CN106780398A - A kind of image de-noising method based on noise prediction - Google Patents
A kind of image de-noising method based on noise prediction Download PDFInfo
- Publication number
- CN106780398A CN106780398A CN201710016951.1A CN201710016951A CN106780398A CN 106780398 A CN106780398 A CN 106780398A CN 201710016951 A CN201710016951 A CN 201710016951A CN 106780398 A CN106780398 A CN 106780398A
- Authority
- CN
- China
- Prior art keywords
- image
- noise
- information
- redundancy
- image sequence
- 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
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000004458 analytical method Methods 0.000 claims abstract description 6
- 230000006835 compression Effects 0.000 claims description 18
- 238000007906 compression Methods 0.000 claims description 18
- 230000009466 transformation Effects 0.000 claims description 3
- 239000000470 constituent Substances 0.000 claims description 2
- 238000012880 independent component analysis Methods 0.000 abstract description 4
- 230000009467 reduction Effects 0.000 abstract description 3
- 244000025254 Cannabis sativa Species 0.000 abstract description 2
- 239000011159 matrix material Substances 0.000 description 7
- 238000000926 separation method Methods 0.000 description 3
- 238000009499 grossing Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000011946 reduction process Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- 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
- G06T2207/20081—Training; Learning
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of image de-noising method based on noise prediction, belong to the technical field of image procossing.The present invention is first by redundancy unnecessary in KSVD algorithms removal original image, then remaining image redundancy information carries out non-average value processing after neighborhood window is to KSVD algorithm process, avoid the inaccurate of long-range prediction grass, enable that the image removed after prediction noise retains detailed information to a greater extent, observed image with the difference of image after original image and non-average value processing as independent composition analysis algorithm again, ensure that the useful information component isolated and noise component(s) are separate by independent component analysis, tried one's best while denoising and retain the useful information of image, ensure that the image after noise reduction is optimal.
Description
Technical field
The invention discloses a kind of image de-noising method based on noise prediction, belong to the technical field of image procossing.
Background technology
With the development of the informationization technology based on computer network, requirement of the people to digital picture quality is more next
Higher, the noise such as thermal noise, shot noise and quantum noise that image output equipment and some photoelectric devices are produced can seriously drop
The quality of the image of low final output.Digital picture noise can be divided into internal noise and external noise according to Producing reason,
Defined with the shape of Annual distribution according to noise amplitude and Gaussian noise can be divided into and Rayleigh noise, Gaussian noise and the spiced salt are made an uproar
Sound is common two kinds of noises in digital picture.
Traditional denoising mode puts on an equal footing noise and useful information, it is believed that noise and image detail information become at certain
Change in domain and be distributed in different intervals, however, noise and some useful information images are generally overlapped mutually in certain region,
Can also remove useful detailed information while removal noise, different denoising means can only specific noise signal range it
Interior effective, off-limits noisy image comprehends large area distortion by place.
Blind source separating is a kind for the treatment of technology based on thermal signals such as statistics, and initial classic applications are " cocktail party ".
Blind source separating because can analyze multidimensional data be gradually applied to image co-registration, image enhaucament, feature extraction, artifact eliminate, it is mixed
Close image, eliminate the image procossings such as line at random aspect, and carrying out image noise reduction using blind source separating needs by the priori of noise
Knowledge carries out sparse coding compression by way of training noisy acoustic image.
The content of the invention
Goal of the invention of the invention is directed to the deficiency of above-mentioned background technology, there is provided a kind of image based on noise prediction
Denoising method, by the progressively noise reduction process to original image, remains image detail information and solves blind source to greatest extent
Technical problem of the separation algorithm it is necessary to have multiple image signal sources.
The present invention employs following technical scheme for achieving the above object:
Patent of the present invention is that a kind of image that certain noise prediction is carried out from noisy image based on blind source separating goes
Method for de-noising, when gained image only, it is necessary to carry out a part of denoising to this image, first have to carry out is pressure
Contracting, the redundancy of image is obtained by compression, non-average value processing is carried out to redundancy and obtains the different image of noise intensity,
This step is repeated, multiple images containing varying strength noise can be obtained, blind source point is carried out further according to independent composition analysis algorithm
From obtaining denoising image, it is not necessary to shoot photo in large quantities and can to greatest extent retain detailed information.
The method is specifically:Non-homogeneous treatment is carried out to the redundancy through KSVD dictionary compressions to obtain containing varying strength
The image of Gaussian noise, raw noise image subtracts the image containing varying strength Gaussian noise and obtains a preliminary denoising figure
Picture, then blind source separating is carried out to preliminary denoising image, specific method is as follows:
1st, sparse compression raw noise image
The raw noise image for shooting and obtaining is processed using KSVD compression algorithms, that is, meets the contractible iteration condition of KSVD,
The object function of KSVD algorithms is expressed as:
Wherein, D ∈ Rn×K y∈Rn x∈RK, D, Y, X represent the rarefaction representation of dictionary, training signal, training signal respectively
Vector, D ∈ Rn×K, Y ∈ Rn, X ∈ RK, Rn×KIt is the vector space of n × K, RnIt is n-dimensional vector space, RKIt is K gts,It is the set of N number of training signal,It is the solution vector set of Y, T0It is non-zero in rarefaction representation coefficient
Maximum difference degree in the upper limit of the number of amount, i.e. coefficient vector.
After iterated transform, many noise informations in raw noise image are together rejected, and what is remained is superfluous
Remaining information but contains many image detail informations, i.e. through the treated noisy acoustic intelligence of redundancy information for obtaining of KSVD compressions and
Substantial amounts of image detail information, non-average value processing is to these excessive further rarefactions of redundancy;
2nd, non-average value processing redundancy
Search window with D pixels as window size, neighborhood window with d pixels as window size, due to KSVD compressions treatment
The redundancy for producing afterwards also retains substantial amounts of image detail information while partial information is remained, according between neighborhood
Similitude determines the weights of pixel to estimate a relatively small noise signalBetween by calculating two neighborhood windows
Similarity degree is assigned to weights for pixel yThe prediction noise of pixel xFor:
Wherein, weightsThe similarity between pixel x and pixel y is represented, its value is by with pixel x, pixel
Distance between rectangular neighborhood V (x) centered on point y, V (y) | | V (x)-V (y) | |2Determine:
Wherein,
Z (x) is normalization coefficient, and h is the smoothing parameter for controlling Gaussian function attenuation degree, and d is the size of neighborhood window,
Ds is transform domain, repeats the image sequence that non-average value processing can obtain being made up of multiple images containing varying strength noise, former
Beginning noise image X' and prediction noiseMake the difference and obtain preliminary denoising image
3rd, the preliminary denoising image of blind source separating
Just become new observed image, according to blind source separating:
Subscript i is order of every frame observed image in image sequence, and the value 1 ..., m, m of i are observed image
Totalframes, aijIt is component mixed coefficint, S is represented by picture content sjThe matrix of composition, S=[s1,s2,...,sj...,sn]T, under
Mark j represents component sequence number, and the value 1 ..., n, n of j are component sum;M, n are integer, and m >=n >=2, T representing matrix transposition;
Separation matrix W is obtained by fixed point independent component analysis FastICA, so as to obtain unlimited approximation component SjPoint
Amount yi, its relational expression is expressed as below:
Y=WT=WAS → S (8)
In formula (8), Y represents the matrix that the component by isolating is constituted, Y=[y1,y2,...yn]T, yiRepresent image information
Component.
Because non-average value processing is carried out by neighborhood window, it is ensured that will not be made because of long-range prediction noise
It is inaccurate into calculating so that the image after removal prediction noise signal can to a greater extent retain detailed information, to pass through
Picture after non-average value processing carries out blind source separating with the difference of raw noise image as observed image can obtain optimal going
Make an uproar image.
For the image sequence that imaging device shoots, the present invention proposes a kind of image de-noising method based on noise prediction,
Gray proces are carried out to raw noise image sequence, the raw noise image sequence after gray proces is carried out by KSVD algorithms
Dictionary compression obtain it is sparse after redundant image sequence, similarity transformation in neighborhood is carried out to redundancy image sequence and is obtained newly
Noisy image sequence, carry out blind source point by isolated component of the noise information in redundancy image and image useful information
From the standard variance or variance of each isolated component that statistics is separated, according to isolated component standard variance or the spy of variance
Property filters out image useful information isolated component.
Present invention employs above-mentioned technical proposal, have the advantages that:Need not in removing image by KSVD first
The redundancy wanted, then remaining image redundancy information carries out non-average value processing after neighborhood window is to KSVD treatment, keeps away
The inaccurate of long-range prediction grass is exempted from so that removing the image after prediction noise can to a greater extent retain details
Information, then the observed image with the difference of image after original image and non-average value processing as independent composition analysis algorithm, by independence
Constituent analysis ensures that the useful information component isolated and noise component(s) are separate, is tried one's best while denoising and retains image
Useful information, it is ensured that the image after noise reduction is optimal.
Brief description of the drawings
Fig. 1 is the flow chart of image de-noising method of the present invention.
Specific embodiment
The technical scheme invented is described in detail below in conjunction with the accompanying drawings.
An image sequence is shot using imaging device, captured target is in relative static conditions;Afterwards by right
The gray level image for obtaining carries out KSVD compressions, image information is retained to greatest extent and obtains having comprising noise information and image information
With the redundancy image sequence of information, similarity transformation's treatment in neighborhood is carried out to redundancy image sequence and obtains new containing
Make an uproar image sequence;Independent component analysis method using higher order statistical in blind source separating is divided new noisy image sequence
From, the standard variance (Standard Deviation, SD) or variance (Vanriation) of each component that calculating is separated,
Value the greater is required image useful information component, and the standard variance of multiple noise component(s)s is closer to each other, but relative to
The standard variance of image useful information is smaller.
Picture noise information and image useful information are considered as the isolated component of composition image, different noise images are combined and is obtained
The less standard variance (Standard Deviation, SD) or variance (Vanriation) for arriving are roughly the same, and larger mark
Quasi- variance or the corresponding isolated component of variance represent image information, export larger standard variance or the corresponding isolated component institute of variance
The image of composition, that is, complete noise reduction process.
When image only, it is necessary to carry out a part of denoising to this image, first have to carry out is compression,
The redundancy of image is obtained by compression, non-average value processing is carried out to redundancy and is obtained the different image of noise intensity, with
This analogizes, and can obtain multiple images containing varying strength noise, carries out blind source separating further according to independent composition analysis algorithm and obtains
To denoising image.Do not need substantial amounts of shooting photo and can to greatest extent retain detailed information.
Its method is as shown in Figure 1:The redundancy of KSVD dictionary compressions is carried out non-homogeneous treatment obtain containing it is different by force
The image of Gaussian noise is spent, raw noise image subtracts the image containing varying strength Gaussian noise and obtains a preliminary denoising figure
Picture, then blind source separating is carried out to preliminary denoising image, specific method is as follows:
1st, sparse compression raw noise image
The raw noise image for shooting and obtaining is processed using KSVD compression algorithms, that is, meets the contractible iteration condition of KSVD,
The object function of KSVD algorithms is expressed as:
Wherein, D ∈ Rn×K y∈Rn x∈RK, D, Y, X represent the rarefaction representation of dictionary, training signal, training signal respectively
Vector, D ∈ Rn×K, Y ∈ Rn, X ∈ RK, Rn×KIt is the vector space of n × K, RnIt is n-dimensional vector space, RKIt is K gts,It is the set of N number of training signal,It is the solution vector set of Y, T0It is non-zero in rarefaction representation coefficient
Maximum difference degree in the upper limit of the number of amount, i.e. coefficient vector.
After iterated transform, many noise informations are together rejected in raw noise image, the redundancy for remaining
The information but detailed information containing many images, i.e. through the treated noisy acoustic intelligence of redundancy information for obtaining of KSVD compressions and
Substantial amounts of image detail information, non-average value processing is to these excessive further rarefactions of redundancy.
2nd, non-average value processing redundancy
Search window with D pixels as window size, neighborhood window with d pixels as window size, due to KSVD compressions treatment
The redundancy for producing afterwards also retains substantial amounts of image detail information while partial noise information is remained, according to neighborhood
Between similitude determine the weights of pixel to estimate a relatively small noise signal.By calculating two neighborhood windows
Similarity degree between mouthful is assigned to weights for pixel yThe prediction noise of pixel xFor:
Wherein, weightsThe similarity between pixel x and pixel y is represented, its value is by with pixel x, pixel
Distance between rectangular neighborhood V (x) centered on point y, V (y) | | V (x)-V (y) | |2Determine:
Wherein,
Z (x) is normalization coefficient, and h is the smoothing parameter for controlling Gaussian function attenuation degree, and d is the size of neighborhood window,
Ds is transform domain, repeats the image sequence that non-average value processing can obtain being made up of multiple images containing varying strength noise, former
Beginning noise image X' and prediction noiseMake the difference and obtain preliminary denoising image
3rd, the preliminary denoising image of blind source separating
Just become new observed image, according to blind source separating:
Subscript i is order of every frame observed image in image sequence, and the value 1 ..., m, m of i are observed image
Totalframes, aijIt is component mixed coefficint, S is represented by picture content sjThe matrix of composition, S=[s1,s2,...,sj,...,sn]T,
Subscript j represents component sequence number, and the value 1 ..., n, n of j are component sum;M, n are integer, and m >=n >=2, T representing matrixs turn
Put;
Separation matrix W is obtained by fixed point independent component analysis FastICA, so as to obtain unlimited approximation component SjPoint
Amount yi, its relational expression is expressed as below:
Y=WT=WAS → S (8)
In formula (8), Y represents the matrix that the component by isolating is constituted, Y=[y1,y2,...,yi,...,yn]T, yiRepresent
Image information component.
Because non-average value processing is carried out by neighborhood window, it is ensured that will not be made because of long-range prediction noise
It is inaccurate into what is calculated so that the image after removal prediction noise signal can to a greater extent retain detailed information, to pass through
The picture crossed after non-average value processing carries out blind source separating and can obtain optimal with the difference of raw noise image as observed image
Denoising image.
Claims (6)
1. a kind of image de-noising method based on noise prediction, it is characterised in that be compressed to individual original image and obtain superfluous
Remaining information, heterogeneous treatment is carried out to redundancy and obtains multiple images containing varying strength noise information, by raw noise figure
The difference of picture and each image containing varying strength noise information constitutes observed image sequence, and blind source point is carried out to observed image sequence
From obtaining last denoising image.
2. a kind of image de-noising method based on noise prediction according to claim 1, it is characterised in that calculated by KSVD
Method raw noise image is carried out dictionary compression obtain it is sparse after redundancy, the redundancy contains noise information and figure
As detailed information.
3. a kind of image de-noising method based on noise prediction according to claim 1, it is characterised in that only using fixed point
Vertical constituent analysis FastICA algorithms carry out blind source separating to observed image sequence and obtain last denoising image.
4. a kind of image de-noising method based on noise prediction, it is characterised in that be compressed to raw noise image sequence
To redundancy image sequence, similarity transformation in neighborhood is carried out to redundancy image sequence and obtains new noisy image sequence
Row, are separated using the method for higher order statistical isolated component standard variance or variance in blind source separating from new noisy image sequence
Go out image information.
5. a kind of image de-noising method based on noise prediction according to claim 4, it is characterised in that utilize blind source separating
The method of middle higher order statistical isolated component standard variance or variance isolates image information from new noisy image sequence, specifically
Method is:Blind source separating is carried out by isolated component of the noise information in redundancy image and image useful information, quilt is counted
The standard variance or variance of each isolated component isolated, figure is filtered out according to the characteristic of isolated component standard variance or variance
As useful information isolated component.
6. a kind of image de-noising method based on noise prediction according to claim 4, it is characterised in that to raw noise figure
Obtain the method for redundancy image sequence and be as sequence is compressed:First, raw noise image sequence is carried out at gray scale
Reason, then, by KSVD algorithms the raw noise image sequence after gray proces is carried out dictionary compression obtain it is sparse after it is superfluous
Remaining image sequence.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710016951.1A CN106780398B (en) | 2017-01-10 | 2017-01-10 | A kind of image de-noising method based on noise prediction |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710016951.1A CN106780398B (en) | 2017-01-10 | 2017-01-10 | A kind of image de-noising method based on noise prediction |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106780398A true CN106780398A (en) | 2017-05-31 |
CN106780398B CN106780398B (en) | 2019-07-23 |
Family
ID=58948846
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710016951.1A Active CN106780398B (en) | 2017-01-10 | 2017-01-10 | A kind of image de-noising method based on noise prediction |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106780398B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111428596A (en) * | 2020-03-16 | 2020-07-17 | 重庆邮电大学 | Grinding sound signal detection method based on three sound pickups |
CN117928819A (en) * | 2024-03-21 | 2024-04-26 | 西安思坦仪器股份有限公司 | Underground pressure monitoring method and system of permanent wireless pressure gauge |
CN117928819B (en) * | 2024-03-21 | 2024-05-24 | 西安思坦仪器股份有限公司 | Underground pressure monitoring method and system of permanent wireless pressure gauge |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040158821A1 (en) * | 2002-12-18 | 2004-08-12 | Scott Rickard | System and method for non-square blind source separation under coherent noise by beamforming and time-frequency masking |
US8391509B2 (en) * | 2009-08-03 | 2013-03-05 | National Chiao Tung University | Audio-separating apparatus and operation method thereof |
CN103400358A (en) * | 2013-08-27 | 2013-11-20 | 中国兵器工业导航与控制技术研究所 | Wavelet information redundancy based infrared image denoising algorithm |
CN106204461A (en) * | 2015-05-04 | 2016-12-07 | 南京邮电大学 | Compound regularized image denoising method in conjunction with non local priori |
CN106251315A (en) * | 2016-08-23 | 2016-12-21 | 南京邮电大学 | A kind of image de-noising method based on full variation |
-
2017
- 2017-01-10 CN CN201710016951.1A patent/CN106780398B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040158821A1 (en) * | 2002-12-18 | 2004-08-12 | Scott Rickard | System and method for non-square blind source separation under coherent noise by beamforming and time-frequency masking |
US8391509B2 (en) * | 2009-08-03 | 2013-03-05 | National Chiao Tung University | Audio-separating apparatus and operation method thereof |
CN103400358A (en) * | 2013-08-27 | 2013-11-20 | 中国兵器工业导航与控制技术研究所 | Wavelet information redundancy based infrared image denoising algorithm |
CN106204461A (en) * | 2015-05-04 | 2016-12-07 | 南京邮电大学 | Compound regularized image denoising method in conjunction with non local priori |
CN106251315A (en) * | 2016-08-23 | 2016-12-21 | 南京邮电大学 | A kind of image de-noising method based on full variation |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111428596A (en) * | 2020-03-16 | 2020-07-17 | 重庆邮电大学 | Grinding sound signal detection method based on three sound pickups |
CN117928819A (en) * | 2024-03-21 | 2024-04-26 | 西安思坦仪器股份有限公司 | Underground pressure monitoring method and system of permanent wireless pressure gauge |
CN117928819B (en) * | 2024-03-21 | 2024-05-24 | 西安思坦仪器股份有限公司 | Underground pressure monitoring method and system of permanent wireless pressure gauge |
Also Published As
Publication number | Publication date |
---|---|
CN106780398B (en) | 2019-07-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110930327B (en) | Video denoising method based on cascade depth residual error network | |
CN111798396A (en) | Multifunctional image processing method based on wavelet transformation | |
CN109685730B (en) | Wavelet denoising method based on self-adaptive non-local mean value | |
CN108876741A (en) | A kind of image enchancing method under the conditions of complex illumination | |
CN107292316A (en) | A kind of method of the improving image definition based on rarefaction representation | |
CN111915518B (en) | Hyperspectral image denoising method based on triple low-rank model | |
Liu et al. | Image reconstruction using deep learning | |
CN107784639B (en) | Improved multilateral filtering denoising method for remote sensing image of unmanned aerial vehicle | |
CN104135664A (en) | Method for digital processing of medical image | |
CN109003247B (en) | Method for removing color image mixed noise | |
CN105719254A (en) | Image denoising method and system | |
CN109064413B (en) | Image contrast enhancement method and image acquisition medical equipment adopting same | |
CN110349101A (en) | A kind of multiple dimensioned united Wavelet image denoising method | |
CN106780398B (en) | A kind of image de-noising method based on noise prediction | |
CN113450340A (en) | Skin texture detecting system | |
CN110175959B (en) | Typhoon cloud picture enhancement method | |
CN115760773A (en) | Noise suppression method and system for hyperspectral medical image acquisition | |
Budhiraja et al. | Effect of pre-processing on MST based infrared and visible image fusion | |
Thanh et al. | Single image dehazing with optimal color channels and nonlinear transformation | |
CN113888405A (en) | Denoising and demosaicing method based on clustering self-adaptive expansion convolutional neural network | |
CN108038828B (en) | Image denoising method based on self-adaptive weighted total variation | |
CN108010000B (en) | Compressed image sharpening method based on adjacent feature subspace fitting | |
CN111539894A (en) | Novel image enhancement method | |
Kushwaha et al. | Robust denoising technique for ultrasound images by splicing of low rank filter and principal component analysis | |
KR102480448B1 (en) | Image compression system and method based on CNN |
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 | ||
EE01 | Entry into force of recordation of patent licensing contract | ||
EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20170531 Assignee: Anhui Sutong Intelligent Technology Co.,Ltd. Assignor: NANJING University OF POSTS AND TELECOMMUNICATIONS Contract record no.: X2021980006185 Denomination of invention: An image denoising method based on noise prediction Granted publication date: 20190723 License type: Common License Record date: 20210716 |