CN104754183B - A kind of real-time monitor video adaptive filter method and its system - Google Patents
A kind of real-time monitor video adaptive filter method and its system Download PDFInfo
- Publication number
- CN104754183B CN104754183B CN201510166972.2A CN201510166972A CN104754183B CN 104754183 B CN104754183 B CN 104754183B CN 201510166972 A CN201510166972 A CN 201510166972A CN 104754183 B CN104754183 B CN 104754183B
- Authority
- CN
- China
- Prior art keywords
- image
- denoising
- block
- noise
- video
- 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
Landscapes
- Image Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention provides a kind of real-time monitor video adaptive filter method and its system.Divided-fit surface is carried out to current video image f (x, y), the three-dimensional data group of each similar block is constructed;Wavelet transformation is carried out to the three-dimensional data group;Based on noise variance iteration, make self-adaptive solution processing respectively to low-and high-frequency coefficient of soft and hard threshold method and obtain the image f after denoising1(x, y).Vision signal is recovered and handled, the influence of the complex jammings such as gaussian sum impulsive noise is overcome;The marginal information of video image can be kept, denoising effect is relatively good.
Description
Technical field
The present invention relates to a kind of real-time monitor video adaptive filter method and its system, more particularly to one kind is applied to
Real-time monitor video adaptive filter method and its system that field of video monitoring is introduced due to noise.
Background technology
Video monitoring is widely used in the fields such as traffic, criminal investigation, bank, Property Management of residence, industrial production monitoring now.
The processing system for video is handled each frame of video flowing using image processing techniques, overcomes noise in image acquisition process
Caused image degradation and fuzzy.
The image appearance that monitoring system is collected is susceptible to various factors, and what is particularly introduced in gatherer process makes an uproar
Sound so that the quality of image is seriously degenerated, or even objective fuzzy, it is difficult to recognized.Such case has had a strong impact on monitoring system
Normal work, it is impossible to play a role well.As can be seen here, the research that overcome noise influences on monitor video has great show
Sincere justice.
The content of the invention
The technical problem to be solved in the present invention is to provide one kind the complex jammings such as gaussian sum impulsive noise can be overcome to influence
Real-time monitor video adaptive filter method and its system.
Video image always causes some of image to degrade in transmission and transfer process so that original image is mixed with various
Noise, is mainly shown as Gaussian noise and impulsive noise, it is therefore necessary to carry out improvement processing to the image that these degrade.Scheme at present
As noise filtering generally comprises mean filter, medium filtering, the conventional method such as Weighted median filtering, but the application of these methods
Each isolated state is lain substantially in, (mean filter is adapted to filter out Gaussian noise, and medium filtering is fitted for the consideration of shortage globality
Conjunction filters out impulsive noise, and weighted median is then adapted to the impulsive noise for filtering out detail pictures).Because these methods are all with stronger
Specific aim, so effect is not good in terms of image keeps detailing filtering.
The technical solution adopted by the present invention is as follows:A kind of real-time monitor video adaptive filter method, specific method is:
First, divided-fit surface is carried out to current video image f (x, y), constructs the three-dimensional data group of each similar block;2nd, to three dimension
Wavelet transformation is carried out according to group;3rd, based on noise variance iteration, low-and high-frequency coefficient is made adaptively respectively of soft and hard threshold method
Denoising obtains the image f after denoising1(x, y).
Preferably, the specific method of the step one is:Current video two field picture f (x, y) is read in, if its signal to noise ratio is
SNR;M × N image f (x, y) is divided into the non overlapping blocks that size is fixed as t=p × q, wherein p and q represents image respectively
The row and column size of block, calculates the distance of image blockWherein fm,nFor with reference to mark
Remember block, fs,jFor sub-block, as d (fm,n,fs,j)<During t, it is believed that then the sub-block folds these obtained sub-blocks with referring to Block- matching
Into three-dimensional data group.
Preferably, the specific method of the step 2 is:ByObtain
The coefficient that m layers of 2-d wavelet is decomposed in each group, so as to obtain the 3D wavelet transform results of correspondence group;In formula, Cj+1,k,mRepresent one
Width image, Cj,k,mThe low frequency subgraph after decomposing is represented, Respectively represent decompose after correspond to level,
Vertically, the high frequency subgraph in three directions of diagonal.
Preferably, the specific method step of the step 3 is:
(1) initial threshold is setWherein, β is adjustable coefficient, and σ is graphics standard side
Difference, n=t is image block size size;
(2) utilizeWithRespectively to LL and HH
Frequency band uses hard threshold method denoising, and Soft thresholding denoising is used to LH, HL frequency band;
(3) inverse transformation, reconstructed image are carried out to each piece of wavelet coefficient, and estimates the standard deviation sigma of the imagen;
(4) calculateIf Δ≤K, step (5) is performed;If Δ>K, then adjusting parameter β=1+ α
Δ, wherein α are step-length, repeat step (2);
(5) T now is optimal threshold, with obtaining each frequency band wavelet coefficient after the threshold denoising;
(6) the DC coefficients r (0) that LL frequency bands are extracted in one-dimensional inverse transformation is carried out to image block group, performs and sharpens computing:
(7) each sub-block wavelet coefficient after the above-mentioned optimal threshold denoising of two-dimentional inverse transformation, reconstructed image is obtained after denoising
Image f1(x, y).
Preferably, methods described also includes:To next video frame images f2When (x, y) is handled, define first residual
Poor R=f (x, y)-f1(x, y), then calculates the signal to noise ratio difference D of this two field pictures, if D<D0, then only need to f2(x, y)
Subtract residual error R and can obtain the image after denoising;The D0=0.15.
Preferably, judge current video image whether need carry out adaptive-filtering method as:Introduce image noise
The updating formula of ratio:SNR=1.04b-7, wherein b are the variance ratio of picture signal and noise signal, are defined as:B=10*
log10max(max(v(i,j)))/min(min(v(i,j))), wherein v (i, j) is the variance of image, is defined as:V (i, j)=(F (i, j)-F
(i,j)*h(q))2* h (q), F (i, j) are present image, and h (q) is matrix template, setting Gaussian noise and impulsive noise image
Snr threshold S0=25, work as SNR<S0When, judge that current video image needs to carry out adaptive-filtering.
Avaptive filtering system based on above-mentioned real-time monitor video adaptive filter method, it is characterised in that including
Similar block three-dimensional data set constructor module, divided-fit surface is carried out to current video image f (x, y), is constructed each similar
The three-dimensional data group of block;
Wavelet transformation module, wavelet transformation is carried out to the three-dimensional data group;
Denoising module, based on noise variance iteration, makes and adaptively goes respectively of soft and hard threshold method to low-and high-frequency coefficient
Processing of making an uproar obtains the image f after denoising1(x, y).
Preferably, also including next frame video image denoising method chooses judge module, for next frame video image,
Decide whether to use the signal to noise ratio of previous video frame images by the poor signal to noise for calculating it and previous video frame images for foundation
Carry out denoising.
Preferably, also including current video image adaptive-filtering judge module:Judge whether current video image needs
Carry out adaptive-filtering.
Compared with prior art, the beneficial effects of the invention are as follows:Vision signal is recovered and handled, Gauss is overcome
With the influence of the complex jamming such as impulsive noise;The marginal information of video image can be kept, denoising effect is relatively good;Define first residual
Poor R=f (x, y)-f1(x, y), then calculates the signal to noise ratio difference D of this two field pictures, if D<D0(D0Taken here for threshold values
0.15) then only need to f2(x, y) subtracts residual error R and can obtain the image after denoising, is so considerably reduced video denoising
Amount of calculation.
Brief description of the drawings
Fig. 1 is the adaptive-filtering handling process schematic diagram of a wherein embodiment of the invention.
Embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, 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, not
For limiting the present invention.
Any feature disclosed in this specification (including any accessory claim, summary and accompanying drawing), except non-specifically is chatted
State, can alternative features equivalent by other or with similar purpose replaced.I.e., unless specifically stated otherwise, each feature
A simply example in a series of equivalent or similar characteristics.
As shown in figure 1, a kind of real-time monitor video adaptive filter method, specific method is:First, to current video image
F (x, y) carries out divided-fit surface, constructs the three-dimensional data group of each similar block;2nd, wavelet transformation is carried out to the three-dimensional data group;
3rd, based on noise variance iteration, make self-adaptive solution processing respectively to low-and high-frequency coefficient of soft and hard threshold method and obtain denoising
Image f afterwards1(x, y).
Vision signal is recovered and handled, the influence of the complex jammings such as gaussian sum impulsive noise is overcome;It can keep
The marginal information of video image, denoising effect is relatively good.
The specific method of the step one is:Current video two field picture f (x, y) is read in, if its signal to noise ratio is SNR;By M ×
N image f (x, y) is divided into the non overlapping blocks that size is fixed as t=p × q, and wherein p and q represents the row and column of image block respectively
Size, calculates the distance of image blockWherein fm,nFor reference marker block, fs,jFor
Sub-block, as d (fm,n,fs,j)<During t (t is threshold values), it is believed that then the sub-block folds these obtained sub-blocks with referring to Block- matching
Into three-dimensional data group.
The specific method of the step 2 is:ByObtain
The coefficient that m layers of 2-d wavelet is decomposed in each group, so as to obtain the 3D wavelet transform results of correspondence group;In formula, Cj+1,k,mRepresent one
Width image, Cj,k,mThe low frequency subgraph after decomposing is represented, Represent to correspond to water after decomposing respectively
Flat, vertical, the high frequency subgraph in three directions of diagonal.
The specific method step of the step 3 is:
(1) initial threshold is setWherein, β is adjustable coefficient, and σ is graphics standard side
Difference, n=t is image block size size;
(2) utilizeWithRespectively to LL and HH
Frequency band uses hard threshold method denoising, and Soft thresholding denoising is used to LH, HL frequency band;
(3) inverse transformation, reconstructed image are carried out to each piece of wavelet coefficient, and estimates the standard deviation sigma of the imagen;
(4) calculateIf Δ≤K, step (5) is performed;If Δ>K, then adjusting parameter β=1+ α
Δ, wherein α are step-length, repeat step (2);
(5) T now is optimal threshold, with obtaining each frequency band wavelet coefficient after the threshold denoising;
(6) the DC coefficients r (0) that LL frequency bands are extracted in one-dimensional inverse transformation is carried out to image block group, performs and sharpens computing:
(7) each sub-block wavelet coefficient after the above-mentioned optimal threshold denoising of two-dimentional inverse transformation, reconstructed image is obtained after denoising
Image f1(x, y).
Methods described also includes:To next video frame images f2When (x, y) is handled, define first residual error R=f (x,
y)-f1(x, y), next two field picture f to entering adaptive-filtering module2(x, y), its signal to noise ratio is SNR1;Then this is calculated
The signal to noise ratio difference D of two field pictures, signal to noise ratio difference D=| SNR-SNR1|, if D<D0, then only need to subtract f2 (x, y)
Residual error R is that can obtain the image after denoising, and the image after denoising is f'2(x, y)=f2(x, y)-R, otherwise repeating said steps one
To three;So it is considerably reduced the amount of calculation of video denoising;In this specific embodiment, D0Take 0.15.
Judge current video image whether need carry out adaptive-filtering method as:The correction for introducing signal noise ratio (snr) of image is public
Formula:SNR=1.04b-7, wherein b are the variance ratio of picture signal and noise signal, are defined as:B=10*log10max(max(v(i ,j)))/min(min(v(i,j))), wherein v (i, j) is the variance of image, is defined as:V (i, j)=(F (i, j)-F (i, j) * h (q))2*h
(q), F (i, j) is present image, and h (q) is the snr threshold S of matrix template, setting Gaussian noise and impulsive noise image0
=25, work as SNR<S0When, judge that current video image needs to carry out adaptive-filtering.
Claims (6)
1. a kind of real-time monitor video adaptive filter method, specific method is:First, current video image f (x, y) is divided
Block- matching, constructs the three-dimensional data group of each similar block;2nd, wavelet transformation is carried out to the three-dimensional data group;3rd, with noise variance
Based on iteration, make self-adaptive solution processing respectively to low-and high-frequency coefficient of soft and hard threshold method and obtain the image f after denoising1(x,
y);
The specific method of the step one is:Current video two field picture f (x, y) is read in, if its signal to noise ratio is SNR;By M × N's
Image f (x, y) is divided into the non overlapping blocks that size is fixed as t=p × q, and wherein p and q represent the row and column chi of image block respectively
Very little size, calculates the distance of image blockWherein fm,nFor reference marker block, fs,jFor son
Block, as d (fm,n,fs,j)<During t, it is believed that then these obtained sub-blocks are built up three-dimensional data by the sub-block with referring to Block- matching
Group;
The specific method of the step 2 is:ByObtain in each group
The coefficient that m layers of 2-d wavelet is decomposed, so as to obtain the 3D wavelet transform results of correspondence group;In formula, Cj+1,k,mRepresent piece image,
Cj,k,mThe low frequency subgraph after decomposing is represented,Represent to correspond to after decomposing respectively level, it is vertical,
The high frequency subgraph in three directions of diagonal;
The specific method step of the step 3 is:
(1) initial threshold is setWherein, β is adjustable coefficient, and σ is graphics standard variance, n=
T is image block size size;
(2) utilizeWithRespectively to LL and HH frequency bands
Using hard threshold method denoising, Soft thresholding denoising is used to LH, HL frequency band;
(3) inverse transformation, reconstructed image are carried out to each piece of wavelet coefficient, and estimates the standard deviation sigma of the imagen;
(4) calculateIf Δ≤K, step (5) is performed;If Δ>K, then adjusting parameter β=1+ α Δs, its
Middle α is step-length, repeat step (2);
(5) T now is optimal threshold, with obtaining each frequency band wavelet coefficient after the threshold denoising;
(6) the DC coefficients r (0) that LL frequency bands are extracted in one-dimensional inverse transformation is carried out to image block group, performs and sharpens computing:
(7) each sub-block wavelet coefficient after the above-mentioned optimal threshold denoising of two-dimentional inverse transformation, reconstructed image obtains the image after denoising
f1(x, y).
2. real-time monitor video adaptive filter method according to claim 1, methods described also includes:To next video
Two field picture f2When (x, y) is handled, residual error R=f (x, y)-f is defined first1(x, y), then calculates the noise of this two field pictures
Than difference D, if D<D0, then only need to f2(x, y) subtracts residual error R and can obtain the image after denoising;The D0=0.15.
3. the real-time monitor video adaptive filter method according to one of claim 1 to 2, judges that current video image is
It is no need carry out adaptive-filtering method be:Introduce the updating formula of signal noise ratio (snr) of image:SNR=1.04b-7, wherein b are figure
As signal and the variance ratio of noise signal, it is defined as:B=10*log10max(max(v(i,j)))/ min (min (v (i, j))), wherein
V (i, j) is the variance of image, is defined as:V (i, j)=(F (i, j)-F (i, j) * h (q))2* h (q), F (i, j) are current figure
Picture, h (q) is the snr threshold S of matrix template, setting Gaussian noise and impulsive noise image0=25, work as SNR<S0When, judge
Current video image needs to carry out adaptive-filtering.
4. the Avaptive filtering system based on real-time monitor video adaptive filter method described in claim 1, it is characterised in that
Including
Similar block three-dimensional data set constructor module, divided-fit surface is carried out to current video image f (x, y), constructs each similar block
Three-dimensional data group;
Wavelet transformation module, wavelet transformation is carried out to the three-dimensional data group;
Denoising module, based on noise variance iteration, makes at self-adaptive solution respectively of soft and hard threshold method to low-and high-frequency coefficient
Reason obtains the image f after denoising1(x, y).
5. real-time monitor video Avaptive filtering system according to claim 4, in addition to next frame video image denoising
Method chooses judge module, for next frame video image, is determined by the poor signal to noise for calculating it and previous video frame images
Whether the signal to noise ratio of previous video frame images is used for according to progress denoising.
6. real-time monitor video Avaptive filtering system according to claim 4, in addition to current video image are adaptive
Filter judge module:Judge whether current video image needs to carry out adaptive-filtering.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510166972.2A CN104754183B (en) | 2015-04-10 | 2015-04-10 | A kind of real-time monitor video adaptive filter method and its system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510166972.2A CN104754183B (en) | 2015-04-10 | 2015-04-10 | A kind of real-time monitor video adaptive filter method and its system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104754183A CN104754183A (en) | 2015-07-01 |
CN104754183B true CN104754183B (en) | 2017-09-29 |
Family
ID=53593242
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510166972.2A Active CN104754183B (en) | 2015-04-10 | 2015-04-10 | A kind of real-time monitor video adaptive filter method and its system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104754183B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107622488A (en) * | 2017-09-27 | 2018-01-23 | 上海交通大学 | A kind of Confocal Images block similarity measurement method and system |
CN108198567A (en) * | 2018-02-22 | 2018-06-22 | 成都启英泰伦科技有限公司 | A kind of novel voice is except system of making an uproar |
CN108769586A (en) * | 2018-05-30 | 2018-11-06 | 深圳汇通智能化科技有限公司 | A kind of accurate monitoring system of monitoring |
CN109961408B (en) * | 2019-02-26 | 2023-03-14 | 山东理工大学 | Photon counting image denoising method based on NSCT and block matching filtering |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2011058327A2 (en) * | 2009-11-12 | 2011-05-19 | Onzo Limited | Method and apparatus for noise reduction and data compression |
CN102663687A (en) * | 2012-03-22 | 2012-09-12 | 西安电子科技大学 | Method of de-noising surface wave transformation video based on space adaptive Gaussian mixture model (GMM) |
CN102682429A (en) * | 2012-04-13 | 2012-09-19 | 泰山学院 | De-noising method of filtering images in size adaptive block matching transform domains |
CN103700072A (en) * | 2013-12-17 | 2014-04-02 | 北京工业大学 | Image denoising method based on self-adaptive wavelet threshold and two-sided filter |
-
2015
- 2015-04-10 CN CN201510166972.2A patent/CN104754183B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2011058327A2 (en) * | 2009-11-12 | 2011-05-19 | Onzo Limited | Method and apparatus for noise reduction and data compression |
CN102663687A (en) * | 2012-03-22 | 2012-09-12 | 西安电子科技大学 | Method of de-noising surface wave transformation video based on space adaptive Gaussian mixture model (GMM) |
CN102682429A (en) * | 2012-04-13 | 2012-09-19 | 泰山学院 | De-noising method of filtering images in size adaptive block matching transform domains |
CN103700072A (en) * | 2013-12-17 | 2014-04-02 | 北京工业大学 | Image denoising method based on self-adaptive wavelet threshold and two-sided filter |
Non-Patent Citations (3)
Title |
---|
一种基于小波变换的图像自适应去噪新方法;刘江涛;《硅谷》;20141231;全文 * |
基于小波变换与阀值收缩法的图像增强去噪;高飞 等;《四川理工学院学报(自然科学版)》;20060430;第19卷(第2期);全文 * |
自适应模糊阈值的图像去噪算法;王欣;《计算机仿真》;20110630;第28卷(第6期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN104754183A (en) | 2015-07-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105913393B (en) | A kind of adaptive wavelet threshold image de-noising method and device | |
CN104754183B (en) | A kind of real-time monitor video adaptive filter method and its system | |
CN101944230B (en) | Multi-scale-based natural image non-local mean noise reduction method | |
CN101950414B (en) | Non-local mean de-noising method for natural image | |
CN103116875B (en) | Self-adaptation bilateral filtering image de-noising method | |
CN103093433B (en) | Natural image denoising method based on regionalism and dictionary learning | |
CN101847257B (en) | Image denoising method based on non-local means and multi-level directional images | |
CN107464226B (en) | A kind of image de-noising method based on improvement two-dimensional empirical mode decomposition algorithm | |
CN103020918B (en) | Shape-adaptive neighborhood mean value based non-local mean value denoising method | |
CN104182939B (en) | Medical image detail enhancement method | |
CN108596848B (en) | Image denoising method based on improved wavelet threshold function | |
CN101094312A (en) | Self-adapting method for filtering image with edge being retained | |
Chen et al. | Coupling denoising algorithm based on discrete wavelet transform and modified median filter for medical image | |
CN101477679A (en) | Image denoising process based on Contourlet transforming | |
CN111242854B (en) | Image denoising method | |
CN104881847A (en) | Match video image enhancement method based on wavelet analysis and pseudo-color processing | |
CN103761707A (en) | Average filtering method eliminating image impulse noise fast and efficiently | |
CN101504769B (en) | Self-adaptive noise intensity estimation method based on encoder frame work | |
CN107169932A (en) | A kind of image recovery method based on Gauss Poisson mixed noise model suitable for neutron imaging system diagram picture | |
CN110706186B (en) | Underwater information anti-interference method based on compressed sensing and Renyi entropy | |
CN102314675B (en) | Wavelet high-frequency-based Bayesian denoising method | |
CN111311508B (en) | Noise reduction method for pavement crack image with noise | |
CN112801897A (en) | Image denoising method based on wide convolution neural network | |
CN105701781A (en) | Image denoising method based on multi-resolution singular value decomposition | |
CN102547074A (en) | Surfacelet domain BKF model Bayes video denoising method |
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 |