CN105678704B - A kind of non local intermediate value blind landing method for de-noising of view-based access control model perception - Google Patents
A kind of non local intermediate value blind landing method for de-noising of view-based access control model perception Download PDFInfo
- Publication number
- CN105678704B CN105678704B CN201510738612.5A CN201510738612A CN105678704B CN 105678704 B CN105678704 B CN 105678704B CN 201510738612 A CN201510738612 A CN 201510738612A CN 105678704 B CN105678704 B CN 105678704B
- Authority
- CN
- China
- Prior art keywords
- pixel
- noise
- image
- vision
- peels
- 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
- 238000000034 method Methods 0.000 title claims abstract description 46
- 230000008447 perception Effects 0.000 title claims abstract description 16
- 230000004438 eyesight Effects 0.000 claims abstract description 56
- 230000009467 reduction Effects 0.000 claims abstract description 28
- 238000004364 calculation method Methods 0.000 claims abstract description 17
- 230000000007 visual effect Effects 0.000 claims abstract description 15
- 230000008439 repair process Effects 0.000 claims abstract description 12
- 238000011946 reduction process Methods 0.000 claims abstract description 11
- 230000003044 adaptive effect Effects 0.000 claims abstract description 10
- 238000010276 construction Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 10
- 230000004927 fusion Effects 0.000 claims description 8
- 238000005259 measurement Methods 0.000 claims description 7
- 238000001514 detection method Methods 0.000 claims description 6
- 230000016776 visual perception Effects 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 5
- 238000012804 iterative process Methods 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims description 2
- 238000013139 quantization Methods 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 7
- 238000012545 processing Methods 0.000 description 6
- 238000013461 design Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000004069 differentiation Effects 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000014759 maintenance of location Effects 0.000 description 2
- 235000002566 Capsicum Nutrition 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 241001504519 Papio ursinus Species 0.000 description 1
- 239000006002 Pepper Substances 0.000 description 1
- 235000016761 Piper aduncum Nutrition 0.000 description 1
- 235000017804 Piper guineense Nutrition 0.000 description 1
- 244000203593 Piper nigrum Species 0.000 description 1
- 235000008184 Piper nigrum Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 210000001367 artery Anatomy 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002955 isolation Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 230000000116 mitigating effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 210000003462 vein Anatomy 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
Abstract
The present invention provides a kind of non local intermediate value blind landing method for de-noising of view-based access control model perception, includes the following steps：Vision based on pixel in digital picture, which peels off, estimates construction impulsive noise blind tester, and vision peels off the vision general character estimated by quantifying different model impulsive noises, merges different visual signature quantized results and obtains；The non-local information for extracting image, constructs non local median calculation model；It peels off to estimate according to vision and calculates regularization parameter with non-local information, establish non local intermediate value regularization term；Non local intermediate value noise reduction functional model is built, it is adaptive to repair noise in image pixel.The blind landing method for de-noising of the present invention, according to impulsive noise visual characteristic, image itself self-similarity and outlier data digging, be uniformly processed in digital picture different models, density impulsive noise, noise pixel is difficult to effectively repair caused by can solving the problems, such as in practical noise reduction process unknown impulse model, high density noise and the multi-modal complexity of image.
Description
Technical field
The present invention relates to technical field of image processing, especially denoising digital picture technology, in particular to a kind of base
In the non local intermediate value blind landing method for de-noising of visual perception, suitable for making an uproar to the blind landing of Unknown Model impulsive noise in digital picture.
Background technology
Impulsive noise is a kind of common interference signal in digital picture, in the acquisition, transmission and storing process of image
In, it is generated because of factors such as the not perfect of imaging system, transmission medium and recording equipment, mistakes.It is usual according to brightness Distribution value
Impulsive noise is divided into three kinds, is fixed value, random value and fixed value random value mixed type respectively.Inhibit in digital picture
Impulsive noise is the premise and basis of image analysis, understanding and identification and a key points and difficulties problem in the field.For
Specific impulse model, domestic and international research institution and researcher have carried out extensive research, have obtained a large amount of noise-reduction method, always
For, two class of transform domain noise reduction and spatial domain noise reduction can be divided into.
The thinking of transform domain noise-reduction method is to convert observed image, inhibits noise in the transform domain as illustrated, then passes through
Inverse transformation obtains final noise reduction result.Such methods are first with the characteristic distributions of coefficient in transform domain and the sparsity of dictionary representative domain
Test, have powerful multi-resolution analysis and rarefaction representation ability, but coefficient it is complicated for operation, to parameter setting and primary condition dependence
By force, and usually not global solution is easily introduced deceptive information when repairing high density noise image, complicated image, breaks ring comparison
Degree such as generates " ring ", " ladder ", " overlapping ".
Spatial domain noise-reduction method is the impulse noise mitigation directly in the spatial domain of image, and in contrast, this method is existing
Have in technology and apply relative maturity, noise reduction result is also closer to visual perception.The spatial domain minimizing technology of impulsive noise substantially may be used
It is divided into two class of linear and nonlinear.Mean filter, medium filtering and its innovatory algorithm are spatial domain filter algorithms the most typical, but
Just with mean value, intermediate value and its simple deformation to noise pixel reparation, assignment precision is low, may result in noise reduction result mould
Paste or detailed information are lost.Theoretical and experiment shows that the regularization impulsive noise minimizing technology based on energy functional model can be with
Noise is effectively inhibited, and more fully protects the details of image.Around the design of regularization model, the choosing of regularization parameter
It takes, three work of solution of object function, domestic and international researcher proposes l1Norm+guarantor side regularization term, l1Norm+full change
Poor item, l1Norm+partial differential constraint, l1Norm+lpThe outstanding algorithm such as norm constraint item.Under normal circumstances, these methods have mostly
There is ideal anti-acoustic capability, can inhibit pulse and be effectively protected the details of image, but on condition that prior-constrained accurate and reliable, just
Then change parameter to choose rationally.When choosing regularization parameter, algorithm uses advance unified definition mostly at present, then by a large amount of real
Test the mode of optimization.But parameter of consistency value is defined to the noise pixel of different characteristic in image so that the complex region of image is high
Density noise region fidelity and smooth unbalance, complicated image, the reparation accuracy reduction of high density noise image.
To sum up, existing most method impulse model, noise life density it is known that when relatively easy complex pattern to be repaired can obtain
Obtain preferable noise reduction result.But in view of in practical noise reduction process, the specific mould of impulse noise form images can seldom be known in advance
The complexity of type, density and complex pattern to be repaired, thus it is related to the blind Detecting in image to Unknown Model impulsive noise, to not
With characteristic area pixel self-adapting detecting, repair and to high density impulsive noise the problems such as effectively removing when, it is existing
Noise-reduction method is difficult to be effectively treated.
Invention content
It is in view of the defects existing in the prior art or insufficient, the present invention is directed to propose a kind of perception of view-based access control model it is non local in
It is worth blind landing method for de-noising, can be effectively inhibited in the case of the complexity of unknown pulse noise model, noise density and image
Noise, and completely protect the detailed information of image.
Another object of the present invention is to provide a kind of blind denoising devices of the impulsive noise of view-based access control model perception, and
The computer system that a kind of non local intermediate value blind landing for realizing the perception of aforementioned view-based access control model is made an uproar.
The above-mentioned purpose of the present invention realizes that dependent claims are to select else or have by the technical characteristic of independent claims
The mode of profit develops the technical characteristic of independent claims.
To reach above-mentioned purpose, the first aspect of the present invention proposes that a kind of non local intermediate value blind landing of view-based access control model perception is made an uproar
Method includes the following steps：
Step 1, the vision based on pixel in digital picture peel off and estimate, and construct impulsive noise blind tester, described regards
Feel the vision general character for peeling off and estimating by quantifying different model impulsive noises, merges different visual signature quantized results and obtain；
Step 2, the non-local information for extracting image, construct non local median calculation model；
Step 3, peel off to estimate according to vision calculates regularization parameter with non-local information, establishes non local intermediate value regularization
；
Step 4 establishes non local intermediate value noise reduction functional model according to step 2,3, adaptive to repair noise in image pixel.
According to the disclosure, another aspect of the present invention also proposes that a kind of blind landing of the impulsive noise of view-based access control model perception is made an uproar dress
It sets, including：
It peels off and estimates for the vision based on pixel in digital picture, construct the first module of impulsive noise blind tester,
The vision peels off the vision general character estimated by quantifying different model impulsive noises, merges different visual signature quantized results
And it obtains；
Non-local information for extracting image constructs the second module of non local median calculation model；
Regularization parameter is calculated with non-local information for peeling off to estimate according to vision, establishes non local intermediate value regularization term
Third module；
For according to constructed by aforementioned second module non local median calculation model and the non-office that is established of third module
Portion's intermediate value regularization term builds non local intermediate value noise reduction functional model, which is configured for certainly
It adapts to repair noise in image pixel.
Improvement according to the present invention, the third aspect of the present invention also propose a kind of non-office perceived for realizing view-based access control model
The computer system that portion's intermediate value blind landing is made an uproar, the computer system include：
Memory；
One or more processors；
One or more modules, the one or more module are stored in the memory and are configured to by described one
A or multiple processors execute, and one or more of modules include the module for executing following processing：
It peels off and estimates for the vision based on pixel in digital picture, construct the first module of impulsive noise blind tester,
The vision peels off the vision general character estimated by quantifying different model impulsive noises, merges different visual signature quantized results
And it obtains；
Non-local information for extracting image constructs the second module of non local median calculation model；
Regularization parameter is calculated with non-local information for peeling off to estimate according to vision, establishes non local intermediate value regularization term
Third module；
For according to constructed by aforementioned second module non local median calculation model and the non-office that is established of third module
Portion's intermediate value regularization term builds non local intermediate value noise reduction functional model, which is configured for certainly
It adapts to repair the noise pixel in image.
Compared with prior art, blind noise reduction schemes proposed by the invention have significant advantageous effect：
1. quantify and merged different model impulsive noises to peel off characteristic from visual angle, propose that pixel vision peels off survey
Degree, constructs impulsive noise blind tester, is the unified differentiation of impulsive noise of different models, realizes that Unknown Model impulsive noise is blind
Detection；
2. devising non local intermediate value impulsive noise minimizing technology, adaptive regularization parameter, in conjunction with non local intermediate value structure
Noise reduction functional model is made, the prior-constrained of object function is increased, to improve the reparation precision of noise pixel；
3. noise is effectively inhibited in the case of unknown pulse noise model noise density and the complexity of image,
And completely protect the detailed information of image.
It should be appreciated that as long as aforementioned concepts and all combinations additionally conceived that describe in greater detail below are at this
Sample design it is not conflicting in the case of can be viewed as the disclosure subject matter a part.In addition, required guarantor
All combinations of the theme of shield are considered as a part for the subject matter of the disclosure.
Can be more fully appreciated from the following description in conjunction with attached drawing present invention teach that foregoing and other aspect, reality
Apply example and feature.The feature and/or advantageous effect of other additional aspects such as illustrative embodiments of the present invention will be below
Description in it is obvious, or by according to present invention teach that specific implementation mode practice in learn.
Description of the drawings
Attached drawing is not intended to drawn to scale.In the accompanying drawings, identical or approximately uniform group each of is shown in each figure
It can be indicated by the same numeral at part.For clarity, in each figure, not each component part is labeled.
Now, by example and the embodiments of various aspects of the invention will be described in reference to the drawings, wherein：
Fig. 1 is the flow of the non local intermediate value blind landing method for de-noising perceived according to the view-based access control model of certain embodiments of the invention
Figure.
Fig. 2 a-2d are the image schematic diagram of two class impulse disturbances respectively (noise density is 30%).
Fig. 3 a-3c are the X-ray images and its noise reduction process result schematic diagram by 50% random value noise jamming respectively.
Fig. 4 a-4c's is the tongue fur image done by 70% fixed value noise and its noise reduction process result schematic diagram respectively.
Specific implementation mode
In order to know more about the technology contents of the present invention, spy lifts specific embodiment and institute's accompanying drawings is coordinated to be described as follows.
Various aspects with reference to the accompanying drawings to describe the present invention in the disclosure, shown in the drawings of the embodiment of many explanations.
It is not intended to cover all aspects of the invention for embodiment of the disclosure.It should be appreciated that a variety of designs and reality presented hereinbefore
Those of apply example, and describe in more detail below design and embodiment can in many ways in any one come it is real
It applies, this is because design disclosed in this invention and embodiment are not limited to any embodiment.In addition, disclosed by the invention one
A little aspects can be used alone, or otherwise any appropriately combined be used with disclosed by the invention.
According to an embodiment of the invention, on the whole, the non local intermediate value of view-based access control model proposed by the invention perception is blind
Digital picture is uniformly processed according to impulsive noise visual characteristic, image itself self-similarity and outlier data digging in noise-reduction method
The impulsive noise of middle difference model, density, it is intended to solve unknown impulse model in practical noise reduction process, high density noise and image
Noise pixel caused by multi-modal complexity is difficult to effective reparation problem.
Two stages are generally comprised in the process that entire blind landing is made an uproar, are respectively：1) vision based on pixel in digital picture
It peels off and estimates construction impulsive noise blind tester, to the blind differentiation of Unknown Model impulsive noise in digital picture；2) in noise pixel
Assignment phase establishes cost functional model based on non local intermediate value noise reduction algorithm, the adaptive noise pixel repaired in image.
Vision above-mentioned, which peels off, to be estimated, and by quantifying the vision general character of different model impulsive noises, it is special to merge different visions
The acquisition for levying quantized result, the vision that each pixel is calculated with this are peeled off and estimate, and measurement foundation is provided for the detection of noise pixel.
As described in following specific embodiments, the vision general character of the noise pulse noise of different models is mainly reflected in
Three aspects：Spatial distribution is isolated, connectivity is poor, brightness is abnormal.The blind landing method for de-noising of the present invention is intended to total using these visions
Property is quantified, and is merged to quantized result, and estimated to calculate the vision of pixel in digital picture and peel off with this.
On the basis of fusion, we build blind tester, to the blind differentiation of Unknown Model impulsive noise in digital picture.
Then, using adaptive to be carried out to the noise in digital picture based on the cost functional model of non local intermediate value noise reduction algorithm
It repairs.
Shown in below in conjunction with the accompanying drawings, the realization of the blind landing method for de-noising of previous embodiment is more specifically described.
In conjunction with shown in Fig. 1, according to an embodiment of the invention, a kind of non local intermediate value blind landing side of making an uproar of view-based access control model perception
Method includes the following steps：
Step 1, the vision based on pixel in digital picture peel off and estimate, and construct impulsive noise blind tester, described regards
Feel the vision general character for peeling off and estimating by quantifying different model impulsive noises, merges different visual signature quantized results and obtain；
Step 2, the non-local information for extracting image, construct non local median calculation model；
Step 3, peel off to estimate according to vision calculates regularization parameter with non-local information, establishes non local intermediate value regularization
；
Step 4 establishes non local intermediate value noise reduction functional model according to step 2,3, adaptive to repair noise in image pixel.
As optional example, abovementioned steps 1 are when realizing, including the quantization of impulsive noise vision general character, fusion quantization knot
Fruit simultaneously builds two processes of blind tester, is specifically described separately below.
First, in conjunction with the vision common feature of the noise pulse noise of different models, we are to impulsive noise vision general character
The processing of quantization includes：Quantization that the Spatial Outlier of pixel quantifies and the brightness of pixel peels off.
A. the Spatial Outlier quantization of pixel：According to the data characteristics of digital picture part unicom, estimated using Spatial Outlier,
Using the abnormal number mining algorithm based on connectivity, (IM is estimated in the space for calculating arbitrary pixel i:isolation
measurement)IM(i)；
B. the brightness of pixel peels off quantization：Based on Weber-Fechner law, goal in research pixel regional area minimum brightness
It is poor to feel, is peeled off relative to the brightness of its background in conjunction with the local space Quantified pixel i that peels off with this and estimates (LTM：
luminance transition measurement)LTM(i)。
Merge aforementioned pixel Spatial Outlier quantization and pixel brightness peel off quantization as a result, we can obtain often
The vision of a pixel, which peels off, estimates (VPOM:Visual perception outlier measurement) VPOM (i), and
The impulsive noise blind tester estimated that peels off based on pixel vision is constructed on the basis of this.
Since the computation model of non-local mean Denoising Algorithm is derived to solving minimum by independent variable function of P.
minP(Σj∈P(i)ωi,j|P-Pj|2)
In formula, i indicates that object pixel, j indicate non local pixel, PjIt is the image block centered on j, P (i) is pixel i
Self-similar pixel search window, ωi,jIt is the similarity of pixel i and j.
In view of the nonlinear characteristic of impulsive noise, mean value solution in non-local mean algorithm is converted into intermediate value solution can
The assignment accuracy for improving noise pixel, seeks the weighted median of all self-similar pixels of object pixel i --- and non local intermediate value is calculated
Method solves the minimum that P is independent variable function and obtains intermediate value.
Therefore, in aforementioned step 2, we construct non local median calculation model by following manner：
minP(Σj∈P(i)ωi,j|P-Pj|2)
As the aforementioned, i indicates that object pixel, j indicate non local pixel, PjIt is the image block centered on j, P (i) is pixel
The self-similar pixel search window of i, ωi,jIt is the similarity of pixel i and j.
Therefore it solves the minimum that obtained P is independent variable function and obtains non local intermediate value.
Simultaneously as noise pixel contributes very little to the reparation of i, so combining the survey that peels off of pixel in some embodiments
Degree and the fuzzy membership of pixel optimize weights omegai,j.In order to rapidly solve corresponding weighted mean, as preferred example,
The method that self-similar pixel number is limited using given threshold solves minimum, to reduce computation complexity, improves computation model
Convergence rate.
Certainly, in further embodiments, can also be solved using mode well known to existing others, herein no longer
It repeats.
We construct regular terms using energy functional model as frame, using non local intermediate value, establish noise reduction functional model, will
It is more specifically described in the following contents.
In step 3, we are with the vision degree of peeling off of pixel, testing result (the pixel i quilts in iterative process of noise pixel
It is judged as the number T (i) of noise pixel), the local message in region residing for pixel is analyzed, regularization is adaptively determined for pixel
Parameter：
λ (i)=λ0f1(VPOM(i))f2(T(i))
λ0It is Initial regularization parameters, f1And f2It is weighting function.
In step 4, we combine local regular terms, non local median calculation model and adaptive regularization parametric configuration
Non local intermediate value regularization term.
Shown in Fig. 1, to the non local intermediate value blind landing method for de-noising of the view-based access control model perception using previous embodiment
The flow that blind noise reduction process is carried out to digital picture is illustratively illustrated.
Step 1 inputs observed image u of the width by impulse noise interference
The image u of this step input is by impulse noise interference, but specific impulse model is unknown, it may be possible to fixed value arteries and veins
Punch die type or random value impulse model, certainly it could even be possible to being the mixed model of the two.Such as the example of Fig. 2 a-2d, in figure
The noise density of picture is 30%.
The space of any pixel i, brightness peel off and estimate in step 2, calculating observation image u, and fusion calculation result obtains figure
The vision of each pixel, which peels off, as in estimates
A. the Spatial Outlier for calculating any pixel i in image u is estimated
9 × 9 fields according to human eye to the visual perception of brightness centered on by pixel i, utilize the variable thresholding of following formula
LUT(ul), wherein l=i+k, k ∈ [- 4,4], the unicom pixel chain for calculating the pixel includes with maximum unicom pixel chain
The number of pixels of the most pixel chain of number of pixels defines the connectivity parameters C of the pixel.
In formula, ulIndicate that the brightness of current pixel l in pixel chain, LUT (l) are using the brightness of pixel l as the variable of background
Threshold value.
10 pixels with pixel i luminance difference minimums are calculated in 5 × 5 window centered on by pixel l, find out these pictures
The connectivity parameters of element take its intermediate value C1, then again in rounding 5 × 5 windows the connectivity parameters of all pixels intermediate value C2,
Take connectivity measurement IM (i) of the ratio of the two as pixel i.
B. the brightness for calculating any pixel i in image u peels off and estimates
Estimated according to Spatial Outlier, the α for calculating the image block in 5 × 5 window according to the following formula centered on by pixel i is cut
Cut out background luminance of the mean value as regional area：
In formula, uαIt is that α cuts out mean value, n is the number of pixel in image block, ukIt is after arranging n pixel from small to large
K-th of value, takes α=18 here.
Centered on by pixel i in 5 × 5 window, 10 pixels with pixel i luminance difference minimums are calculated, are calculated
The luminance difference S of these pixels and pixel it, t ∈ [1,10], according to the local visual luminance difference of Fechner's law calculating pixel, such as
Following formula：
C. fusion brightness peels off to estimate estimates with Spatial Outlier, and the vision of any pixel i, which peels off, in calculating image u estimates
VPOM (i), the numerical value are to judge whether pixel i belongs to the foundation of noise pixel.
VPOM (i)=β IM (i)+γ LTM (i)
In above formula, β and γ are that brightness peels off the fusion coefficients estimated and estimated with Spatial Outlier, take β=γ=0.5 here.
Step 3 is peeled off using the vision of each pixel in image u and is estimated, and is built the blind tester of following formula, is passed through threshold
Value TkNoise pixel in detection image：
Tk=Tk-10.9, k=1,2,3 ... Kmax
In above formula, k is iterations (noise reduction process of the invention takes iterative processing), KmaxIt is maximum iteration.
The non-local information of any pixel i in step 4, extraction image u
The image block for choosing centered on pixel i 21 × 21 calculates pixel i's in this block of pixels using following kernel function
Self-similar pixel weights omegai,j：
In above formula, uiAnd ujIt is the pixel value of i and j, λ=16 respectively.
Step 5, the regularization parameter of any pixel i in image u is calculated
With the testing result of the vision degree of peeling off of pixel, noise pixel, (pixel i is judged as noise pixel in iterative process
Number T (i)), adaptively determine regularization parameter for pixel.
λ (i)=λ0f1(VPOM(i))f2(T(i))
λ0It is Initial regularization parameters, takes λ here0=0.01, f1And f2It is weighting function, wherein
In above formula, VPOM (i+k) is that the vision of pixel in 3 × 3 field centered on pixel i peels off measure value.
K in above formulamaxIndicate the maximum iteration during noise reduction process.
Step 6, target noise reduction functional model F is establishedr:RM×N→ R, to pixel to be repaired in image u, (step 3 detects
Noise pixel) carry out valuation reparation
In above formula, V ≡ { 1,2 ..., M } × { 1,2 ..., N } indicates that a width size is the image of M × N, and r indicates noise reduction
Image is repaired, i here indicates target pixel points, and j is the self-similar pixel of pixel i, and Q is most like with i in self-similar pixel
49 pixels, riIndicate the reparation of pixel i as a result, rjThe reparation of pixel j is indicated as a result, the value of λ is 16.
By seeking Fr(u) for minimum to the noise pixel valuation reparation in image, the noise reduction of the present embodiment is to change
Finally image u is repaired by iteration gradually to the impulse noise detection reparation in image for algorithm.
The blind landing method for de-noising as described in previous embodiment drops digital picture using this method with reference to some
It makes an uproar the example of processing, further describes its noise reduction.
1) experiment condition windows8, CPU Inter (R) Core (TM) i5,2.5GHz, software platform are
Matlab7.9.1。
First data that emulation is chosen be by 50% random value noise pollution X-ray images, such as Fig. 3 a, second
A data are by 70% fixed value noise pollution tongue fur image 4a.Third data are interfered by 30% mixed noise
Lena images, Baboon images, Goldhill images, Boat images, Pepper images.
2) experiment content and result
Use the side of traditional non-local mean method (NLM methods) and previous embodiment respectively under these experimental conditions
Method handles first, second experimental data.The noise reduction result that NLM methods obtain such as Fig. 3 b, Fig. 4 b；The noise reduction that the present invention obtains
As a result such as Fig. 3 c, Fig. 4 c.
Compare Fig. 3 b, Fig. 3 c and Fig. 4 b, Fig. 4 c can be seen that the NLM method noise-reduced image loss of detail of the prior art
Than more serious, there is a degree of destruction in the edge of image, and the method for present invention not only can be effective
Removal noise can keep the details of image again simultaneously, and the NLM methods of the prior art are substantially better than from visual effect.
It utilizes NLM methods and the method for the present invention to third data noise reduction process under these experimental conditions, calculates noise reduction
The overall peaks signal-to-noise ratio PSNR and mean absolute error MAE of result images, as a result such as table 1.
The PSNR values and MAE values of the image repair result of 1-30% impulse noise interference of table
As it can be seen from table 1 the method for present invention is apparent in the noise reduction of mixed model impulsive noise
Better than NLM algorithms, reflect that the PSNR values of picture quality are higher, and reflects that the MAE values of image detail loss are smaller.
To sum up, the method for the present invention has compared with traditional NLM methods in terms of removing various types of impulsive noises bright
Aobvious advantage, anti-acoustic capability is more preferable, and acquired results image PSNR is significantly improved, and details protection is more complete, and realizes unknown
The blind Detecting of impulse model noise and adaptive reparation to noise pixel in high density noise image and complicated image, improve
The accuracy of noise pixel reparation.
According to the disclosure, the invention further relates to a kind of blind denoising devices of the impulsive noise of view-based access control model perception, including：
It peels off and estimates for the vision based on pixel in digital picture, construct the first module of impulsive noise blind tester,
The vision peels off the vision general character estimated by quantifying different model impulsive noises, merges different visual signature quantized results
And it obtains；
Non-local information for extracting image constructs the second module of non local median calculation model；
Regularization parameter is calculated with non-local information for peeling off to estimate according to vision, establishes non local intermediate value regularization term
Third module；
For according to constructed by aforementioned second module non local median calculation model and the non-office that is established of third module
Portion's intermediate value regularization term builds non local intermediate value noise reduction functional model, which is configured for certainly
It adapts to repair noise in image pixel.
It should be appreciated that the first module, the second module, third module and the 4th module that the present embodiment is proposed, work(
It can, act on and effect is said in the description of the non local intermediate value blind landing method for de-noising of above view-based access control model perception
It is bright, realization method and exemplary illustration is done in the embodiment previously with regard to blind landing method for de-noising, details are not described herein.
Aforementioned embodiments according to the present invention, for example, view-based access control model perception non local intermediate value blind landing method for de-noising and base
In the blind denoising device of non local intermediate value of visual perception, the present invention also propose it is a kind of perceived for realizing view-based access control model it is non local
The computer system that intermediate value blind landing is made an uproar, the computer system include：
Memory；
One or more processors；
One or more modules, the one or more module are stored in the memory and are configured to by described one
A or multiple processors execute, and one or more of modules include the module for executing following processing：
It peels off and estimates for the vision based on pixel in digital picture, construct the first module of impulsive noise blind tester,
The vision peels off the vision general character estimated by quantifying different model impulsive noises, merges different visual signature quantized results
And it obtains；
Non-local information for extracting image constructs the second module of non local median calculation model；
Regularization parameter is calculated with non-local information for peeling off to estimate according to vision, establishes non local intermediate value regularization term
Third module；
For according to constructed by aforementioned second module non local median calculation model and the non-office that is established of third module
Portion's intermediate value regularization term builds non local intermediate value noise reduction functional model, which is configured for certainly
It adapts to repair the noise pixel in image.
It should be appreciated that memory above-mentioned is used to execute for the processor for storing program and data.These storages
Device, such as can be the memory using disk as storage medium, or the memory etc. based on flash chip.
Obviously, in the computer system of the present embodiment, the module of these storages can be executed by one or more processors
And realize blind noise reduction process described in the non local intermediate value blind landing method for de-noising of aforementioned view-based access control model perception, reach to unknown pulse
The blind Detecting of plant noise and adaptive reparation to noise pixel in high density noise image and complicated image improve noise picture
The accuracy that element is repaired.
Although the present invention has been disclosed as a preferred embodiment, however, it is not to limit the invention.Skill belonging to the present invention
Has usually intellectual in art field, without departing from the spirit and scope of the present invention, when can be used for a variety of modifications and variations.Cause
This, the scope of protection of the present invention is defined by those of the claims.
Claims (1)
1. a kind of non local intermediate value blind landing method for de-noising of view-based access control model perception, which is characterized in that include the following steps：
Step 1, the vision based on pixel in digital picture, which peel off, estimates construction impulsive noise blind tester, and the vision peels off
Estimate the vision general character by quantifying different model impulsive noises, merges different visual signature quantized results and obtain；
Step 2, the non-local information for extracting image, construct non local median calculation model；
Step 3, peel off to estimate according to vision calculates regularization parameter with non-local information, establishes non local intermediate value regularization term；
Step 4 establishes non local intermediate value noise reduction functional model according to step 2,3, adaptive to repair noise in image pixel；
Wherein, in the step 1, the vision based on pixel in digital picture, which peels off, estimates the tool of construction impulsive noise blind tester
Body is realized：
1-1) one is calculated by the Spatial Outlier of any pixel i in the image u of impulse noise interference to be estimated；
The brightness for 1-2) calculating any pixel i in aforementioned image u peels off and estimates；
1-3) fusion brightness peels off to estimate estimates with Spatial Outlier, and the vision of any pixel i, which peels off, in calculating image u estimates
VPOM (i), the numerical value are to judge whether pixel i belongs to the foundation of noise pixel；And
It 1-4) is peeled off using the vision of each pixel in image u and estimates VPOM (i), built the blind tester of following formula, pass through
Threshold value TkNoise pixel in detection image：
Tk=Tk-10.9, k=1,2,3 ... Kmax
In above formula, k is iterations, KmaxIt is maximum iteration；
Wherein, the step 1-1) in, the calculation that the Spatial Outlier of any pixel i is estimated is as follows：
9 × 9 neighborhoods according to human eye to the visual perception of brightness centered on by pixel i, utilize the variable thresholding LUT of following formula
(ul), wherein l=i+k, k ∈ [- 4,4] calculate the connected pixel chain of the pixel, with maximum connected pixel chain, that is, include picture
The number of pixels of the most pixel chain of prime number mesh defines the connectivity parameters C of the pixel：
In formula, ulIndicate the brightness of current pixel l in pixel chain, LUT (ul) it is using the brightness of pixel l as the variable thresholding of background；
10 pixels with pixel i luminance difference minimums are calculated in 5 × 5 window centered on by pixel l, find out these pixels
Connectivity parameters take its intermediate value C1, then again in rounding 5 × 5 windows the connectivity parameters of all pixels intermediate value C2, take two
The ratio of person the connectivity measurement IM (i) as pixel i, the connectivity measurement IM (i) are that the Spatial Outlier of pixel i is estimated；
Also, abovementioned steps 1-2) in the brightness calculation estimated that peels off it is as follows：
Estimated according to Spatial Outlier, the α for calculating the image block in 5 × 5 image block according to the following formula centered on by pixel i is cut out
Background luminance of the mean value as regional area：
In formula, uαIt is that α cuts out mean value, n is the number of pixel in image block, ukIt is the kth after arranging n pixel from small to large
A value, α=18；
Centered on by pixel i in 5 × 5 window, 10 pixels with pixel i luminance difference minimums are calculated, these pixels are calculated
With the luminance difference S of pixel it, t ∈ [1,10], according to the local visual luminance difference of Fechner's law calculating pixel, such as following formula：
The local visual luminance difference LTM (i) being calculated according to this is that the brightness of pixel i peels off and estimates；
And in step 1-3) in, it peels off to estimate using following formula fusion brightness and estimate with Spatial Outlier, calculate any in image u
The vision of pixel i, which peels off, estimates VPOM (i)：
VPOM (i)=β IM (i)+γ LTM (i)
In above formula, β and γ are that brightness peels off the fusion coefficients estimated and estimated with Spatial Outlier, take β=γ=0.5；
Wherein, the specific implementation of abovementioned steps 2 includes：
The image block for choosing centered on pixel i 21 × 21, in this image block using following kernel function calculate pixel i from phase
Like pixel weight ωi,j：
In above formula, uiAnd ujIt is the pixel value of i and j, λ=16 respectively；
Wherein, the specific implementation of the step 3 includes：
With the testing result of the vision degree of peeling off of pixel, noise pixel, regularization parameter λ (i) is adaptively determined for pixel：
λ (i)=λ0f1(VPOM(i))f2(T(i))
λ0It is Initial regularization parameters, takes λ here0=0.01, f1And f2It is weighting function, T (i) is the detection knot of noise pixel
Fruit, i.e., pixel i is judged as the number of noise pixel in iterative process；
Wherein,
In above formula, VPOM (i+k) is that the vision of pixel in 3 × 3 neighborhood centered on pixel i peels off measure value；
K in above formulamaxIndicate the maximum iteration during noise reduction process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510738612.5A CN105678704B (en) | 2015-11-02 | 2015-11-02 | A kind of non local intermediate value blind landing method for de-noising of view-based access control model perception |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510738612.5A CN105678704B (en) | 2015-11-02 | 2015-11-02 | A kind of non local intermediate value blind landing method for de-noising of view-based access control model perception |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105678704A CN105678704A (en) | 2016-06-15 |
| CN105678704B true CN105678704B (en) | 2018-09-25 |
Family
ID=56946801
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510738612.5A Active CN105678704B (en) | 2015-11-02 | 2015-11-02 | A kind of non local intermediate value blind landing method for de-noising of view-based access control model perception |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105678704B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107832967B (en) * | 2017-11-23 | 2021-09-14 | 福建农林大学 | Sound scene coordination degree dynamic evaluation method suitable for bamboo forest space |
| KR20200116763A (en) | 2019-04-02 | 2020-10-13 | 삼성전자주식회사 | Method and apparatus for processing similarity using key-value coupling |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102209385A (en) * | 2011-05-25 | 2011-10-05 | 厦门雅迅网络股份有限公司 | Method for calculating position of base station based on spatial outlier data mining algorithm |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9063226B2 (en) * | 2009-01-14 | 2015-06-23 | Microsoft Technology Licensing, Llc | Detecting spatial outliers in a location entity dataset |
-
2015
- 2015-11-02 CN CN201510738612.5A patent/CN105678704B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102209385A (en) * | 2011-05-25 | 2011-10-05 | 厦门雅迅网络股份有限公司 | Method for calculating position of base station based on spatial outlier data mining algorithm |
Non-Patent Citations (3)
| Title |
|---|
| Adaptive Impulse Detection Using Center-Weighted;Tao chen et al;《SIGNAL PROCESSING LETTERS》;20010131;第8卷(第1期);全文 * |
| 一种有效的脉冲噪声去除算法;胡积宝 等;《计算机工程与应用》;20140601;第50卷(第11期);全文 * |
| 一种有效的脉冲噪声去除算法研究;朱柱;《信息与电脑》;20150723(第14期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105678704A (en) | 2016-06-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yu et al. | Image denoising using trivariate shrinkage filter in the wavelet domain and joint bilateral filter in the spatial domain | |
| CN108154519A (en) | Dividing method, device and the storage medium of eye fundus image medium vessels | |
| CN108629343B (en) | License plate positioning method and system based on edge detection and improved Harris corner detection | |
| CN104574413B (en) | Blood vessel bifurcation extracting method and system of lung CT picture | |
| CN105184216A (en) | Cardiac second region palm print digital extraction method | |
| CN104504686A (en) | Hyper-spectral image abnormity detection method adopting local self-adaptive threshold segmentation | |
| US20230014823A1 (en) | Defect detection in image space | |
| Zhao et al. | Infrared and visible image fusion algorithm based on saliency detection and adaptive double-channel spiking cortical model | |
| CN115841447A (en) | Detection method for surface defects of magnetic shoe | |
| CN104657951A (en) | Multiplicative noise removal method for image | |
| CN113837974A (en) | NSST (non-subsampled contourlet transform) domain power equipment infrared image enhancement method based on improved BEEPS (Bayesian particle swarm optimization) filtering algorithm | |
| Almoussa et al. | Automated vasculature extraction from placenta images | |
| Mehena | Medical image edge detection using modified morphological edge detection approach | |
| CN105678704B (en) | A kind of non local intermediate value blind landing method for de-noising of view-based access control model perception | |
| Ferraioli et al. | A novel cost function for despeckling using convolutional neural networks | |
| Hue et al. | Image denoising with overlapping group sparsity and second order total variation regularization | |
| CN114399480A (en) | Method and device for detecting severity of vegetable leaf disease | |
| Asghar et al. | Automatic enhancement of digital images using cubic Bézier curve and Fourier transformation | |
| Amil et al. | Bilateral histogram equalization with pre-processing for contrast enhancement | |
| Ghosh et al. | A novel intuitionistic fuzzy soft set based colonogram enhancement for polyps localization | |
| Siddique et al. | A robust segmentation of blood vessels in retinal images | |
| CN109559290A (en) | A kind of image denoising method of the asymmetric jump connection of depth | |
| Malik et al. | Contrast enhancement and smoothing of CT images for diagnosis | |
| CN115272141A (en) | Hyperspectral image denoising method, device, system and medium | |
| Keerthivasan et al. | Speckle noise reduction in ultrasound images using fuzzy logic based on histogram and directional differences |
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 |