CN104240184B - The evaluation method and system of noise criteria difference - Google Patents
The evaluation method and system of noise criteria difference Download PDFInfo
- Publication number
- CN104240184B CN104240184B CN201310227562.5A CN201310227562A CN104240184B CN 104240184 B CN104240184 B CN 104240184B CN 201310227562 A CN201310227562 A CN 201310227562A CN 104240184 B CN104240184 B CN 104240184B
- Authority
- CN
- China
- Prior art keywords
- noise
- border
- pixel
- raw
- pattern
- 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
- 238000011156 evaluation Methods 0.000 title claims abstract description 17
- 238000000034 method Methods 0.000 claims abstract description 45
- 238000004140 cleaning Methods 0.000 claims abstract description 36
- 230000015572 biosynthetic process Effects 0.000 claims description 14
- 238000005755 formation reaction Methods 0.000 claims description 14
- 210000000988 bone and bone Anatomy 0.000 claims description 9
- 210000001519 tissue Anatomy 0.000 claims description 9
- 238000002059 diagnostic imaging Methods 0.000 claims 3
- 230000006870 function Effects 0.000 description 35
- 230000009467 reduction Effects 0.000 description 19
- 238000004422 calculation algorithm Methods 0.000 description 13
- 238000013170 computed tomography imaging Methods 0.000 description 8
- 230000008859 change Effects 0.000 description 4
- 238000001914 filtration Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000008520 organization Effects 0.000 description 3
- 230000002146 bilateral effect Effects 0.000 description 2
- 230000005855 radiation Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 238000012360 testing method 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
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
- G06T11/005—Specific pre-processing for tomographic reconstruction, e.g. calibration, source positioning, rebinning, scatter correction, retrospective gating
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- 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
- G06T2207/10081—Computed x-ray tomography [CT]
-
- 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
- G06T2207/20004—Adaptive image processing
- G06T2207/20012—Locally adaptive
-
- 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
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The present invention relates to a kind of poor evaluation method of noise criteria, in the method, have received the input picture of a Noise and the input picture is being filtered and after obtaining a smoothed image, the difference for taking the input picture and the smoothed image is raw noise figure.Border in raw noise figure described in pre cleaning, threshold value T or threshold function table T is estimated from the pre cleaning in the noise pattern on borderg, then with the threshold value T or threshold function table TgThe border in raw noise figure is determined, and identified border is removed from raw noise figure, then the estimated noise standard deviation from the noise pattern for removing border.
Description
Technical field
The present invention relates to a kind of evaluation method of noise criteria difference and system, in particular it relates to institute in image noise reduction algorithm
The evaluation method and system of noise criteria difference.
Background technology
Standard deviation refers to a mathematical formulae of the degree for representing the average value that one group of data deviates this group of data.Standard deviation
It is usually used in measurement and the sign of noise.In noise reduction algorithm, such as, in non-local mean filter noise reduction algorithm, noise criteria
Difference is an important parameter of the smoothness of the reconstructed image for controlling finally to obtain.Accurately estimated noise standard deviation has
Very important meaning, if can not accurately estimated noise standard deviation, be able to may have in the image obtained after noise reduction bad
Character, such as, and too many noise, lose resolution ratio, there is artifact etc..However, because noise criteria difference is for different calculating
Machine tomoscan（CT）It is uneven for image, accurately standard of appraisal difference has very high difficulty.
The content of the invention
The present invention relates to a kind of poor evaluation method of noise criteria, in the method, the input picture of Noise have received
And the input picture is filtered and obtained after a smoothed image, take the difference of the input picture and the smoothed image
For raw noise figure.Border in raw noise figure described in pre cleaning, estimates threshold from the pre cleaning in the noise pattern on border
Value T or threshold function table Tg, then by the threshold value T or threshold function table TgFor determining the border in raw noise figure, and by really
Fixed border is removed from raw noise figure, then the estimated noise standard deviation from the noise pattern for removing border.
The invention further relates to a kind of poor evaluation method of noise criteria, in the method, the input figure of Noise have received
Picture is simultaneously filtered to the input picture and obtained after a smoothed image, takes the difference of the input picture and the smoothed image
Volume is raw noise figure.The noise criteria of pixel in raw noise Tu Zhong uniform formations region difference is estimated,
The noise criteria difference and the coordinate of these pixels for the pixel being then based in obtained uniform formation region, pass through
Curve fitting technique obtains the function of noise criteria difference.
Brief description of the drawings
By being described with reference to accompanying drawing for embodiments of the invention, the present invention may be better understood, in the accompanying drawings：
Fig. 1 is the figure of one embodiment of CT imaging devices.
Fig. 2 is the schematic block diagram of CT imaging devices shown in Fig. 1
Fig. 3 be one embodiment of the invention in standard deviation evaluation method flow chart.
Fig. 4 be another embodiment of the present invention in standard deviation evaluation method flow chart.
Fig. 5 is the flow chart of the standard deviation evaluation method in another of the invention embodiment.
Fig. 6（a）Show the noise criteria difference estimated with the method described in the embodiment of the present invention to Noise
Image carries out the image that non-local mean filter noise reduction is obtained.
Fig. 6（b）Show and non-local average filter is integrally carried out to the image of a Noise with 90HU noise criteria difference
The image that ripple noise reduction is obtained.
Fig. 6（c）Show and non-local mean filter drop is carried out to the image of a Noise with 110HU noise criteria difference
Make an uproar obtained image.
Fig. 6（d）Show and non-local mean filter drop is carried out to the image of a Noise with 120HU noise criteria difference
Make an uproar obtained image.
Embodiment
The language of approximation used herein can be used for quantitative expression, show in the case where not changing basic function
Quantity can be allowed to have certain variation.Therefore, the numerical value corrected with the language such as " about ", " left and right " is not limited to the exact value
Itself.In certain embodiments, " about " or " left and right " represent allow its correct numerical value positive and negative 10（10%）Model
Interior change is enclosed, such as, " about 100 " represent can be any numerical value between 90 to 110.In addition, in " the about first numerical value
To second value " statement in, " about " is while correct the first numerical value and second value two values.In some cases, closely
May be relevant with the precision of measuring instrument like language.
In addition to being defined, technology used herein and scientific terminology are with universal with art personnel of the present invention
The identical meanings of understanding.Term " first " used herein, " second " etc. are not offered as any order, quantity or importance, and
It is only intended to distinguish a kind of element and another element.Also, the "a" or "an" does not indicate that the restriction of quantity, but table
Show the relevant item in the presence of one.
Embodiments of the invention are related to a kind of method of standard of appraisal difference, and this method is generally used for various CT imaging systems
In noise reduction algorithm, such as, available for a kind of such as the patent of the U.S. the 7th, 706,497（It is entitled in application on March 14th, 2008
“Methods and apparatus for noise estimation for multi-resolution anisotropic
diffusion filtering”）Described in CT imaging systems noise reduction algorithm in.
Fig. 1 is a kind of schematic diagram of CT imaging systems 10.Fig. 2 is the schematic block diagram of the system 10 shown in Fig. 1.Show described
In the embodiment of example property, CT imaging systems 10 include a pallet 12 for representing third generation CT imaging systems, and the pallet 12 has one
Individual radiographic source 14, can launch X-ray pencil-beam 16 to the detector array 18 of the opposite side on the pallet 12.
The detector array 18 is by a plurality of detector rows（It is not shown）Composition, wherein the detector row includes plural number
Individual detecting element 20, together for sensing the X-ray beam through certain object, such as medical patient 22.The detecting element 20 is each
An electric signal for representing collision radiation beam density is produced, wherein the electric signal for representing collision radiation beam density can table
Show light beam through the Weaken degree after object or patient.During scanning obtains ray projection data, the He of pallet 12
The element installed thereon rotates around a pivot 24.Fig. 2 only show a line detecting element 20（That is detector row）, but
Multi-slice detector array 18 includes the plurality of parallel detector row being made up of detecting element 20, so, just can be by once
Scanning obtains the data for projection of the plurality of parallel layer of correspondence or quasi-parallel layer simultaneously.
The rotation of the pallet 12 and the running of the radiographic source 14 are carried out by the controlling organization 26 of CT imaging systems 10
Control and management.The controlling organization 26 includes providing the He of ray controller 28 of energy and clock signal to the radiographic source 14
For the pallet engine controller 30 for the position for controlling velocity of rotation and pallet 12.Data collecting system in controlling organization 26
32 from the detecting element 20 sampling collection analogue data, and converts the data into data signal, for subsequent treatment.Imagescope
（image reconstructor）34 receive the sampling collection and digitized number of rays from the data collecting system 32
According to progress high speed image reproduction.The image of the reproduction is transfused to computer 36, and described image is stored in great Rong by the computer
Measure in storage device 38.
The computer also receives the instruction inputted by operator's console 40 and sweep parameter, and the console includes key
Disk and（Or）Other users input unit.By a united display system 42, operator can observe institute from computer 36
State reproduction image and other data.Instruction that computer 36 is provided using operator and parameter come to data collecting system 32, penetrate
Lane controller 28 and pallet engine controller 30 provide control signal and information.In addition, computer 36 is by running a work
Make platform engine controller 44 to control a mobile working platform 46 for being used for being arranged to patient 22 in pallet 12.Especially,
The workbench 46 is the position of the removing patient of opening 48 by the pallet.
In one embodiment, computer 36 is used for from machine readable medium 52 including one, such as floppy disk, CD（CD-
ROM）Or Digital versatile disc（DVD）The middle device for reading instruction or data, such as disk drive, CD-ROM drive or CD-ROM device.Ying Li
The thing of solution, also in the presence of other suitable machine readable type of memory（At will fors two examples for, such as CD-rewritable with
Flash memory）, any of which is not excluded for herein.In another embodiment, computer 36, which is performed, is stored in admittedly
Part（It is not shown）In instruction.Usually, data collecting system 32, imagescope 34 and computer 36 as shown in Figure 2 are installed on
In at least one in processor be composed of the program for performing following method and steps.But following methods and be not limited to the use of CT into
As system 10, it can also be connected and use with other different types of imaging systems.In one embodiment, computer 36 is composed of realization
The program of function described herein, therefore, computer referred to herein are not limited to industry commonly called integrated circuit institute's generation
The computer of table, and should have wider scope, including computer, processor, single-chip microcomputer, miniature electronic meter, programmable patrol
Collect controller, application specific integrated circuit and other programming devices.
The computer 36 passes through program composition, can perform noise criteria difference estimation, and detailed content will be carried out hereinafter
Description.
One aspect of the present invention provides a kind of evaluation method of noise criteria difference, and it can accurately estimate noise mark
Quasi- difference., hereafter will be main because this method is particularly suitable for use in non-local mean filter noise reduction algorithm carrying out standard deviation estimation
It is described with reference to non-local mean filter noise reduction algorithm, it should be appreciated that method of the invention can also be applied to other and be related to make an uproar
The algorithm of sound standard deviation estimation, such as bilateral filtering algorithm (bilateral filtering).
In certain embodiments, the noise removing method is based on reconstructed image.Input the original image of a Noise
In, use low pass filter（low pass filter）It is filtered, a smoothed image I is obtaineds, can be by the smoothed image
As being muting image." low pass filter " referred to herein refers to low frequency signal can be allowed to pass through, and allows frequency to be higher than and cut
The only electronic filter of the signal attenuation of frequency.Input picture（That is original image In）With smoothed image IsBetween difference recognized
To be the noise pattern of input picture, N is expressed ase。
Ne=Is-In
The noise pattern NeIn not only contain input picture InIn noise, also contains in input picture InMiddle presence,
And in smoothed image IsIn the border structure that has but been eliminated.If the estimation of noise criteria difference is based on containing making an uproar for border
Sound spectrogram NeCarry out, then the accuracy estimated will be influenceed by these borders.Therefore, before noise criteria difference estimation is carried out,
Preferably by noise pattern NeMiddle border structure is removed clean.
In general, with noise pattern NeIn the pixel of non-boundary member compare, the pixel on border has higher
Absolute value of noise.One or more threshold values or a threshold function table can be set, by noise pattern NeIn each pixel noise
Absolute value is compared with the threshold value or threshold function table, to determine noise pattern NeIn which pixel be border structure.
In certain embodiments, can be to whole noise pattern NeA threshold value is estimated, by noise pattern NeIn each pixel
Absolute value of noise is compared with the threshold value, to determine noise pattern NeIn border structure.Assuming that PiRepresent noise pattern NeIn certain
The absolute value of noise of one pixel, T is noise pattern NeThreshold value, if Pi＞ nT, then it is assumed that the pixel is boundary pixel point,
Can be by it from noise pattern N before noise criteria difference estimation is carried outeIt is middle to remove.Compare formula above-mentioned, n is for controlling side
Remove the scale factor constant of intensity in boundary.
In view of the inhomogeneities of CT noise in image, its usual noise criteria difference is become as the position in image is different
Change.In most cases, the threshold value for whole noise pattern is not a fixed numerical value, and it can be with image
Position is different and produces change.Therefore, in certain embodiments, the noise pattern NeIt is divided into N number of or N*N region, region
Shape can be rectangle, square, circular, annular or object adaptive shape, a threshold value can be estimated to each region
Ti, for determining border structure contained in the region.
The threshold value T in each regioniIt is determined that afterwards, both can directly use, it may also be used for being formed one for whole noise
The threshold function table of figure.Especially, curve fitting technique can be passed through based on the threshold data in each region（curve fitting
technology）Obtain a threshold function table Tg.The threshold function table TgThe threshold value map of one correspondence noise pattern can be provided, this
Sample, noise pattern NeIn each pixel（I, j）All correspond to threshold value map TgIn a reference noise value, it is determined that border tie
The reference noise value can be used as the benchmark of multilevel iudge during structure.The reference noise value can be with corresponding threshold value Tg（I, j）It is directly proportional.
Such as, in a specific embodiment, pixel（I, j）Corresponding noise reference value is n*Tg（I, j）, wherein Tg（I, j）For
Pixel（I, j）Corresponding threshold value, n is scale factor constant.As noise pattern NeIn a certain pixel absolute value of noise it is big
When its corresponding reference noise value, such as, work as P（I, j）＞ n*Tg（I, j）When, then it is assumed that the pixel is boundary pixel point,
Can be by it from noise pattern N before noise criteria difference estimation is carried outeIt is middle to remove.
The threshold value T or the threshold value T in each region of noise pattern for whole noise patterniIt can be estimated by different methods
Meter and determination.
In one embodiment, can be based on the warp relevant with the border structure in noise pattern in threshold value T method
Test knowledge and pre cleaning is carried out to the border structure in noise pattern, such as, understood based on Heuristics, the border structure in noise pattern
It is generally located near the position of bone-tissue or air-tissue boundary.Using this kind of Heuristics, it is estimated that input
The quantity e_N of possible boundary pixel point in image.Especially, with low pass filter to input picture InIt is filtered acquisition
Muting image Is, and pass through the InAnd IsDifference obtain noise pattern NeIn, the volume that border structure passes through boundary pixel point
Product（convolution）With the kernel of filter（filter kernel）It is determined.NeThe approximate number of middle boundary pixel point
Ea_N can in the input picture quantity e_N of possible boundary pixel point and low pass filter interior nuclear operator（kernel
size）K*k is calculated by below equation and obtained：
ea_N=k*e_N
Because typically the pixel with higher absolute value of noise is boundary pixel point, once boundary pixel point is general
Quantity ea_N is determined, it is possible to which it is boundary pixel point to be inferred to ea_N maximum pixel of absolute value of noise.By that will make an uproar
The absolute value of noise of each pixel is arranged from high to low in sound spectrogram, can be with it can be easily ascertained that ea_N maximum noise is absolute
Value.After boundary pixel point is determined, by these pixels from noise pattern NeMiddle pre cleaning falls, and obtains making an uproar behind pre cleaning border
Sound spectrogram, is expressed as Nep.Also can be by making an uproar before pre cleaning and removing border in the noise pattern of different phase, the application in order to distinguish
Sound spectrogram NeReferred to as raw noise figure.
Then, can be from the noise pattern N behind the pre cleaning borderepIn to image InMiddle pixel（I, j）Neighbouring noise
Standard deviation is estimated that the formula of estimation is as follows in advance：
Sigmap(i, j)=std (Nep(i-w:i+w,j-w:j+w))。
Wherein, " std " is that, for calculating the known function of standard deviation, 2w+1 is the poor pre- estimation of the progress noise criteria
Width neighborhood used, w therein can adaptively rebuild the visual field（Field of view, FOV）.To the standard deviation map
SigmapPre- estimation can based in image each pixel carry out, sparsely partial pixel can also be taken to click through from image
Row is calculated.
Then can the standard deviation map Sigma based on the pre- estimationpObtain a threshold value T.A medium filtering can be used
Standard deviation map Sigma of the device to the pre- estimationpIt is filtered, takes SigmapIntermediate value be used as threshold value T, such threshold value meter
Calculation mode is highly stable.
T=mediannSigmap)
Obtain after the threshold value T, the boundary pixel point that can be gone out by the rule judgment being described above in noise pattern, i.e. will
Absolute value of noise Pi＞ nT pixel treats as boundary pixel point.After boundary pixel point is determined and removed, a removing can be obtained
Noise pattern N behind borderer。
When a threshold value not fixed value for whole noise pattern, incited somebody to action by the above method based on Heuristics
Boundary pixel is from noise pattern after pre cleaning, the noise pattern N behind pre cleaning borderepIt is divided into multiple regions, passes through preceding method
Standard deviation to each region is estimated that obtain the region estimates standard deviation map Sigma in advancep_i, and take Sigmap_iIn
It is worth the threshold value T as the regioni。
Ti=median (Sigmap_i)
After the threshold data in all regions is obtained, curve fitting technique acquisition one can be passed through based on the threshold data
Threshold function table Tg。
Obtain the threshold function table TgAfterwards, the boundary pixel point that can be gone out by the rule judgment being described above in noise pattern,
That is, by absolute value of noise P（I, j）＞ n*Tg（I, j）Pixel treat as boundary pixel point.Boundary pixel point is determined and removed
Afterwards, a noise pattern N removed behind border can be obtaineder。
As it was previously stated, the relatively large pixel of absolute value of noise typically belongs to boundary pixel point.Further or can
Alternatively, the relatively large pixel of noise gradient typically falls within boundary pixel point.Therefore, as by by each pixel
Absolute value of noise and reference noise value be compared to determine supplement or the replacement of the method on border, border also can be by will be every
The noise gradient of one pixel is compared to determine with reference gradient value.Wherein described reference gradient value can be with corresponding threshold value
It is directly proportional.The gradient in x-axis and y-axis direction in one CT image can be calculated by below equation：
[FX, FY]=gradient (In).
In certain embodiments, the absolute value of noise of each pixel can be compared with a reference noise value, will
The noise gradient of each pixel and a reference gradient value are compared to determine border.Such as, in a specific implementation
In example, boundary pixel point can be determined by below equation：
Wherein, " abs " is the function that takes absolute value, and " n " and " m " is constant, and two constants can be with identical, can also not
Together.The pixel for meeting Edge_In (i, j)=1 is defined as boundary pixel point, removed from noise pattern, just can be by noise pattern
Border structure remove clean, obtain the noise pattern N removed behind borderer。
Based on the noise pattern N behind the removing borderer, an accurate noise criteria can be obtained and poorly scheme Sigma.Especially
Ground, input picture InIn pixel（I, j）Neighbouring noise criteria difference can be from the noise pattern N behind the removing bordererIn lead to
Cross below equation and calculate acquisition：
Sigma (i, j)=std (Ner(i-w:i+w,j-w:J+w)) (2),
Wherein, " std " is that, for calculating the known function of standard deviation, 2w+1 calculates used to carry out the noise criteria difference
Width neighborhood, w therein can adaptively rebuild the visual field.Similarly, the calculating to the standard deviation map Sigma can be based on
Each pixel in image is carried out, and can also sparsely take partial pixel point to be calculated from image.
In certain embodiments, the noise criteria obtained by methods described poorly schemes Sigma can be logical with reference to threshold function table
Below equation is crossed further to be modified：
Wherein, Sigma (i, j) is pixel（I, j）Neighbouring noise criteria is poor, Tg（I, j）For pixel（I, j）Near
Threshold value, L and H are respectively a relatively small constant and a relatively large constant.
Another aspect provides the method for another estimated noise standard deviation.
Input after original image, processing is filtered to the original image and obtains a smoothed image, by as previously described
Method, the original image of desirable input and the difference of smoothed image are used as noise pattern Ne.In the noise pattern NeIn have homogeneity range
Domain, the noise criteria difference estimated from homogeneous area is accurate.In certain embodiments, only to the picture in uniform tissue regions
Vegetarian refreshments carries out the estimation of noise criteria difference, without to skeletal structure and low attenuating structure（low attenuation
structures）In pixel carry out noise criteria difference estimation because to pixel in skeletal structure and low attenuating structure
The noise criteria difference estimation potentially unstable of point.In addition, in order to obtain in the poor estimation result of accurate noise criteria, air
Pixel is also excluded from outside and without the estimation of noise criteria difference.
Because the noise criteria difference in air is poor much smaller than the noise criteria of scanned object, therefore air section is excluded
It is easy-to-use.Can be by by the noise average of each pixel neighborhood of a point in original image and an air threshold and one
Individual bone threshold value is compared, to judge whether each pixel is located in tissue regions.If a certain pixel is in the input image
The noise average of neighborhood be more than the air threshold and less than the bone threshold value, then it is assumed that the pixel is located at tissue area
In domain, rather than in bone structure, low attenuating structure or air.
In noise pattern NeIn, compared with the borderline pixel of skeletal structure and low attenuating structure, if pixel（I, j）
In uniform region, the noise average of its neighborhood is relatively low.Therefore, can be by by each pixel in noise pattern
The threshold value of noise average and a noise average of neighborhood be compared, to judge the pixel whether in uniform area
In domain.
In one particularly example, the noise average difference of neighborhood that can be by each pixel in original image
It is compared with an air threshold and a bone threshold value, and the noise average of its neighborhood in noise pattern is made an uproar with one
The threshold value of sound average value is compared, to judge whether each pixel is located in uniform tissue regions.If a pixel exists
The noise average of neighborhood in input picture is more than the air threshold and less than the bone threshold value, and it is in noise pattern
Neighborhood noise average be less than the noise average threshold value, then it is assumed that the pixel be located at uniform tissue regions
It is interior.In a specific embodiment, as shown in following equation, only the pixel in uniform tissue regions is made an uproar
The estimation of sound standard deviation, i.e. only when the pixel that a pixel belongs in tissue, and its be located at homogeneous area in when,
The estimation of noise criteria difference is carried out to it, its noise criteria difference is otherwise set to 0.
Wherein, tnFor the threshold value of noise criteria difference, taFor air threshold, tbFor bone threshold value.
Based on described in estimating in uniform tissue regions（That is Sigma in aforesaid equationeIt is not equal to 0 region）
The noise criteria difference and the coordinate of these pixels of interior pixel（I, j）, one can be obtained by curve fitting technique
The function of individual noise criteria difference（Map）.In such an embodiment, SigmaeIn non-zero noise criteria difference be in close proximity to
Real noise criteria is poor.Due to noise criteria, poorly figure should be relatively smooth, can be integrated by curve fitting technique
SigmaeIn non-zero noise criteria difference carry out fit surface, and obtain the noise mark of the function representation by a noise criteria difference
Standard is poorly schemed.Such as, by polynomial curve fitting method, a Gaussian function can be obtained from above-mentioned data（general
Gaussian function）.Certainly, the function of the noise criteria difference is not limited to Gaussian function, and it can be an integration
The smooth surface of noise criteria difference data is obtained in the uniform formation region.
To obtain noise pattern from highly uniform input picture, due to homogeneous area therein and Nonuniform Domain Simulation of Reservoir it
Between not obvious boundary, thus be difficult to determine where be homogeneous area.In this case, the chi of neighborhood can be reduced
It is very little, to allow the determination of homogeneous area to become easier to.Such as, can be by above-mentioned SigmaeEquation in parameter w be set to 1,
So the size of neighborhood is 3*3, and size is sufficiently small, can conveniently determine homogeneous area.
In certain embodiments, after the function that noise criteria difference is obtained by the above method, directly the function can be used for
Noise reduction algorithm.
In some other embodiments, the function of noise criteria difference can be treated as threshold function table TgFor foregoing side
Formula（1）In determine the border structure in raw noise figure, so that border structure be removed from raw noise figure, then pass through
Aforesaid equation（2）Calculating noise criteria based on the noise pattern for removing border is poor, obtains noise criteria and poorly schemes.Institute
Obtain noise criteria poorly to scheme poorly scheme output directly as final noise criteria, can also further pass through aforementioned equation
Formula（3）Exported again after being modified., will compared with the foregoing embodiment directly by the function of noise criteria difference for noise reduction algorithm
The function of noise criteria difference treats as threshold function table TgFor determining and removing the border structure in raw noise figure, then removing
The mode of estimated noise standard deviation can further increase the accuracy of noise criteria difference in the noise pattern on border.
Noise reduction algorithm can be used it for after obtaining noise criteria difference by the above method.In certain embodiments, obtained
Noise criteria difference can be applied in non-local mean filter noise reduction algorithm.
As shown in figure 3, in a kind of poor evaluation method of noise criteria, the input of a Noise is received in step s 11
Image, is filtered to obtain a smoothed image to the input picture, the input is taken in step s 13 in step s 12
The difference of image and the smoothed image is raw noise figure, and the border in the raw noise figure is carried out in step S14
Pre cleaning, estimates threshold value T or threshold function table T in step S15 from the pre cleaning in the noise pattern on borderg, in step S16
It is middle by the threshold value T or threshold function table TgFor determining the border in raw noise figure, by identified border in step S17
From raw noise figure remove, in step S18 from the noise pattern for removing border estimated noise standard deviation.
In certain embodiments, in the step S14, input picture can be estimated by the Heuristics about border
In boundary pixel point quantity e_N.As it was previously stated, with low pass filter to input picture InIt is filtered acquisition noiseless
Image Is, and pass through the InAnd IsDifference obtain noise pattern NtIn, the convolution that border structure passes through boundary pixel point
（convolution）With the kernel of filter（filter kernel）It is determined.NeThe approximate number of middle boundary pixel point
Ea_N can in the input picture quantity e_N of possible boundary pixel point and low pass filter interior nuclear operator（kernel
size）K*k is calculated by below equation and obtained：
ea_N=k*e_N。
Due to noise pattern NeMiddle boundary pixel point generally has relatively bigger absolute value of noise than non-border pixel point, because
This, ea_N maximum pixel of absolute value of noise, i.e. correspondingly the pixel of ea_N maximum absolute value of noise is considered as
Boundary pixel point, so as to be eliminated.Such as, once the quantity of the boundary pixel point in noise pattern, such as 100, it is determined that after, based on one
As for have relatively large absolute value of noise pixel be boundary pixel point premise, can be inferred that correspondence maximum
100 pixels of 100 absolute value of noise are boundary pixel point, should be eliminated.For containing ea_N boundary pixel point
Noise pattern NeFor, the pixel of ea_N maximum absolute value of noise of correspondence is boundary pixel point.By by noise pattern
Pixel is arranged from big to small according to the size of its absolute value of noise, and removes ea_N absolute value of noise for wherein corresponding to maximum
Ea_N pixel, just reached the effect on the border in pre cleaning noise pattern.
In certain embodiments, in step S15, threshold value T is estimated in the noise pattern on border from the pre cleaning,
The intermediate value for the noise criteria difference estimated for pre cleaning in the noise pattern on border.With the threshold value T come really in the step S16
, can be by the noise figure of each pixel and a reference noise value being directly proportional to the threshold value T when determining the border in noise pattern
Compare, the pixel that absolute value of noise is more than the reference noise value is defined as boundary pixel point.
In certain embodiments, in step S15, easily noise pattern is divided into multiple regions by the pre cleaning, is
Each region estimates that noise criteria is poor in advance, take pre- estimation noise criteria difference intermediate value be the region threshold value Ti.At some
In embodiment, with the threshold function table T in the step S16gIt is to determine during the border in noise pattern, absolute value of noise is big
In a reference noise value being directly proportional with corresponding threshold value pixel and at least one noise gradient be more than one with it is right
The pixel for the reference gradient that threshold value is directly proportional is answered to be defined as boundary pixel point.
As shown in figure 4, in a kind of poor evaluation method of noise criteria, the input of a Noise is received in the step s 21
Image, is filtered to obtain a smoothed image to the input picture in step S22, the input is taken in step S23
The difference of image and the smoothed image is raw noise figure, in step s 24 to area of the raw noise Tu Zhong uniform formations
The noise criteria difference of pixel in domain is estimated, in step s 25 based on the area of uniform formation obtained in the step S24
The noise criteria difference and the coordinate of these pixels of pixel in domain, a noise mark is obtained by curve fitting technique
The function of quasi- difference.
Fig. 5 shows a kind of poor evaluation method of the noise criteria similar with Fig. 4, and difference is that it further exists
The function of the noise criteria obtained difference is used as threshold function table in step S26, to determine the border in raw noise figure, in step
In rapid S27 identified border is removed from raw noise figure, the noise pattern on border is removed from described in step S28
Estimated noise standard deviation.It is similar to the step 16 in Fig. 3, it is in certain embodiments, in step S26, absolute value of noise is big
In a reference noise value being directly proportional with corresponding threshold value pixel and at least one noise gradient be more than one with it is right
The pixel for the reference gradient that threshold value is directly proportional is answered to be defined as boundary pixel point.
Method described in the embodiment of the present invention can improve the accuracy of noise criteria difference estimation.Fig. 6 shows one
Comparative example, compares the CT images of Noise have been carried out after non-local mean filter noise reduction with different parameters wherein
Experimental result.Fig. 6（a）Show that the local noise criteria that the method described in an a kind of the present embodiment of use is obtained is poor（local
noise standard deviation）The image obtained after non-local mean filter noise reduction is carried out to the image of Noise.Fig. 6
（b）Show that one carries out the figure obtained after non-local mean filter noise reduction with 90HU noise criteria difference to the image of Noise
Picture.Because 90HU noise criteria difference is less than the real noise standard deviation in the centre of image, centre is caused to filter
Gently, therefore, itself and Fig. 6（a）Shown in image compare, contain more noises at the position.Fig. 6（c）With 6（d）Show respectively
The image obtained after non-local mean filter noise reduction carried out to the image of Noise with 110HU and 120HU noise criteria difference.
Because 110HU or 120HU noise criteria difference is higher than real noise standard deviation, cause image excess smoothness, therefore, itself and Fig. 6
（a）Compare, border is more obscured.This example shows that the poor evaluation method of noise criteria in the present invention can be provided and more accurately made an uproar
Sound standard deviation estimation result, so as to realize more preferable noise reduction.
The present invention can be summarized with others without prejudice to the concrete form of the spirit or essential characteristics of the present invention.Therefore, nothing
By from the point of view of which point, the embodiment above of the invention can only all be considered the description of the invention and can not limit this hair
Bright, the scope of the present invention is to be defined by tbe claims, rather than is defined by above-mentioned, therefore, will with the right of the present invention
Any change in book suitable implication and scope is asked, is all considered as being included within the scope of the claims.
Claims (22)
1. a kind of evaluation method of noise criteria difference, it includes：
Receive the input picture of Noise；
The input picture is filtered to obtain smoothed image；
The difference for taking the input picture and the smoothed image is raw noise figure；
Border in raw noise figure described in pre cleaning, to obtain the noise pattern on pre cleaning border；
Threshold value T or threshold function table T is estimated in the noise pattern on border from the pre cleaningg；
By the threshold value T or threshold function table TgFor determining the border in raw noise figure；
Identified border is removed from raw noise figure, to obtain the noise pattern for removing border；And
The estimated noise standard deviation from the noise pattern for removing border.
2. the step of border in raw noise figure described in a kind of the method for claim 1, wherein pre cleaning, includes：
The quantity e_N of possible boundary pixel point in the input picture is estimated based on the Heuristics about border；
Quantity e_N based on possible boundary pixel point in the input picture calculates boundary pixel point in the raw noise figure
Approximate number ea_N；And
Ea_N pixel of ea_N maximum absolute value of noise of correspondence is removed from raw noise figure.
3. a kind of the method for claim 1, wherein estimation threshold value T in the noise pattern on border from the pre cleaning
Step includes：
Pre- estimation noise criteria is poor in the noise pattern on border from the pre cleaning；And
Intermediate value using the noise criteria difference of the pre- estimation is threshold value T.
4. a kind of method as claimed in claim 3, wherein, the threshold value T is used to determine the border in raw noise figure
Step includes：
By the absolute value of noise of each pixel in raw noise figure and a reference noise value being directly proportional to the threshold value T
Compare；And
The pixel that absolute value of noise is more than the reference noise value is defined as boundary pixel point.
5. the method for claim 1, wherein a kind of estimate threshold function table from the pre cleaning in the noise pattern on border
TgThe step of include：
The noise pattern on border is divided into multiple regions by the pre cleaning；
Noise criteria is poor to be estimated in advance to each region, and using the poor intermediate value of the noise criteria estimated in advance as the threshold value T in the regioni；With
And
Threshold data based on the multiple region sets up a threshold function table Tg。
6. a kind of method as claimed in claim 5, wherein, by the threshold function table TgFor determining the side in raw noise figure
The step of boundary, includes：
The absolute value of noise of each pixel in raw noise figure is directly proportional with a threshold value corresponding with the pixel
Reference noise value compares；And
The pixel that absolute value of noise is more than its corresponding reference noise value is defined as boundary pixel point.
7. a kind of method as claimed in claim 5, wherein, by the threshold function table TgFor determining the side in raw noise figure
The step of boundary, includes：
The absolute value of noise of each pixel in raw noise figure threshold value corresponding with one and the pixel is directly proportional
Reference noise value compares；
By the noise gradient and one and the pixel pair of each pixel in raw noise figure in one or more directions
The reference gradient value that the threshold value answered is directly proportional compares；And
The noise gradient that absolute value of noise is more than its corresponding reference noise value or at least one direction is more than corresponding reference
The pixel of gradient is defined as boundary pixel point.
8. a kind of estimating system of noise criteria difference, including：
For the device for the input picture for receiving Noise；
For being filtered to the input picture to obtain the device of smoothed image；
Difference for taking the input picture and the smoothed image is the device of raw noise figure；
For the border in raw noise figure described in pre cleaning with the device for the noise pattern for obtaining pre cleaning border；
For estimating threshold value T or threshold function table T in the noise pattern on border from the pre cleaninggDevice；
By the threshold value T or threshold function table TgDevice for determining the border in raw noise figure；
For identified border to be removed from raw noise figure to obtain the device for the noise pattern for removing border；And
Device for the estimated noise standard deviation from the noise pattern for removing border.
9. system as claimed in claim 8, wherein, the device for the border in pre cleaning raw noise figure is further wrapped
Include：
Quantity e_N for estimating possible boundary pixel point in the input picture based on the Heuristics about border
Device；
Border picture in the raw noise figure is calculated for the quantity e_N based on possible boundary pixel point in the input picture
The approximate number ea_N of vegetarian refreshments device；And
For the device for removing ea_N pixel of ea_N maximum absolute value of noise of correspondence from raw noise figure.
10. system as claimed in claim 8, wherein, the device for estimation threshold value T in the noise pattern on border from pre cleaning
Further comprise：
Device for pre- estimation noise criteria difference in the noise pattern on border from the pre cleaning；And
For the device using the intermediate value of the noise criteria difference of the pre- estimation as threshold value T.
11. system as claimed in claim 8, wherein, for estimation threshold function table T in the noise pattern on border from pre cleaningg's
Device further comprises：
Noise pattern for the border by the pre cleaning is divided into the device in multiple regions；
For estimating that noise criteria is poor in advance to each region, and using the poor intermediate value of the noise criteria estimated in advance as the threshold value in the region
TiDevice；And
A threshold function table T is set up for the threshold data based on the multiple regiongDevice.
12. a kind of medical imaging apparatus, it includes：
Radiographic source and ray detector on rotary pallet are installed on, wherein the ray detector is used for receiving being penetrated by described
Line source passes through the ray of scanned object after sending；
Data collecting system, sets and is used for receiving the data from the ray detector when scanning object, and to computer
Data for projection collection is provided；
The display device for the image that the computer is obtained with the data for projection collection is shown by responding the computer；With
And
Processor, setting is used for：
Receive the input picture of Noise；
The input picture is filtered to obtain smoothed image；
The difference for taking the input picture and the smoothed image is raw noise figure；
Border in raw noise figure described in pre cleaning, to obtain the noise pattern on pre cleaning border；
Threshold value T or threshold function table T is estimated in the noise pattern on border from the pre cleaningg；
By the threshold value T or threshold function table TgFor determining the border in raw noise figure；
Identified border is removed from raw noise figure, to obtain the noise pattern for removing border；And
The estimated noise standard deviation from the noise pattern for removing border.
13. a kind of evaluation method of noise criteria difference, it includes：
Receive the input picture of Noise；
The input picture is filtered to obtain smoothed image；
The difference for taking the input picture and the smoothed image is raw noise figure；
Noise criteria difference to the pixel in the raw noise Tu Zhong uniform formations region is estimated；And
Noise criteria difference and the coordinate of these pixels based on the pixel in the uniform formation region obtained, pass through
Curve fitting technique obtains the function of noise criteria difference.
14. method as claimed in claim 13, wherein the uniform formation region does not include air and bony areas.
15. method as claimed in claim 13, wherein, the uniform formation region is through the following steps that determine：
By the noise average of each pixel neighborhood of a point in input picture respectively with an air threshold and a bone threshold
Value is compared；
The threshold value of the noise average and a noise average of each pixel neighborhood of a point in raw noise figure is compared
Compared with；And
If the noise average of the neighborhood of a pixel in the input image is more than the air threshold and less than the bone
Threshold value, and threshold value of the noise average less than the noise average of its neighborhood in raw noise figure, then judge the picture
Vegetarian refreshments is located in uniform tissue regions.
16. method as claimed in claim 13, wherein, the function of the noise criteria difference has been fitted in uniform formation region
The coordinate of each pixel and its Gaussian function of noise criteria difference.
17. method as claimed in claim 13, it further comprises：
The border in raw noise figure is determined using the function of noise criteria difference as threshold function table；
Identified border is removed from raw noise figure, to obtain the noise pattern for removing border；And
The estimated noise standard deviation from the noise pattern for removing border.
18. method as claimed in claim 17, wherein, the function using noise criteria difference is original to determine as threshold function table
The step of border in noise pattern, includes：
The absolute value of noise of each pixel in raw noise figure threshold value corresponding with one and the pixel is directly proportional
Reference noise value compares；
By the noise gradient and one and the pixel pair of each pixel in raw noise figure in one or more directions
The reference gradient value that the threshold value answered is directly proportional compares；And
The noise gradient that absolute value of noise is more than its corresponding reference noise value or at least one direction is more than corresponding reference
The pixel of gradient is defined as boundary pixel point.
19. a kind of estimating system of noise criteria difference, including：
For the device for the input picture for receiving Noise；
For being filtered to the input picture to obtain the device of smoothed image；
Difference for taking the input picture and the smoothed image is the device of raw noise figure；
The device estimated for the noise criteria difference to the pixel in the raw noise Tu Zhong uniform formations region；With
And
For noise criteria difference and the coordinate of these pixels based on the pixel in the uniform formation region obtained,
The device of the function of noise criteria difference is obtained by curve fitting technique.
20. system as claimed in claim 19, wherein, further comprise：
For determining the device on the border in raw noise figure using the function of noise criteria difference as threshold function table；
For identified border to be removed from raw noise figure to obtain the device for the noise pattern for removing border；And
Device for the estimated noise standard deviation from the noise pattern for removing border.
21. a kind of medical imaging apparatus, it includes：
Radiographic source and ray detector on rotary pallet are installed on, wherein the ray detector is used for receiving being penetrated by described
Line source passes through the ray of scanned object after sending；
Data collecting system, sets and is used for receiving the data from the ray detector when scanning object, and to computer
Data for projection collection is provided；
The display device for the image that the computer is obtained with the data for projection collection is shown by responding the computer；With
And
Processor, setting is used for：
Receive the input picture of Noise；
The input picture is filtered to obtain smoothed image；
The difference for taking the input picture and the smoothed image is raw noise figure；
Noise criteria difference to the pixel in the raw noise Tu Zhong uniform formations region is estimated；And
Noise criteria difference and the coordinate of these pixels based on the pixel in the uniform formation region obtained, pass through
Curve fitting technique obtains the function of noise criteria difference.
22. a kind of medical imaging apparatus as claimed in claim 21, processor therein, which is further set, to be used for：
The border in raw noise figure is determined using the function of noise criteria difference as threshold function table；
Identified border is removed from raw noise figure, to obtain the noise pattern for removing border；And
The estimated noise standard deviation from the noise pattern for removing border.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310227562.5A CN104240184B (en) | 2013-06-08 | 2013-06-08 | The evaluation method and system of noise criteria difference |
| PCT/US2014/041015 WO2014197658A1 (en) | 2013-06-08 | 2014-06-05 | Method and system for noise standard deviation estimation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310227562.5A CN104240184B (en) | 2013-06-08 | 2013-06-08 | The evaluation method and system of noise criteria difference |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104240184A CN104240184A (en) | 2014-12-24 |
| CN104240184B true CN104240184B (en) | 2017-09-26 |
Family
ID=52008579
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310227562.5A Active CN104240184B (en) | 2013-06-08 | 2013-06-08 | The evaluation method and system of noise criteria difference |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN104240184B (en) |
| WO (1) | WO2014197658A1 (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106994021B (en) * | 2016-01-22 | 2022-10-14 | 通用电气公司 | Method and device for calculating noise on CT image |
| US10013780B2 (en) * | 2016-02-29 | 2018-07-03 | General Electric Company | Systems and methods for artifact removal for computed tomography imaging |
| WO2019134871A1 (en) | 2018-01-02 | 2019-07-11 | Koninklijke Philips N.V. | Learning-based voxel evolution for regularized reconstruction |
| CN113129235A (en) * | 2021-04-22 | 2021-07-16 | 深圳市深图医学影像设备有限公司 | Medical image noise suppression algorithm |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5050227A (en) * | 1988-12-15 | 1991-09-17 | Dainippon Screen Mfg. Co., Ltd. | Method of and apparatus for image smoothing along a tangential direction of a contour |
| CN101933042A (en) * | 2008-01-25 | 2010-12-29 | 模拟逻辑有限公司 | Edge detection |
| CN102177705A (en) * | 2008-09-04 | 2011-09-07 | 晶像股份有限公司 | System, method, and apparatus for smoothing of edges in images to remove irregularities |
| CN102667852A (en) * | 2009-11-25 | 2012-09-12 | 皇家飞利浦电子股份有限公司 | Enhanced image data/dose reduction |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7643699B2 (en) * | 2005-07-05 | 2010-01-05 | Hewlett-Packard Development Company, L.P. | Image processing based on local noise statistics |
| JP2011512999A (en) * | 2008-03-04 | 2011-04-28 | トモセラピー・インコーポレーテッド | Improved image segmentation method and system |
| EA017302B1 (en) * | 2011-10-07 | 2012-11-30 | Закрытое Акционерное Общество "Импульс" | Method of noise reduction of digital x-ray image series |
-
2013
- 2013-06-08 CN CN201310227562.5A patent/CN104240184B/en active Active
-
2014
- 2014-06-05 WO PCT/US2014/041015 patent/WO2014197658A1/en active Application Filing
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5050227A (en) * | 1988-12-15 | 1991-09-17 | Dainippon Screen Mfg. Co., Ltd. | Method of and apparatus for image smoothing along a tangential direction of a contour |
| CN101933042A (en) * | 2008-01-25 | 2010-12-29 | 模拟逻辑有限公司 | Edge detection |
| CN102177705A (en) * | 2008-09-04 | 2011-09-07 | 晶像股份有限公司 | System, method, and apparatus for smoothing of edges in images to remove irregularities |
| CN102667852A (en) * | 2009-11-25 | 2012-09-12 | 皇家飞利浦电子股份有限公司 | Enhanced image data/dose reduction |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104240184A (en) | 2014-12-24 |
| WO2014197658A1 (en) | 2014-12-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP4820582B2 (en) | Method to reduce helical windmill artifact with recovery noise for helical multi-slice CT | |
| Gravel et al. | A method for modeling noise in medical images | |
| US9036771B2 (en) | System and method for denoising medical images adaptive to local noise | |
| US8294717B2 (en) | Advanced clustering method for material separation in dual energy CT | |
| US6819735B2 (en) | Histogram-based image filtering in computed tomography | |
| US8532744B2 (en) | Method and system for design of spectral filter to classify tissue and material from multi-energy images | |
| US10157483B2 (en) | Backprojection approach for photoacoustic image reconstruction | |
| CN102622743B (en) | For comparing the method and apparatus of 3D and 2D view data | |
| JP2005296605A (en) | Method of segmenting a radiographic image into diagnostically relevant and diagnostically irrelevant regions | |
| CN104240184B (en) | The evaluation method and system of noise criteria difference | |
| CN108888284A (en) | Image adjusting method, device and equipment, storage medium | |
| JPH04246982A (en) | Method of correcting measurement of optical density formed onto x-ray film | |
| CN111402150B (en) | CT image metal artifact removal method and device | |
| US20080144904A1 (en) | Apparatus and Method for the Processing of Sectional Images | |
| US20080118128A1 (en) | Methods and systems for enhanced accuracy image noise addition | |
| US9763636B2 (en) | Method and system for spine position detection | |
| CN102419864B (en) | Method and device for extracting skeletons of brain CT (computerized tomography) image | |
| WO2015040547A1 (en) | Method and system for spine position detection | |
| CN106108932B (en) | Full-automatic kidney region of interest extraction element and method | |
| Liao et al. | Noise Estimation for Single‐Slice Sinogram of Low‐Dose X‐Ray Computed Tomography Using Homogenous Patch | |
| US8989462B2 (en) | Systems, methods and computer readable storage mediums storing instructions for applying multiscale bilateral filtering to magnetic resonance (RI) images | |
| US11113810B2 (en) | X-ray CT scanner, image generation method, and image generation program | |
| Hsieh | A practical cone beam artifact correction algorithm | |
| JP6878590B2 (en) | Reproduction of conventional computed tomography images from spectral computed tomography data | |
| JP2007248121A (en) | Method, program, and device for extracting contour of tomographic image |
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 |