CN107845081A - A kind of Magnetic Resonance Image Denoising - Google Patents
A kind of Magnetic Resonance Image Denoising Download PDFInfo
- Publication number
- CN107845081A CN107845081A CN201711379469.0A CN201711379469A CN107845081A CN 107845081 A CN107845081 A CN 107845081A CN 201711379469 A CN201711379469 A CN 201711379469A CN 107845081 A CN107845081 A CN 107845081A
- Authority
- CN
- China
- Prior art keywords
- mrow
- image
- msub
- pixel
- image block
- 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
- 238000012937 correction Methods 0.000 claims abstract description 5
- 238000000034 method Methods 0.000 claims description 38
- 239000011159 matrix material Substances 0.000 claims description 21
- 238000001914 filtration Methods 0.000 claims description 18
- 238000005194 fractionation Methods 0.000 claims description 7
- 239000013256 coordination polymer Substances 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 239000000203 mixture Substances 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 10
- 238000012545 processing Methods 0.000 abstract description 5
- 238000002595 magnetic resonance imaging Methods 0.000 description 10
- 238000001678 elastic recoil detection analysis Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000004321 preservation Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 210000000746 body region Anatomy 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000003706 image smoothing Methods 0.000 description 1
- 238000012633 nuclear imaging Methods 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000005303 weighing Methods 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
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a kind of Magnetic Resonance Image Denoising, including step：Tensor resolution；Image block is split；Calculate similarity weight；Pixel value is estimated；Bias Correction；The advantage of the invention is that：Improve the denoising effect of MRI so that more original image informations can be saved, and reduce the fuzzy of image border part.Improve the processing speed of MRI.It can be combined with other non-local mean denoising innovatory algorithms, for lifting the speed of image procossing, and the effect of denoising can be ensured.
Description
Technical field
The present invention relates to technical field of image processing, more particularly to a kind of Magnetic Resonance Image Denoising.
Background technology
Proportion more and more higher of the medical image in clinical diagnosis and research, status are more and more important.It can be noninvasive
In the case of give doctor to provide the complete accurate image information of inside of human body region of interest, checked to doctor and diagnosis provide
Very important reference frame.The technology of imaging has a lot, including：X ray, tomoscan, ultrasonic wave, nuclear imaging, magnetic resonance
The technological means such as imaging.Wherein magnetic resonance imaging (MRI) technology can provide the organ of gem-pure human body and the figure of tissue
Picture, so being widely applied very much in terms of medical treatment.Due to the limitation of condition, the MRI collected under normal circumstances
Signal to noise ratio all than relatively low.In addition, MRI can degenerate because of the influence of artifact and noise.Establish appropriate denoising mould
Type is extremely important in MRI processing.The research for removing modeling type is all very popular problem all the time, its purpose
In the noise and artifact included in image is removed, the details and marginal information of image are preserved, improves the signal to noise ratio of image.
The method of image denoising has a lot, it is generally understood that carrying out denoising during image is gathered and gathering
Denoising work is carried out to image afterwards.Because the technical conditions of magnetic resonance imaging limit, phase is usually carried out after having gathered
The denoising work of pass, the conventional denoising method of MRI can substantially be divided into two major classes, first, the method for filtering, another
Make the method for transform domain.It can be subdivided into the method for filtering, the method for linear filtering and nonlinear filtering method.Its is hollow
Between filtering and time filtering belong to linear filtering, anisotropic filtering, non local PCA, non-local mean filtering belong to non-linear
Filtering method.Denoising method based on transform domain includes wavelet transformation, warp wavelet, Fourier transformation, removes wave conversion, profile
Wavelet transformation.Non-local mean filters and contourlet conversion is denoising effect the best way in two kinds of filtering methods respectively.
This patent is that it is carried out using the method and adjustment weighing computation method of tensor resolution on the basis of non-local mean filtering
Improve, for the efficiency of boosting algorithm denoising and the effect of denoising.
MRI has very high redundancy, and data volume is very big, non-local mean filtering and noise reduction method
The redundancy properties of image in itself are made full use of, the noise included in image is removed, there is the image compared with high s/n ratio at recovery.It is non-
The basic skills of local mean value filtering is first in global scope, to divide an image into the image block of formed objects, Ran Houli
Its basic step, which is, to be estimated to the pixel value of some point with all image blocks：Current pixel is calculated respectively
The image block at place and the Euclidean distance of other block of pixels, because the size of each image block is identical, so institute in image block
Comprising pixel number it is identical, therefore only need to according to the gray value of each pixel calculate Euclidean distance.Euclidean away from
From with the similitude intensity for judging two image blocks, the similitude intensity of two image blocks is bigger, the central point of two image blocks
Phase knowledge and magnanimity it is higher, when the value to pixel is estimated, the bigger point of similarity weight, to the point to be estimated
Contribution it is bigger.Using identical method, all pixels point can be estimated by traveling through pixel all in whole image
Value.
Tensor resolution, the flat site, edge and angle point of image can be distinguished using tensor resolution in piece image
Part.Its method calculated is to calculate image partial derivative in the x, y direction respectively, and then calculating matrix ranks are K and mark H,
Determinant K and mark H are screened by a threshold value, Huo get not flat site, marginal portion, the original image of angular point portions
Pixel value.In this approach, equal-sized three images can be obtained, the structural information of image is included in marginal portion mostly,
The probability that random noise caused by image is included in angle point region is also bigger.When calculating the similitude of image, can use
Structural information is included and more partly calculated.
Original non-local mean filtering application is to noise Reduction of Magnetic Resonance Images, and the time complexity of algorithm is high, original image
Edge and detailed information do not retained well, denoising effect is poor under the conditions of strong noise.
Original non local denoising method travels through all pixels in the scope of the overall situation, and the time for adding calculating answers
Miscellaneous degree.Meanwhile using the similarity weight for equalizing the pixel in same image block, make the edge and detail section of image
Suppressed, reduce the structural information of image.
A part of improved method has a certain upgrade in the performance of denoising, but is the increase in the run time of algorithm.
Some investigators use random sampling, and reducing the method for image block reduces the run time of program, its consequence
It is the image block limited amount chosen, the pixel value accuracy estimated compared with original algorithm, run time subtracts
It is few, while also have lost the effect of denoising.
The improved method of another part has lifting in the speed of denoising, but to sacrifice the effect of a part of denoising work
For cost.
Some researchers have done many work in lifting noise removal capability, also achieve obvious effect.Use two
Or the collaboration filtering of polyalgorithm improves the performance of denoising to the improved method of non-local mean algorithm progress, adds place
The complexity of reason, so that the time increase of processing.
The content of the invention
The defects of present invention is directed to prior art, there is provided a kind of Magnetic Resonance Image Denoising, can effectively solve the problem that
State the problem of prior art is present.
In order to realize above goal of the invention, the technical scheme that the present invention takes is as follows：
A kind of Magnetic Resonance Image Denoising, comprises the following steps：
Step 1：Tensor resolution；
Step 11：Seek the tensor matrix of noise image；
Using the CP decomposition methods in tensor resolution, X, partial derivative Ix, Iy in Y-direction are calculated respectively, passes through two sides
Upward partial derivative, is calculated respectively：
Ix2=Ix.^2；
Iy2=Iy.^2；
Ixy=Ix.*Iy；
Tensor matrix ST=[Ix2, the Ixy of three value composition noise images；Ixy,Iy2].
Step 12：Extract the HFS in image；
By traveling through all pixels, tensor determinant of a matrix K and mark H is calculated, by dividing different K's and H
It is worth size, noise image is carried out to be split as three parts：Flat, marginal portion, angular point portions；Take marginal portion and
The image of angular point portions, referred to as structured image.
Step 2：Image block is split；
Step 21：Split noise image；Noise image is split as size identical image block, each picture block is used to estimate
The value of a pixel is counted, image block is referred to as search window.
Step 22：Split structural images；
The structured image obtained in step 12 is split, the image block of fractionation is referred to as neighborhood window, the neighbour after fractionation
Area image block forms a Neighborhood matrix.
Step 3：Calculate similarity weight；
Step 31：Neighborhood matrix is clustered using K-means；
K-means methods are used to the Neighborhood matrix in step 22, try to achieve the cluster centre of each neighborhood image block.
Step 32：Calculate the similarity weight between image block；
Formula is used by the cluster centre of previous step：
Wherein,Represent the similarity weight between pixel, Z (xi) normalized parameter is represented, E (xi) is represented
The cluster centre of image block centered on i points, E (xj) represent the cluster centre of the image block centered on j points, and h is filtering
Intensity.
The similarity weight between neighborhood image block is calculated, the similarity weight between neighborhood image block is center pixel
The similarity weight of point, using the similarity relationships between pixel, estimates the value related like vegetarian refreshments.
Step 4：Pixel value is estimated；
In each search window, according to the similarity weight of the pixel calculated in step 32, to search window
The value of central pixel point is estimated that the formula of estimation is as follows：
Wherein, W () is the similarity weight between the pixel estimated in previous step,
In(xj) observed value i.e. actual value of pixel, the value for the pixel that I (xi) is estimated for corresponding to.N tables
Show the number of pixel in image block, xi represents in image block to order for i-th vegetarian refreshments, and xj represents in image block to order for j-th vegetarian refreshments.
Each pixel is corresponding to be exactly the point is estimating it is being institute for pixel value with pixel similarity weight to be estimated
The contribution done, similarity weight is higher, and the ratio shared by the point is also bigger.
Step 5：Bias Correction；
Treating method equation below：
Output result=sqrt (max (0, denoising result/similarity weight sum -2* noise variances)).
Further, the size of search window uses empirical value, radius 3 in step 21.
Further, the radius of neighborhood window takes empirical value 2 in step 22.
Compared with prior art the advantage of the invention is that：
Improve the denoising effect of MRI so that more original image informations can be saved, and reduce image side
Edge point obscures.
Improve the processing speed of MRI.
It can be combined with other non-local mean denoising innovatory algorithms, for lifting the speed of image procossing, and
The effect of denoising can be ensured.
Brief description of the drawings
Fig. 1 is flow chart of the embodiment of the present invention；
Fig. 2 is tensor resolution schematic diagram of the embodiment of the present invention；
Fig. 3 is that noise image of the embodiment of the present invention carries out fractionation schematic diagram.
Embodiment
For the objects, technical solutions and advantages of the present invention are more clearly understood, develop simultaneously embodiment referring to the drawings, right
The present invention is described in further details.
As shown in figure 1, a kind of Magnetic Resonance Image Denoising, comprises the following steps：
Step 1：Tensor resolution；
Step 11：Seek the tensor matrix of noise image；
Using the CP decomposition methods in tensor resolution, X, partial derivative Ix, Iy in Y-direction are calculated respectively, passes through two sides
Upward partial derivative, is calculated respectively：
Ix2=Ix.^2；
Iy2=Iy.^2；
Ixy=Ix.*Iy；
Tensor matrix ST=[Ix2, the Ixy of three value composition noise images；Ixy,Iy2].The basic skills of tensor resolution
As shown in Fig. 2 original image array is estimated with less two matrixes, while main structural information can be protected occasionally
Leave and.
Step 12：Extract the HFS in image；
By traveling through all pixels, tensor determinant of a matrix K and mark H is calculated, by dividing different K's and H
It is worth size, noise image is carried out to be split as three parts：Flat, marginal portion, angular point portions.The structure letter of image
Breath is all stored in the structure division of high frequency central, therefore we only choose the image of marginal portion and angular point portions, here I
Be referred to as structured image；Noise image splits as shown in Figure 3.
By the image after tensor resolution, low frequency part and radio-frequency head are separated, HFS deposits the structure of image
Information and marginal information, our main selection HFSs are used for the similitude for calculating image.In addition, we are pair in the present invention
The tensor resolution that entire image is carried out can reduce the time of calculating.Importantly, one of main points of the present invention are only to choose
Structure division carries out similarity system design between image block, eliminates the influence of smooth.
Step 2：Image block is split；
Step 21：Split noise image；
Noise image is split as size identical image block, each picture block is used for the value for estimating a pixel.Figure
As block, we are referred to as search window, the size of search window we use empirical value, radius 3.We only search at one
Rope window is intraoral to be estimated pixel, without being operated in entire image, on the one hand can be reduced amount of calculation, on the other hand may be used
To reduce the image smoothing that global average band is come, preservation is more structural information.
Step 22：Split structural images；
The structured image obtained in step 12 is split, the image block of fractionation is referred to as neighborhood window, neighborhood window
Radius equally takes empirical value 2, and the neighborhood image block after fractionation forms a Neighborhood matrix.Each row storage of Neighborhood matrix
It is the value of all pixels of a neighborhood window, so, the number of the behavior pixel of matrix, it is classified as the number of image block.
Step 3：Calculate similarity weight；
Step 31：Neighborhood matrix is clustered using K-means；
K-means methods are used to the Neighborhood matrix in step 22, try to achieve the cluster centre of each neighborhood image block.Use
It is one of core procedure of this method that the method for k-means clusters, which carries out cluster, to all pixels point in each neighborhood window
Clustered, there is two advantages, when, reduce participate in weight computing pixel number, can save program operation when
Between；Secondly, increase the difference between border in same neighborhood window, enable the preferably quilt of the marginal portion after equalization
Preserve；Finally, this method is that the method for using cluster in neighborhood window interior with other denoising method differences, without
It is the cluster between image block.
Step 32：Calculate the similarity weight between image block；
Formula is used by the cluster centre of previous step：
The similarity weight between neighborhood image block is calculated, the similarity weight between neighborhood image block is center pixel
The similarity weight of point.What the method for calculating similarity weight was sent out different from traditional calculating method, this method is to use image block
Internal cluster centre calculates similarity weight, clusters and maximizes the distance between class, then carries out mean time smoothness meeting
Reduce.Therefore, it is possible to reduce fuzzy, the preservation more details information of marginal portion.
Step 4：Estimate pixel value；
In each search window, according to the similarity weight of the pixel calculated in step 32, to search window
The value of central pixel point is estimated that the formula of estimation is as follows：
Inside same search window, pixel to be estimated and each other pixel have a similitude to weigh
Weight, the value of the value, the as image after denoising that can obtain required pixel is averaging by weighting.
Step 5：Bias Correction；
MRI figure others image is different, and for its noise based on thermal noise, image can produce puppet after obtaining
Shadow, noise type are interpreted as Lay this noise, so to carry out offset correction to image after carrying out denoising.Common treating method
Equation below：
Output result=sqrt (max (0, denoising result/similarity weight sum -2* noise variances).
One of ordinary skill in the art will be appreciated that embodiment described here is to aid in reader and understands this hair
Bright implementation, it should be understood that protection scope of the present invention is not limited to such especially statement and embodiment.Ability
The those of ordinary skill in domain can be made according to these technical inspirations disclosed by the invention it is various do not depart from essence of the invention its
Its various specific deformations and combination, these deformations and combination are still within the scope of the present invention.
Claims (3)
1. a kind of Magnetic Resonance Image Denoising, it is characterised in that comprise the following steps：
Step 1：Tensor resolution；
Step 11：Seek the tensor matrix of noise image；
Using the CP decomposition methods in tensor resolution, X, partial derivative Ix, Iy in Y-direction are calculated respectively, by both direction
Partial derivative, calculate respectively：
Ix2=Ix.^2；
Iy2=Iy.^2；
Ixy=Ix.*Iy；
Tensor matrix ST=[Ix2, the Ixy of three value composition noise images；Ixy,Iy2]；
Step 12：Extract the HFS in image；
By traveling through all pixels, tensor determinant of a matrix K and mark H is calculated, by dividing being worth for different K and H
Size, noise image is carried out to be split as three parts：Flat, marginal portion, angular point portions；Take marginal portion and angle point
Partial image, referred to as structured image；
Step 2：Image block is split；
Step 21：Split noise image；Noise image is split as size identical image block, each picture block is used to estimate one
The value of individual pixel, image block are referred to as search window；
Step 22：Split structural images；
The structured image obtained in step 12 is split, the image block of fractionation is referred to as neighborhood window, the Neighborhood Graph after fractionation
As block forms a Neighborhood matrix；
Step 3：Calculate similarity weight；
Step 31：Neighborhood matrix is clustered using K-means；
K-means methods are used to the Neighborhood matrix in step 22, try to achieve the cluster centre of each neighborhood image block；
Step 32：Calculate the similarity weight between image block；
Formula is used by the cluster centre of previous step：
<mrow>
<mi>W</mi>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mover>
<mi>x</mi>
<mo>&OverBar;</mo>
</mover>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mover>
<mi>x</mi>
<mo>&OverBar;</mo>
</mover>
<mi>j</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mi>Z</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mrow>
<mo>-</mo>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<mi>E</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>x</mi>
<mi>i</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>E</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>x</mi>
<mi>j</mi>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<msup>
<mi>h</mi>
<mn>2</mn>
</msup>
</mfrac>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
Wherein,Represent the similarity weight between pixel, Z (xi) normalized parameter is represented, E (xi) is represented with i points
Centered on image block cluster centre, E (xj) represents the cluster centre of image block centered on j points, and h is filtering strength；
The similarity weight between neighborhood image block is calculated, the similarity weight between neighborhood image block is central pixel point
Similarity weight, using the similarity relationships between pixel, estimate the value related like vegetarian refreshments；
Step 4：Pixel value is estimated；
In each search window, according to the similarity weight of the pixel calculated in step 32, to search window center
The value of pixel is estimated that the formula of estimation is as follows：
<mrow>
<mi>I</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>&Element;</mo>
<mi>N</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</munder>
<mi>W</mi>
<mrow>
<mo>(</mo>
<msub>
<mover>
<mi>x</mi>
<mo>&OverBar;</mo>
</mover>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mover>
<mi>x</mi>
<mo>&OverBar;</mo>
</mover>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<msub>
<mi>I</mi>
<mi>n</mi>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
Wherein, W () is the similarity weight between the pixel that estimates, I in previous stepn(xj) pixel for corresponding to
The observed value i.e. actual value of point, the value for the pixel that I (xi) is estimated；N represents the number of pixel in image block, xi
Represent to order for i-th in image block vegetarian refreshments, xj represents in image block to order for j-th vegetarian refreshments.
Each pixel is corresponding to be exactly this is estimating it is being to be done for pixel value with pixel similarity weight to be estimated
Contribution, similarity weight is higher, and the ratio shared by the point is also bigger；
Step 5：Bias Correction；
Treating method equation below：
Output result=sqrt (max (0, denoising result/similarity weight sum -2* noise variances)).
A kind of 2. Magnetic Resonance Image Denoising according to claim 1, it is characterised in that：Search window in step 21
Size uses empirical value, radius 3.
A kind of 3. Magnetic Resonance Image Denoising according to claim 1, it is characterised in that：Neighborhood window in step 22
Radius takes empirical value 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711379469.0A CN107845081B (en) | 2017-12-20 | 2017-12-20 | Magnetic resonance image denoising method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711379469.0A CN107845081B (en) | 2017-12-20 | 2017-12-20 | Magnetic resonance image denoising method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107845081A true CN107845081A (en) | 2018-03-27 |
| CN107845081B CN107845081B (en) | 2021-05-11 |
Family
ID=61684007
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711379469.0A Expired - Fee Related CN107845081B (en) | 2017-12-20 | 2017-12-20 | Magnetic resonance image denoising method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107845081B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111091514A (en) * | 2019-12-20 | 2020-05-01 | 广州华端科技有限公司 | Oral cavity CBCT image denoising method and system |
| CN112738532A (en) * | 2020-12-29 | 2021-04-30 | 哈尔滨工业大学芜湖机器人产业技术研究院 | DCT blocking effect phenomenon elimination method based on Gaussian filtering |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103150714A (en) * | 2013-03-12 | 2013-06-12 | 华东师范大学 | Method and device for real-time interactive enhancement of magnetic resonance image |
| CN103955903A (en) * | 2014-05-09 | 2014-07-30 | 东南大学 | Weight window self-adaptation non-local mean image denoising method |
| CN103985099A (en) * | 2014-05-30 | 2014-08-13 | 成都信息工程学院 | Dispersion tensor magnetic resonance image tensor domain non-local mean denoising method |
| US8866936B2 (en) * | 2008-07-24 | 2014-10-21 | Florida State University of Research Foundation | Systems and methods for training an active random field for real-time image denoising |
| CN104680495A (en) * | 2015-03-17 | 2015-06-03 | 武汉倍尔生生物科技有限公司 | Adaptive denoising method for ultrasonic image |
-
2017
- 2017-12-20 CN CN201711379469.0A patent/CN107845081B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8866936B2 (en) * | 2008-07-24 | 2014-10-21 | Florida State University of Research Foundation | Systems and methods for training an active random field for real-time image denoising |
| CN103150714A (en) * | 2013-03-12 | 2013-06-12 | 华东师范大学 | Method and device for real-time interactive enhancement of magnetic resonance image |
| CN103955903A (en) * | 2014-05-09 | 2014-07-30 | 东南大学 | Weight window self-adaptation non-local mean image denoising method |
| CN103985099A (en) * | 2014-05-30 | 2014-08-13 | 成都信息工程学院 | Dispersion tensor magnetic resonance image tensor domain non-local mean denoising method |
| CN104680495A (en) * | 2015-03-17 | 2015-06-03 | 武汉倍尔生生物科技有限公司 | Adaptive denoising method for ultrasonic image |
Non-Patent Citations (2)
| Title |
|---|
| VARDAN PAPYAN，AND ETC: "Multi-Scale Patch-Based Image Restoration", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
| 张小丹等: "K均值聚类和支持向量数据描述的图像超分辨率算法", 《中国图象图形学报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111091514A (en) * | 2019-12-20 | 2020-05-01 | 广州华端科技有限公司 | Oral cavity CBCT image denoising method and system |
| CN111091514B (en) * | 2019-12-20 | 2023-05-05 | 合肥登特菲医疗设备有限公司 | Oral cavity CBCT image denoising method and system |
| CN112738532A (en) * | 2020-12-29 | 2021-04-30 | 哈尔滨工业大学芜湖机器人产业技术研究院 | DCT blocking effect phenomenon elimination method based on Gaussian filtering |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107845081B (en) | 2021-05-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Singh et al. | Multimodal medical image sensor fusion model using sparse K-SVD dictionary learning in nonsubsampled shearlet domain | |
| CN104978715B (en) | A kind of non-local mean image de-noising method based on filter window and parameter adaptive | |
| CN103473767B (en) | The method and system that a kind of soft tissues of abdomen nuclear-magnetism image is cut apart | |
| Khullar et al. | Wavelet-based fMRI analysis: 3-D denoising, signal separation, and validation metrics | |
| Armanious et al. | Retrospective correction of rigid and non-rigid mr motion artifacts using gans | |
| CN103985099B (en) | Dispersion tensor magnetic resonance image tensor domain non-local mean denoising method | |
| US8649585B2 (en) | Method and system for retrospective image combination under minimal total deformation constrain for free-breathing cardiac magnetic resonance imaging with motion correction | |
| Rokade et al. | Automatic detection of hard exudates in retinal images using haar wavelet transform | |
| CN104008537A (en) | Novel noise image fusion method based on CS-CT-CHMM | |
| Yang et al. | An automatic brain tumor image segmentation method based on the U-Net | |
| CN107845081A (en) | A kind of Magnetic Resonance Image Denoising | |
| Guo et al. | Brain tumor segmentation based on attention mechanism and multi-model fusion | |
| Wang et al. | Brain MR image super-resolution using 3D feature attention network | |
| Saraswathi et al. | Brain tumor segmentation and classification using self organizing map | |
| Lepcha et al. | Multimodal medical image fusion based on pixel significance using anisotropic diffusion and cross bilateral filter | |
| Trinh et al. | Medical image denoising using kernel ridge regression | |
| Roychowdhury et al. | A modular framework to predict alzheimer’s disease progression using conditional generative adversarial networks | |
| CN106570880B (en) | In conjunction with the brain tissue MRI image dividing method of fuzzy clustering and markov random file | |
| Akbarpour et al. | Medical image fusion based on nonsubsampled shearlet transform and principal component averaging | |
| CN110619633B (en) | Liver image segmentation method based on multipath filtering strategy | |
| CN115222725B (en) | Medical image fusion method based on NSCT domain PCRGF and dual-channel PCNN | |
| Indhumathi et al. | A comprehensive study of image fusion techniques and their applications | |
| Slavakis et al. | Kernel regression imputation in manifolds via bi-linear modeling: The dynamic-MRI case | |
| Saravanan et al. | An Effective Possibilistic Fuzzy Clustering Method for Tumor Segmentation in MRI brain Images | |
| CN115619693A (en) | Space-time fusion model based on nested multi-scale transformation |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210511 Termination date: 20211220 |