CN107295217A - A kind of video noise estimation method based on principal component analysis - Google Patents
A kind of video noise estimation method based on principal component analysis Download PDFInfo
- Publication number
- CN107295217A CN107295217A CN201710524401.0A CN201710524401A CN107295217A CN 107295217 A CN107295217 A CN 107295217A CN 201710524401 A CN201710524401 A CN 201710524401A CN 107295217 A CN107295217 A CN 107295217A
- Authority
- CN
- China
- Prior art keywords
- video
- block
- matrix
- gradient
- noise
- 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.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N5/00—Details of television systems
- H04N5/14—Picture signal circuitry for video frequency region
- H04N5/21—Circuitry for suppressing or minimising disturbance, e.g. moiré or halo
Abstract
The present invention proposes a kind of video noise estimation method based on principal component analysis, make full use of the correlation of video sequence, the search of similar block is carried out between adjacent two field picture, difference image between adjacent two field picture is obtained based on minimum cost criterion, the influence of video motion is eliminated, preliminary weak texture difference image is obtained；Meanwhile, the noise estimation based on video block, adaptive acquisition noise level parameter are introduced again, and propose that normal distyribution function, as the threshold function table that weak texture block is selected in text, reduces computation complexity；Clear and definite iteration index is also set up in addition and make it that the noise level of estimation is more accurate, it is to avoid underestimate phenomenon under high noise levels.Set forth herein the estimation of video noise estimation algorithm it is accurate, blind video denoising field can be applied to, had broad application prospects.
Description
Technical field
The present invention relates to a kind of video noise estimation method, specifically, a kind of regarding based on principal component analysis relate to
Frequency noise estimation method.
Background technology
Vision signal may all introduce noise during catching, recording and transmitting.Seriously reduction is regarded the noise of introducing
Frequency image quality, influences the visual experience of spectators.And video denoising is by the characteristics of digital video image and at existing signal
Reason technology is combined, and a kind of multimedia information of noise jamming in video image is reduced as far as possible.It presently, there are
The performances of most of video denoising algorithms all rely to varying degrees on noise parameter in noisy video.Although by various
The algorithm of various kinds, can reach highly desirable denoising effect, but most of algorithms have individual supposed premise, i.e. noise intensity to be pre-
First know.Artificially given noise parameter or noise parameter are inaccurate, can all cause denoising effect undesirable.Therefore for containing
Make an uproar video noise parameter estimation be video denoising research in a critical problem.
In order to solve the problem of above is present, people are seeking a kind of preferable technical solution always.
The content of the invention
The purpose of the present invention is in view of the shortcomings of the prior art, so as to be made an uproar there is provided a kind of video based on principal component analysis
Sound method of estimation.
To achieve these goals, the technical solution adopted in the present invention is：A kind of video based on principal component analysis is made an uproar
Sound method of estimation, comprises the following steps：
Step 1, noisy frame of video is inputted, Block- matching is carried out to current frame image and next two field picture, and calculating is worked as respectively
The difference image block of prior image frame sub-block corresponding with each pair in next two field picture, chooses a pair optimal of corresponding sub-blocks of matching effect
Difference image block be used as original noisy image；
Step 2, the variance of original noisy image is calculated, as initial noisc level, and the initial noisc leveler is utilized
Calculate initial threshold；
Step 3, all sub-blocks of current frame image are write the form of column vector as, calculate each sub-block gradient matrix,
The eigenvalue of maximum of the corresponding gradient covariance matrix of gradient matrix and gradient covariance matrix；
Step 4, the eigenvalue of maximum for choosing gradient covariance matrix is less than the sub-block of initial threshold, labeled as weak texture maps
As block；
Step 5, all weak texture image blocks are constituted into a big data matrix, and calculates the corresponding association side of data matrix
Poor matrix；
Step 6, feature decomposition is done to covariance matrix, obtains its all characteristic value, using minimum characteristic value as making an uproar
The estimate of sound level, and calculate threshold value using the estimate of the noise level；
Step 7, calculate the corresponding gradient matrix of each weak texture image block, the corresponding covariance matrix of gradient matrix and
The eigenvalue of maximum of covariance matrix；
Step 8, the eigenvalue of maximum for choosing gradient covariance matrix is less than the weak texture image block of threshold value, repeat step 5
To step 7, until stopping when reaching default iteration index.
Based on above-mentioned, normal distyribution function is respectively adopted in step 2 and step 6 and calculates initial threshold and threshold value.
Based on above-mentioned, in step 8, default iteration index is that iterations reaches preset times or front and rear estimated twice
The ratio of noise level difference and previous estimate is less than preset ratio.
The present invention is compared with the prior art with prominent substantive distinguishing features and significantly progressive, specifically, of the invention logical
Cross front and rear two frames noisy image and calculate first estimation noise without artificial given noise parameter, reduce image information to making an uproar
The influence of sound estimation, can widely adapt to various types of pictures, and the error of noise estimation is smaller, and estimated result is more accurate.
Brief description of the drawings
Fig. 1 is the schematic flow sheet of the present invention.
Embodiment
Below by embodiment, technical scheme is described in further detail.
As shown in figure 1, a kind of video noise estimation method based on principal component analysis, comprises the following steps：
Step 1, noisy frame of video is inputted, Block- matching is carried out to current frame image and next two field picture, and calculating is worked as respectively
The difference image block of prior image frame sub-block corresponding with each pair in next two field picture, chooses a pair optimal of corresponding sub-blocks of matching effect
Difference image block be used as original noisy image；
Specifically, the front and rear two field pictures of sequence of video images have very strong correlation in time, gathered around for script
There are many textures and the block of detailed information, if it has stronger correlation in front and rear two frame, then the result of its difference is still
A smooth block can be so produced, so as to reduce the influence that image information is estimated noise；
Assuming that the video image collected is I (i, j, n)=S (i, j, n)+N (i, j, n)
Wherein, S (i, j, n) is original not noisy frame of video, and N (i, j, n) is noise signal, and n is frame number, and i, j are
Pixel coordinate；
By Block- matching, a pair of optimal correspondence sub-blocks of selection matching effect calculate difference image between adjacent two field picture
Block；
Have for a pair of optimal correspondence sub-blocks of matching effect：S (i, j, n)=S (i ', j ', n+1)
Difference image block between adjacent two field picture is obtained according to minimum cost：
D (i, j, n)=I (i, j, n)+I (i ', j ', n+1)=N (i, j, n)-N (i ', j ', n+1)；
Next difference image block between above-mentioned adjacent two field picture is subsequently calculated as original noisy image.
Step 2, if the actual noise variance of raw video image isOriginal noisy image can be obtained according to probability calculation
Initial noisc level have：
Next proceed to calculate in initial threshold, the present invention using the noise variance of original noisy image and use normal distribution
Function calculates threshold value, simplifies algorithm model, reduces computation complexity.
Under normal circumstances, threshold calculations formula is：
Wherein, F-1(δ, α, β) is the inverse function for the cumulative distribution function that Gamma is distributed,For form parameter, β is
Scale parameter, N2It is the pixel count in image block；δ is artificially given significance, σnFor the standard deviation of Gaussian noise；Dh
For the derivation operator of horizontal direction, DvFor the derivation operator of vertical direction；Tr represents to ask mark to operate, the operation of T representing matrixs transposition；
Because the size of derivation operator is only relevant with the size of image block, do not influenceed by original noisy image, therefore when 3
During rank filter operator, the derivation operator D of horizontal directionhWith the derivation operator D of vertical directionvFor normal matrix,
For normal matrix, and then try to achieve
By Gamma probability density figure it is known that Gamma is distributed approximate normal distribution when ɑ is sufficiently large.Meanwhile, according to
Central-limit theorem, independent identically distributed stochastic variable sum tends to normal distribution.Being derived by central-limit theorem to obtain：
Wherein, μ=α β, σ2=α β2=μ β.
Therefore trying to achieve the average and variance of normal distribution is：
Simplify formula so as to obtain threshold calculations：
By initial noisc levelInitial threshold τ can be obtained after substitution0For
Step 3, all sub-blocks of current frame image are write the form of column vector as, is designated as yi,yiIt is i-th sub-block, meter
Calculate the gradient matrix of each sub-blockGradient matrixCorresponding gradient covariance matrixAnd gradient covariance matrixEigenvalue of maximum；
Wherein,
Step 4, gradient covariance matrix is chosenEigenvalue of maximum be less than initial threshold τ0Sub-block, labeled as weak line
Manage image block；
Step 5, all weak texture image blocks are constituted into data matrix M one big, and it is corresponding to calculate data matrix M
Covariance matrix ∑ m；
Step 6, feature decomposition is done to covariance matrix ∑ m, obtains its all characteristic value, using minimum characteristic value as
The estimate of noise levelIt is designated asAnd by the estimate of the noise levelSubstitute into threshold calculations letter
Change formula to calculate threshold tau
Step 7, the corresponding gradient matrix of each weak texture image block is calculatedGradient matrixCorresponding covariance square
Battle arrayAnd covariance matrixEigenvalue of maximum；
Step 8, gradient covariance matrix is chosenEigenvalue of maximum be less than threshold tau weak texture image block, repeat walk
Rapid 5 to step 7, until reaching that iterations reaches that preset times or the front and rear noise level difference estimated twice are estimated with previous
The ratio of evaluation stops iteration when being less than preset ratio.
When it is implemented, when iterations reaches 6 times or the front and rear noise level difference estimated twice and previous estimation
The ratio of value stops iteration when being less than 2%.
Finally it should be noted that:The above embodiments are merely illustrative of the technical scheme of the present invention and are not intended to be limiting thereof；To the greatest extent
The present invention is described in detail with reference to preferred embodiments for pipe, those of ordinary skills in the art should understand that：Still
The embodiment of the present invention can be modified or equivalent substitution is carried out to some technical characteristics；Without departing from this hair
The spirit of bright technical scheme, it all should cover among claimed technical scheme scope of the invention.
Claims (3)
1. a kind of video noise estimation method based on principal component analysis, it is characterised in that comprise the following steps：
Step 1, noisy frame of video is inputted, Block- matching is carried out to current frame image and next two field picture, and calculate present frame respectively
The difference image block of image sub-block corresponding with each pair in next two field picture, chooses the difference of a pair of optimal correspondence sub-blocks of matching effect
Partial image block is used as original noisy image；
Step 2, the variance of original noisy image is calculated, as initial noisc level, and using at the beginning of the initial noisc level calculation
Beginning threshold value；
Step 3, all sub-blocks of current frame image are write the form of column vector as, calculates gradient matrix, the gradient of each sub-block
The eigenvalue of maximum of the corresponding gradient covariance matrix of matrix and gradient covariance matrix；
Step 4, the eigenvalue of maximum for choosing gradient covariance matrix is less than the sub-block of initial threshold, labeled as weak texture image
Block；
Step 5, all weak texture image blocks are constituted into a big data matrix, and calculates the corresponding covariance square of data matrix
Battle array；
Step 6, feature decomposition is done to covariance matrix, obtains its all characteristic value, regard minimum characteristic value as noise water
Flat estimate, and calculate threshold value using the estimate of the noise level；
Step 7, the corresponding gradient matrix of each weak texture image block, the corresponding covariance matrix of gradient matrix and association side are calculated
The eigenvalue of maximum of poor matrix；
Step 8, the eigenvalue of maximum for choosing gradient covariance matrix is less than the weak texture image block of threshold value, repeat step 5 to step
Rapid 7, until stopping when reaching default iteration index.
2. the video noise estimation method according to claim 1 based on principal component analysis, it is characterised in that：Step 2 and
Normal distyribution function is respectively adopted in step 6 and calculates initial threshold and threshold value.
3. the video noise estimation method according to claim 1 or 2 based on principal component analysis, it is characterised in that：Step 8
In, default iteration index is that iterations reaches preset times or the front and rear noise level difference estimated twice and previous estimation
The ratio of value is less than preset ratio.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710524401.0A CN107295217B (en) | 2017-06-30 | 2017-06-30 | Video noise estimation method based on principal component analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710524401.0A CN107295217B (en) | 2017-06-30 | 2017-06-30 | Video noise estimation method based on principal component analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107295217A true CN107295217A (en) | 2017-10-24 |
| CN107295217B CN107295217B (en) | 2020-06-12 |
Family
ID=60098638
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710524401.0A Active CN107295217B (en) | 2017-06-30 | 2017-06-30 | Video noise estimation method based on principal component analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107295217B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109171670A (en) * | 2018-06-25 | 2019-01-11 | 天津海仁医疗技术有限公司 | A kind of 3D blood vessel imaging algorithm based on reverse Principal Component Analysis |
| CN110503625A (en) * | 2019-07-02 | 2019-11-26 | 杭州电子科技大学 | A kind of cmos image signal dependent noise method for parameter estimation |
| CN113298770A (en) * | 2021-05-20 | 2021-08-24 | 武汉工程大学 | Image noise level estimation method, device and computer storage medium |
| CN113469893A (en) * | 2020-05-08 | 2021-10-01 | 上海齐感电子信息科技有限公司 | Method for estimating noise of image in video and video noise reduction method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101919230A (en) * | 2007-12-25 | 2010-12-15 | 梅迪奇视觉-脑科技有限公司 | The reduction of picture noise |
| CN102609914A (en) * | 2012-01-17 | 2012-07-25 | 天津大学 | Signal-correlated noise estimating method for image sensor |
| 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 |
| US20150235350A1 (en) * | 2014-02-18 | 2015-08-20 | Signal Processing, Inc. | Method for Image Denoising |
| CN105678774A (en) * | 2016-01-11 | 2016-06-15 | 浙江传媒学院 | Image noise level estimation method based on principal component analysis |
-
2017
- 2017-06-30 CN CN201710524401.0A patent/CN107295217B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101919230A (en) * | 2007-12-25 | 2010-12-15 | 梅迪奇视觉-脑科技有限公司 | The reduction of picture noise |
| CN102609914A (en) * | 2012-01-17 | 2012-07-25 | 天津大学 | Signal-correlated noise estimating method for image sensor |
| 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 |
| US20150235350A1 (en) * | 2014-02-18 | 2015-08-20 | Signal Processing, Inc. | Method for Image Denoising |
| CN105678774A (en) * | 2016-01-11 | 2016-06-15 | 浙江传媒学院 | Image noise level estimation method based on principal component analysis |
Non-Patent Citations (1)
| Title |
|---|
| 张娟: "噪声大小估计耦合PCA的图像降噪算法", 《计算机工程与设计》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109171670A (en) * | 2018-06-25 | 2019-01-11 | 天津海仁医疗技术有限公司 | A kind of 3D blood vessel imaging algorithm based on reverse Principal Component Analysis |
| CN110503625A (en) * | 2019-07-02 | 2019-11-26 | 杭州电子科技大学 | A kind of cmos image signal dependent noise method for parameter estimation |
| CN110503625B (en) * | 2019-07-02 | 2021-08-17 | 杭州电子科技大学 | CMOS image signal related noise parameter estimation method |
| CN113469893A (en) * | 2020-05-08 | 2021-10-01 | 上海齐感电子信息科技有限公司 | Method for estimating noise of image in video and video noise reduction method |
| CN113469893B (en) * | 2020-05-08 | 2024-03-22 | 上海齐感电子信息科技有限公司 | Noise estimation method of image in video and video noise reduction method |
| CN113298770A (en) * | 2021-05-20 | 2021-08-24 | 武汉工程大学 | Image noise level estimation method, device and computer storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107295217B (en) | 2020-06-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108961198B (en) | Underwater image synthesis method of multi-grid generation countermeasure network and application thereof | |
| Deng et al. | Wavelet domain style transfer for an effective perception-distortion tradeoff in single image super-resolution | |
| CN107295217A (en) | A kind of video noise estimation method based on principal component analysis | |
| CN102360498B (en) | Reconstruction method for image super-resolution | |
| CN109993095B (en) | Frame level feature aggregation method for video target detection | |
| CN105046677B (en) | A kind of enhancing treating method and apparatus for traffic video image | |
| CN103049892B (en) | Non-local image denoising method based on similar block matrix rank minimization | |
| CN104867111B (en) | A kind of blind deblurring method of non-homogeneous video based on piecemeal fuzzy core collection | |
| CN104680510A (en) | RADAR parallax image optimization method and stereo matching parallax image optimization method and system | |
| CN110288550B (en) | Single-image defogging method for generating countermeasure network based on priori knowledge guiding condition | |
| CN108234884B (en) | camera automatic focusing method based on visual saliency | |
| Lie et al. | 2D to 3D video conversion with key-frame depth propagation and trilateral filtering | |
| KR20110014067A (en) | Method and system for transformation of stereo content | |
| CN114463218B (en) | Video deblurring method based on event data driving | |
| CN109544694A (en) | A kind of augmented reality system actual situation hybrid modeling method based on deep learning | |
| CN110349112B (en) | Two-stage image denoising method based on self-adaptive singular value threshold | |
| CN105976337A (en) | Image defogging method based on filtering guiding via medians | |
| CN110458784A (en) | It is a kind of that compression noise method is gone based on image perception quality | |
| CN105023246B (en) | A kind of image enchancing method based on contrast and structural similarity | |
| Zhou et al. | Multicolor light attenuation modeling for underwater image restoration | |
| CN104200434B (en) | Non-local mean image denoising method based on noise variance estimation | |
| CN102810202B (en) | Based on the image multistep residual feedback iterative filtering method of fractional order difference weighting | |
| CN104159098B (en) | The translucent edge extracting method of time domain consistence of a kind of video | |
| CN108830146A (en) | A kind of uncompressed domain lens boundary detection method based on sliding window | |
| CN104537637A (en) | Method and device for estimating depth of single static image |
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 |