
3) calculate
4) repeatedly 1)~3), through successive ignition, obtain picture rich in detail.
2. the image de-noising method based on Demons algorithms according to claim 1, it is characterised in that the Demons is calculated
Method isWherein F is reference picture, and M is floating image,It is (x, y) place from floating
Displacement needed for from motion video to reference picture deformation, div,Respectively divergence operator and gradient operator.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410621477.1A CN104346786B (en) | 2014-11-06 | 2014-11-06 | Image denoising algorithm based on Demons algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410621477.1A CN104346786B (en) | 2014-11-06 | 2014-11-06 | Image denoising algorithm based on Demons algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104346786A CN104346786A (en) | 2015-02-11 |
| CN104346786B true CN104346786B (en) | 2017-05-17 |
Family
ID=52502330
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410621477.1A Active CN104346786B (en) | 2014-11-06 | 2014-11-06 | Image denoising algorithm based on Demons algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104346786B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108428216B (en) * | 2018-01-16 | 2021-09-24 | 辽宁师范大学 | Second-order partial differential equation remote sensing image denoising method based on scatter matrix characteristics |
| CN109840896B (en) * | 2019-02-28 | 2023-06-16 | 华南理工大学 | Image denoising method based on gradient and self-adaptive curvature characteristics |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103236059A (en) * | 2013-04-25 | 2013-08-07 | 深圳先进技术研究院 | Diffeomorphism demons image registration method and system based on mode transformation |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7778488B2 (en) * | 2007-03-23 | 2010-08-17 | Varian Medical Systems International Ag | Image deformation using multiple image regions |
| WO2012155136A2 (en) * | 2011-05-12 | 2012-11-15 | The Johns Hopkins University | Method and system for registering images |
-
2014
- 2014-11-06 CN CN201410621477.1A patent/CN104346786B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103236059A (en) * | 2013-04-25 | 2013-08-07 | 深圳先进技术研究院 | Diffeomorphism demons image registration method and system based on mode transformation |
Non-Patent Citations (4)
| Title |
|---|
| "Image matching as a diffusion process：an analogy with Maxwell"s demons";J.-P.Thirion;《Medical image analysis》;20020429;第2卷(第3期);第243-260页 * |
| "图像去噪中MCM方程的改进AOS数值方案";张奕 等;《中国科技论文在线》;20091231;第1-8页 * |
| "基于加性算子分割的快速静磁场主动轮廓模型";张宁 等;《计算机工程与应用》;20110222;第48卷(第5期);第179-182页 * |
| "改进Demons算法的非刚性医学图像配准";张红颖 等;《光学精密工程》;20070131;第15卷(第1期);第145-150页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104346786A (en) | 2015-02-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2018086289A1 (en) | Background suppression method in automatic optical inspection of display panel, and inspection device | |
| CN104715461B (en) | Image de-noising method | |
| CN105117727B (en) | A kind of Location Method of Vehicle License Plate | |
| CN107292842B (en) | Image deblurring method based on prior constraint and outlier suppression | |
| CN102800094A (en) | Fast color image segmentation method | |
| CN102663702A (en) | Natural image denoising method based on regional division | |
| CN108830857B (en) | Self-adaptive Chinese character copy label image binarization segmentation method | |
| DE112019007550T5 (en) | AUTOMATICALLY SEGMENT AND ADJUST IMAGES | |
| CN104346786B (en) | Image denoising algorithm based on Demons algorithm | |
| WO2022001464A1 (en) | Automatic matting method and system | |
| CN114331869A (en) | Dam face crack semantic segmentation method | |
| Chen et al. | A new watershed algorithm for cellular image segmentation based on mathematical morphology | |
| CN117011704A (en) | Feature extraction method based on dotted line feature fusion and self-adaptive threshold | |
| CN112330613B (en) | Evaluation method and system for cytopathology digital image quality | |
| WO2021114679A1 (en) | System for calculating color conversion time of electronic paper and application thereof | |
| CN109636822A (en) | A kind of improvement Canny adaptive edge extracting method based on new building subordinating degree function | |
| CN116823686B (en) | Night infrared and visible light image fusion method based on image enhancement | |
| CN109559323A (en) | A method of picture edge characteristic is enhanced based on improved prewitt operator | |
| CN106023097A (en) | Iterative-method-based flow field image preprocessing algorithm | |
| Taha et al. | Partial Differential Equations and Digital Image Processing: A Review | |
| CN112070686A (en) | Backlight image cooperative enhancement method based on deep learning | |
| CN104361562B (en) | A kind of Variational PDE image protects side structure-preserving restored method | |
| CN110264434A (en) | A kind of removing rain based on single image method based on low-rank matrix completion | |
| CN116385293A (en) | Foggy-day self-adaptive target detection method based on convolutional neural network | |
| CN105957067A (en) | Color image edge detection method based on color difference |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20201230 Address after: Room 102, building 2, No.15, Zhushan Road, Gaochun Economic Development Zone, Nanjing, Jiangsu, 211300 Patentee after: Nanjing Tongchang Network Technology Co.,Ltd. Address before: 210044 No. 219, Ning six road, Nanjing, Jiangsu Patentee before: Nanjing University of Information Science and Technology |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230406 Address after: Building 21, No. 15 Fengji Avenue, Yuhuatai District, Nanjing City, Jiangsu Province, 211300 Patentee after: Nanjing leyoubai Network Technology Co.,Ltd. Address before: Room 102, building 2, No.15, Zhushan Road, Gaochun Economic Development Zone, Nanjing, Jiangsu, 211300 Patentee before: Nanjing Tongchang Network Technology Co.,Ltd. |