CN108537723A - The three dimensional non-linear method for registering and system of magnanimity brain image data collection - Google Patents
The three dimensional non-linear method for registering and system of magnanimity brain image data collection Download PDFInfo
- Publication number
- CN108537723A CN108537723A CN201810307540.2A CN201810307540A CN108537723A CN 108537723 A CN108537723 A CN 108537723A CN 201810307540 A CN201810307540 A CN 201810307540A CN 108537723 A CN108537723 A CN 108537723A
- Authority
- CN
- China
- Prior art keywords
- registration
- resolution
- linear
- image data
- low
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000013480 data collection Methods 0.000 title claims abstract description 28
- 210000004556 brain Anatomy 0.000 title claims abstract description 24
- 239000011159 matrix material Substances 0.000 claims abstract description 85
- 230000009466 transformation Effects 0.000 claims abstract description 66
- 238000006073 displacement reaction Methods 0.000 claims abstract description 26
- 230000000903 blocking effect Effects 0.000 claims abstract description 13
- 238000013507 mapping Methods 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 3
- FNMKZDDKPDBYJM-UHFFFAOYSA-N 3-(1,3-benzodioxol-5-yl)-7-(3-methylbut-2-enoxy)chromen-4-one Chemical compound C1=C2OCOC2=CC(C2=COC=3C(C2=O)=CC=C(C=3)OCC=C(C)C)=C1 FNMKZDDKPDBYJM-UHFFFAOYSA-N 0.000 description 2
- 241001269238 Data Species 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 1
- 241000208340 Araliaceae Species 0.000 description 1
- 241000257303 Hymenoptera Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/14—Transformations for image registration, e.g. adjusting or mapping for alignment of images
- G06T3/147—Transformations for image registration, e.g. adjusting or mapping for alignment of images using affine transformations
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of three dimensional non-linear method for registering of magnanimity brain image data collection and system, the three dimensional non-linear method for registering of the magnanimity brain image data collection includes the following steps:S0. down-sampled step;S1. low resolution image step of registration, including:S11. linear step of registration;S12. non-linear registration step;S2. full resolution pricture Fast transforms step, including:S21. the transformation matrix of high-resolution registration calculates step;S22. spatial dimension calculates step after high-resolution registration;S23. blocking step;S24. high-resolution non linear transformation step.The present invention utilizes low resolution image Registration of Measuring Data information, is converted to magnanimity high-resolution three-dimensional image data set, realizes the non-linear registration of magnanimity high-resolution three-dimensional image data set.The present invention be suitable for based on transformation matrix institute it is linear registration and all non-linear registrations based on displacement field, can quickly to TB level data collection progress non-linear registration.
Description
Technical field
The present invention relates to image processing techniques more particularly to a kind of three dimensional non-linear registration sides of magnanimity brain image data collection
Method.
Background technology
An important research direction of the image registration techniques as image processing field, be widely used today for remote sensing,
The fields such as self-navigation, pattern-recognition, computer vision, biomedicine.With the arriving in big data epoch, magnanimity 3-D view
Data set constantly increases, and tends to reach TB grades.How to carry out non-linear registration to these magnanimity 3-D data sets is one non-
Often big challenge.
In Chinese invention patent specification CN104361590, it is proposed that a kind of high-resolution of the point self-adapted distribution of control
Rate remote sensing image matching method, self adaptive control point extracting method of this method based on partition strategy, realizes multiple dimensioned J-
Control point extraction in image images, and then multiple dimensioned progress is carried out to control point using normalized mutual information measure NMI
Match.But the data handled by this method are two-dimensional remotely-sensed data, are suitable for specific application field.Moreover, this method institute
The high-resolution of description is also only spatial resolution 10m, and size is 512 × 512 pixels, and size of data is only MB ranks, can not
Non-linear registration is carried out to TB grades of 3 d image datas.
Widely used tool also achieves the non-linear registration of 3-D data set in image registration research, such as
Insight Segmentation and Registration Toolkit (ITK) and Advanced Normalization
Tools(ANTs).Although the method that these tools both provide abundant three dimensional non-linear registration, in registration process it
3-D view is considered as one, disposable import in computer carries out operation, this drastically increases the consumption to memory.For
The 3 d image data of tens MB, some tools can even consume the memory of hundreds of GB, more be difficult to the 3-D view to TB grades
Data carry out non-linear registration.
In summary patented technology and tool, there are many methods for 3-D view non-linear registration at present, can also solve each
The problem of kind field.But as the arriving in big data epoch, magnanimity three-dimensional image data set constantly increase, there are no a kind of energy
Enough handle the non-linear registration method of TB grades of 3 d image datas.
Invention content
It is an object of the present invention to solve above-mentioned technical problem, it is non-thread to provide a kind of three-dimensional of magnanimity brain image data collection
Property method for registering and system, Fast transforms are carried out to magnanimity high-resolution three-dimensional image data set, realize magnanimity high-resolution graphics
As the non-linear registration of data set.
The purpose of the present invention is realized by following technical solution.
A kind of three dimensional non-linear method for registering of magnanimity brain image data collection, includes the following steps:
S0. down-sampled step carries out drop to default high-resolution reference image data and high-resolution image data subject to registration and adopts
Sample obtains low resolution reference image data and low resolution image data subject to registration;
S1. low resolution image step of registration, including:
S11. linear step of registration linearly matches low resolution reference image data and low resolution image data subject to registration
Standard obtains the low transformation matrix differentiated and be linearly registrated, and calculates the low inverse matrix for differentiating registration of its inverse matrix and low resolution line
Property registration after result;
S12. non-linear registration step carries out result after the linear registration of low resolution and low resolution reference image data non-thread
Property registration, obtain it is low differentiate non-linear registration displacement field and low resolution non-linear registration after result;
S2. full resolution pricture Fast transforms step, including:
S21. the transformation matrix of high-resolution registration calculates step, and height is calculated based on the low transformation matrix being linearly registrated of differentiating
Differentiate the transformation matrix of registration and the inverse matrix of its inverse matrix high-resolution registration;
S22. spatial dimension calculates step after high-resolution registration, utilizes the inverse matrix of high-resolution registration obtained by step S21, meter
Calculate spatial dimension after high-resolution is registrated;
S23. blocking step carries out piecemeal to spatial dimension after high-resolution registration obtained by step S22;
S24. high-resolution non linear transformation step carries out parallel computation to each piecemeal obtained by step S23, calculates each
A pixel finally obtains full resolution pricture nonlinear transformation result in the corresponding position of former high-resolution image data subject to registration.
Further, transformed mappings are described as transformation matrix by the linear registration described in the step S11.
Further, transformed mappings, are described as the position of pixel-shift vector by the non-linear registration described in the step S12
Move field.
Further, the piecemeal described in the step S23 carries out piecemeal, picture in parallel computation piecemeal based on computing resource
The transformation of vegetarian refreshments, the computing resource maximum can support n thread parallel, and piecemeal is carried out with length n.
Further, the high-resolution nonlinear transformation described in the step S24, including following sub-step:
S241. the range of former high-resolution image data subject to registration before piecemeal correspondent transform is estimated;
S242. each pixel in piecemeal is traversed, based on the low displacement field for differentiating non-linear registration, interpolation calculation goes out the picture
The corresponding offset vector of vegetarian refreshments calculates the pixel in former high-resolution based on the inverse matrix that offset vector and high-resolution are registrated
The corresponding position of image data subject to registration, using its gray value as the pixel gray value.
Present invention further teaches a kind of three dimensional non-linear registration arrangements of magnanimity brain image data collection, comprise the following modules:
Down-sampled module, it is down-sampled to default high-resolution reference image data and high-resolution image data progress subject to registration,
Obtain low resolution reference image data and low resolution image data subject to registration;
Low resolution image registration module, including:
Linear registration unit is linearly registrated low resolution reference image data and low resolution image data subject to registration,
The low transformation matrix differentiated and be linearly registrated is obtained, and calculates the low inverse matrix for differentiating registration of its inverse matrix and low resolution linearly
Result after registration;
Non-linear registration unit carries out non-linear match to result after the linear registration of low resolution and low resolution reference image data
Standard obtains result after the low displacement field for differentiating non-linear registration and low resolution non-linear registration;
Full resolution pricture Fast transforms module, including:
The transformation matrix computing unit of high-resolution registration calculates high-resolution based on the low transformation matrix being linearly registrated of differentiating
The inverse matrix of the transformation matrix of registration and its inverse matrix high-resolution registration;
Spatial dimension computing unit after high-resolution registration, the transformation matrix computing unit being registrated using high-resolution are secured satisfactory grades
The inverse matrix for distinguishing registration calculates spatial dimension after high-resolution registration;
Blocking unit, spatial dimension is divided after high-resolution registration obtained by spatial dimension computing unit after being registrated to high-resolution
Block;
High-resolution non-linear conversion unit carries out parallel computation to each piecemeal obtained by blocking unit, calculates each
Pixel finally obtains full resolution pricture nonlinear transformation result in the corresponding position of former high-resolution image data subject to registration.
Further, transformed mappings are described as transformation matrix by the linear registration described in the linear registration unit.
Further, transformed mappings are described as pixel-shift by the non-linear registration described in the non-linear registration unit
The displacement field of vector.
Further, the piecemeal described in the blocking unit carries out piecemeal, picture in parallel computation piecemeal based on computing resource
The transformation of vegetarian refreshments, the computing resource maximum can support n thread parallel, and piecemeal is carried out with length n.
Further, the high-resolution nonlinear transformation described in the high-resolution non-linear conversion unit, including with lower unit:
Evaluation unit estimates the range of former high-resolution image data subject to registration before piecemeal correspondent transform;
Converter unit traverses each pixel in piecemeal, and based on the low displacement field for differentiating non-linear registration, interpolation calculation goes out
The corresponding offset vector of the pixel calculates the pixel former high based on the inverse matrix that offset vector and high-resolution are registrated
The corresponding position for differentiating image data subject to registration, using its gray value as the pixel gray value.
The present invention have the advantage that for:Using low resolution image Registration of Measuring Data information, to magnanimity high-resolution 3-D view number
It is converted according to collection, realizes the non-linear registration of magnanimity high-resolution three-dimensional image data set.The present invention is suitable for based on transformation square
The linear registration of institute of battle array and all non-linear registrations based on displacement field can quickly non-linear match the progress of TB level data collection
It is accurate.
Description of the drawings
Fig. 1 is that the flow of the three dimensional non-linear method for registering of the magnanimity brain image data collection of the preferred embodiment for the present invention is shown
It is intended to;
Fig. 2 is the schematic diagram of full resolution pricture Fast transforms in the preferred embodiment for the present invention;
Fig. 3 is that the structure of the three dimensional non-linear registration arrangement of the magnanimity brain image data collection of the preferred embodiment for the present invention is shown
It is intended to.
Specific implementation mode
The present invention is described in further detail below in conjunction with the accompanying drawings.Fig. 1 shows that the preferred embodiment for the present invention carries
The flow diagram of the three dimensional non-linear method for registering of the magnanimity brain image data collection of confession.
A kind of three dimensional non-linear method for registering of magnanimity brain image data collection, includes the following steps:
S0. down-sampled step carries out drop to default high-resolution reference image data and high-resolution image data subject to registration and adopts
Sample obtains low resolution reference image data and low resolution image data subject to registration.
Wherein, the high-resolution reference image data D1, high-resolution image data D2 subject to registration.To two full resolution prictures
Data progress is down-sampled, obtains low resolution reference image data D3 and low resolution image data D4 subject to registration.Down-sampled process
Ensure at 2 points:Low resolution image data three-dimensional all directions resolution ratio is consistent;Two low resolution image data D3, D4 are whole
Resolution ratio is consistent.
S1. low resolution image step of registration, including:
S11. linear step of registration linearly matches low resolution reference image data and low resolution image data subject to registration
Standard obtains the low transformation matrix differentiated and be linearly registrated, and calculates the low inverse matrix for differentiating registration of its inverse matrix and low resolution line
Property registration after result;
Wherein, using the linear registration based on affine transformation, low resolution image data D3, D4 is linearly registrated, are obtained
To the low transformation matrix M1 for differentiating and being linearly registrated, and calculate inverse matrix M2 and low point that the low resolution of its inverse matrix is linearly registrated
Result D5 after litz wire registration.
Wherein, the linear registration described in the step S11, for transformed mappings can be described as to the wired of transformation matrix
Property method for registering, such as:The linear registration such as translation, rotation, scaling.
S12. non-linear registration step carries out result after the linear registration of low resolution and low resolution reference image data non-thread
Property registration, obtain it is low differentiate non-linear registration displacement field and low resolution non-linear registration after result;
Wherein, result D5 and low resolution reference image data D3 after the low linear registration of resolution is carried out being based on displacement field
Non-linear registration, obtain it is low differentiate non-linear registration displacement field T1 and low resolution non-linear registration after result.
Wherein, the non-linear registration described in the step S12, for transformed mappings can be described as to pixel-shift vector
All non-linear registration methods of displacement field, such as:The non-linear registration that elastic model, fluid model convert
S2. full resolution pricture Fast transforms step, including:
S21. the transformation matrix of high-resolution registration calculates step, and height is calculated based on the low transformation matrix being linearly registrated of differentiating
Differentiate the transformation matrix of registration and the inverse matrix of its inverse matrix high-resolution registration;
Wherein, based on the low transformation matrix M3 for differentiating the transformation matrix M1 being linearly registrated and calculating high-resolution registration, and
The inverse matrix M4 of its inverse matrix high-resolution registration, formula are as follows:
M3=scaleM × M3 × scaleM '
M4=M3-1
Wherein scaleM and scaleM ' is to reduce matrix and amplification matrix respectively, the multiple and high-resolution for reducing and amplifying
It is consistent with low resolution multiple.
S22. spatial dimension calculates step after high-resolution registration, utilizes the inverse matrix of high-resolution registration obtained by step S21, meter
Calculate spatial dimension after high-resolution is registrated;
Wherein, it is converted using 8 vertex of high-resolution inverse matrix M4 image data D2s subject to registration to high-resolution, takes him
X, Y, Z-direction maximin as high-resolution images after registration data D6 high-resolution registration after spatial dimension.
S23. blocking step carries out piecemeal to spatial dimension after high-resolution registration obtained by step S22;
Wherein, it is based on computing resource and carries out piecemeal, the transformation of pixel in parallel computation piecemeal.It is right according to computing environment
Spatial dimension carries out Z-direction piecemeal after the high-resolution registration.If computing environment maximum can support n thread parallel, with length
It spends n and carries out piecemeal, this makes it possible to the same one piece of data of parallel computation.If it is more than memory that each piecemeal, which need to be loaded into internal storage data,
Piecemeal length reduces half, until meeting memory requirements.
S24. high-resolution non linear transformation step carries out parallel computation to each piecemeal obtained by step S23, calculates each
A pixel finally obtains full resolution pricture nonlinear transformation result in the corresponding position of former high-resolution image data subject to registration.
High-resolution nonlinear transformation described in the step S24, including following sub-step:
S241. the range of former high-resolution image data subject to registration before piecemeal correspondent transform is estimated;
S242. each pixel in piecemeal is traversed, based on the low displacement field for differentiating non-linear registration, interpolation calculation goes out the picture
The corresponding offset vector of vegetarian refreshments calculates the pixel in former high-resolution based on the inverse matrix that offset vector and high-resolution are registrated
The corresponding position of image data subject to registration, using its gray value as the pixel gray value.
Wherein, Z-direction model where it corresponds to former high-resolution image data D2 subject to registration is calculated first for each piecemeal
It encloses.Calculating piecemeal corresponds to the low piecemeal for differentiating image data D4 subject to registration and is preceded by low to the position of each pixel in piecemeal
The offset vector in the corresponding positions displacement field T1 three directions of non-linear registration is differentiated, is then linearly registrated multiplied by with low differentiate
Inverse matrix M2 obtains it and corresponds to the former low resolution positions image data D4 subject to registration.Piecemeal all pixels point is found out to correspond to
Behind the former low position for differentiating image data D4 subject to registration, the maximin in three directions is calculated, it is subject to registration to obtain former low resolution
The Z-direction range of image data D4 finally calculates the Z-direction range of corresponding former high-resolution image data D2 subject to registration again.
As shown in Figure 2.For pixel p0(x0,y0,z0), calculate its seat in low resolution image data D4 subject to registration
Mark pixel p1(x1,y1,z1), then it is (x that it, which corresponds to the coordinate of the low displacement field T1 for differentiating non-linear registration,1,y1,z1), due to x1,
y1,z1Not necessarily integer, so the corresponding position to the low displacement field T1 for differentiating non-linear registration is needed to carry out Tri linear interpolation
Calculate pixel p1Corresponding offset vector d1.According to the p being calculated1Point offset vector d1Calculate p0The corresponding offset of point
Vectorial d0.By pixel p0Coordinate value is added d with offset vector0, the coordinate p that will obtain2Multiplied by the inverse matrix being registrated with high-resolution
M4 finally obtains coordinate p3For p0The coordinate of inverse transformation, i.e. pixel p0The corresponding former high-resolution of inverse transformation image data subject to registration
Point in D2, using the gray value of this point as p0Gray value.
Described calculates the corresponding gray value of each pixel in piecemeal, after the completion of every section of calculating in the form of two dimensional image
Disk is written.Each piecemeal of cycle calculations is completed until entire high-resolution image data subject to registration calculates.
The present invention utilizes low resolution image Registration of Measuring Data information it can be seen from the above content, three-dimensional to magnanimity high-resolution
Image data set is converted, and realizes the non-linear registration of magnanimity high-resolution three-dimensional image data set.The present invention is suitable for being based on
Transformation matrix it is linear registration and all non-linear registrations based on displacement field, can quickly to TB level data collection carry out it is non-
Linear registration.
It is shown in Figure 3, the three dimensional non-linear registration for the magnanimity brain image data collection that an embodiment of the present invention provides
The structural schematic diagram of system.
A kind of three dimensional non-linear registration arrangement of magnanimity brain image data collection, comprises the following modules:
Down-sampled module, it is down-sampled to default high-resolution reference image data and high-resolution image data progress subject to registration,
Obtain low resolution reference image data and low resolution image data subject to registration;
Low resolution image registration module, including:
Linear registration unit is linearly registrated low resolution reference image data and low resolution image data subject to registration,
The low transformation matrix differentiated and be linearly registrated is obtained, and calculates the low inverse matrix for differentiating registration of its inverse matrix and low resolution linearly
Result after registration;
Non-linear registration unit carries out non-linear match to result after the linear registration of low resolution and low resolution reference image data
Standard obtains result after the low displacement field for differentiating non-linear registration and low resolution non-linear registration;
Full resolution pricture Fast transforms module, including:
The transformation matrix computing unit of high-resolution registration calculates high-resolution based on the low transformation matrix being linearly registrated of differentiating
The inverse matrix of the transformation matrix of registration and its inverse matrix high-resolution registration;
Spatial dimension computing unit after high-resolution registration, the transformation matrix computing unit being registrated using high-resolution are secured satisfactory grades
The inverse matrix for distinguishing registration calculates spatial dimension after high-resolution registration;
Blocking unit, spatial dimension is divided after high-resolution registration obtained by spatial dimension computing unit after being registrated to high-resolution
Block;
High-resolution non-linear conversion unit carries out parallel computation to each piecemeal obtained by blocking unit, calculates each
Pixel finally obtains full resolution pricture nonlinear transformation result in the corresponding position of former high-resolution image data subject to registration.
Transformed mappings are described as transformation matrix by the linear registration described in the linear registration unit.It is described non-linear
Transformed mappings are described as the displacement field of pixel-shift vector by the non-linear registration described in registration unit.
In the blocking unit, the piecemeal carries out piecemeal based on computing resource, pixel in parallel computation piecemeal
Transformation, the computing resource maximum can support n thread parallel, and piecemeal is carried out with length n.
High-resolution nonlinear transformation described in the high-resolution non-linear conversion unit, including with lower unit:Pro form bill
Member estimates the range of former high-resolution image data subject to registration before piecemeal correspondent transform;Converter unit traverses each pixel in piecemeal
Point, based on the low displacement field for differentiating non-linear registration, interpolation calculation goes out the corresponding offset vector of the pixel, is based on offset vector
With high-resolution registration inverse matrix, calculate the pixel former high-resolution image data subject to registration corresponding position, by its ash
Angle value is as the pixel gray value.
The present invention utilizes low resolution image Registration of Measuring Data information it can be seen from the above content, three-dimensional to magnanimity high-resolution
Image data set is converted, and realizes the non-linear registration of magnanimity high-resolution three-dimensional image data set.The present invention is suitable for being based on
Transformation matrix it is linear registration and all non-linear registrations based on displacement field, can quickly to TB level data collection carry out it is non-
Linear registration.
It should be noted last that the above examples are only used to illustrate the technical scheme of the present invention and are not limiting, although ginseng
It is described the invention in detail according to preferred embodiment, it will be understood by those of ordinary skill in the art that, it can be to the present invention
Technical solution be modified or replaced equivalently, without departing from the spirit of the technical scheme of the invention and range, should all cover
In the scope of the claims of the present invention.
Claims (10)
1. a kind of three dimensional non-linear method for registering of magnanimity brain image data collection, which is characterized in that include the following steps:
S0. down-sampled step, it is down-sampled to default high-resolution reference image data and high-resolution image data progress subject to registration,
Obtain low resolution reference image data and low resolution image data subject to registration;
S1. low resolution image step of registration, including:
S11. linear step of registration is linearly registrated low resolution reference image data and low resolution image data subject to registration,
The low transformation matrix differentiated and be linearly registrated is obtained, and calculates the low inverse matrix for differentiating registration of its inverse matrix and low resolution linearly
Result after registration;
S12. non-linear registration step carries out non-linear match to result after the linear registration of low resolution and low resolution reference image data
Standard obtains result after the low displacement field for differentiating non-linear registration and low resolution non-linear registration;
S2. full resolution pricture Fast transforms step, including:
S21. the transformation matrix of high-resolution registration calculates step, and high-resolution is calculated based on the low transformation matrix being linearly registrated of differentiating
The inverse matrix of the transformation matrix of registration and its inverse matrix high-resolution registration;
S22. spatial dimension calculates step after high-resolution registration, using the inverse matrix of high-resolution registration obtained by step S21, calculates
Spatial dimension after high-resolution registration;
S23. blocking step carries out piecemeal to spatial dimension after high-resolution registration obtained by step S22;
S24. high-resolution non linear transformation step carries out parallel computation to each piecemeal obtained by step S23, calculates each picture
Vegetarian refreshments finally obtains full resolution pricture nonlinear transformation result in the corresponding position of former high-resolution image data subject to registration.
2. according to the three dimensional non-linear method for registering of the magnanimity brain image data collection described in claim 1, which is characterized in that institute
The linear registration described in step S11 is stated, transformed mappings are described as transformation matrix.
3. according to the three dimensional non-linear method for registering of the magnanimity brain image data collection described in claim 1, which is characterized in that institute
The non-linear registration described in step S12 is stated, transformed mappings are described as to the displacement field of pixel-shift vector.
4. according to the three dimensional non-linear method for registering of the magnanimity brain image data collection described in claim 1, which is characterized in that institute
The piecemeal described in step S23 is stated, piecemeal is carried out based on computing resource, the transformation of pixel, the meter in parallel computation piecemeal
N thread parallel can be supported by calculating resource maximum, and piecemeal is carried out with length n.
5. according to the three dimensional non-linear method for registering of the magnanimity brain image data collection described in claim 1, which is characterized in that institute
State the high-resolution nonlinear transformation described in step S24, including following sub-step:
S241. the range of former high-resolution image data subject to registration before piecemeal correspondent transform is estimated;
S242. each pixel in piecemeal is traversed, based on the low displacement field for differentiating non-linear registration, interpolation calculation goes out the pixel
Corresponding offset vector is calculated the pixel and is waited matching in former high-resolution based on the inverse matrix that offset vector and high-resolution are registrated
The corresponding position of quasi- image data, using its gray value as the pixel gray value.
6. a kind of three dimensional non-linear registration arrangement of magnanimity brain image data collection, which is characterized in that comprise the following modules:
Down-sampled module, it is down-sampled to default high-resolution reference image data and high-resolution image data progress subject to registration, it obtains
Low resolution reference image data and low resolution image data subject to registration;
Low resolution image registration module, including:
Linear registration unit is linearly registrated low resolution reference image data and low resolution image data subject to registration, is obtained
It is low to differentiate the transformation matrix being linearly registrated, and calculate the low inverse matrix for differentiating registration of its inverse matrix and the linear registration of low resolution
Result afterwards;
Non-linear registration unit carries out non-linear registration to result after the linear registration of low resolution and low resolution reference image data,
Obtain it is low differentiate non-linear registration displacement field and low resolution non-linear registration after result;
Full resolution pricture Fast transforms module, including:
The transformation matrix computing unit of high-resolution registration calculates high-resolution registration based on the low transformation matrix being linearly registrated of differentiating
Transformation matrix and its inverse matrix high-resolution registration inverse matrix;
Spatial dimension computing unit after high-resolution registration is matched using the transformation matrix computing unit gained high-resolution of high-resolution registration
Accurate inverse matrix calculates spatial dimension after high-resolution registration;
Blocking unit, spatial dimension carries out piecemeal after high-resolution registration obtained by spatial dimension computing unit after being registrated to high-resolution;
High-resolution non-linear conversion unit carries out parallel computation to each piecemeal obtained by blocking unit, calculates each pixel
Point finally obtains full resolution pricture nonlinear transformation result in the corresponding position of former high-resolution image data subject to registration.
7. according to the three dimensional non-linear registration arrangement of the magnanimity brain image data collection described in claim 6, which is characterized in that institute
The linear registration described in linear registration unit is stated, transformed mappings are described as transformation matrix.
8. according to the three dimensional non-linear registration arrangement of the magnanimity brain image data collection described in claim 6, which is characterized in that institute
The non-linear registration described in non-linear registration unit is stated, transformed mappings are described as to the displacement field of pixel-shift vector.
9. according to the three dimensional non-linear registration arrangement of the magnanimity brain image data collection described in claim 6, which is characterized in that institute
The piecemeal described in blocking unit is stated, piecemeal is carried out based on computing resource, the transformation of pixel, the meter in parallel computation piecemeal
N thread parallel can be supported by calculating resource maximum, and piecemeal is carried out with length n.
10. according to the three dimensional non-linear registration arrangement of the magnanimity brain image data collection described in claim 6, which is characterized in that
High-resolution nonlinear transformation described in the high-resolution non-linear conversion unit, including with lower unit:
Evaluation unit estimates the range of former high-resolution image data subject to registration before piecemeal correspondent transform;
Converter unit traverses each pixel in piecemeal, and based on the low displacement field for differentiating non-linear registration, interpolation calculation goes out the picture
The corresponding offset vector of vegetarian refreshments calculates the pixel in former high-resolution based on the inverse matrix that offset vector and high-resolution are registrated
The corresponding position of image data subject to registration, using its gray value as the pixel gray value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810307540.2A CN108537723B (en) | 2018-04-08 | 2018-04-08 | Three-dimensional nonlinear registration method and system for massive brain image data sets |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810307540.2A CN108537723B (en) | 2018-04-08 | 2018-04-08 | Three-dimensional nonlinear registration method and system for massive brain image data sets |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108537723A true CN108537723A (en) | 2018-09-14 |
CN108537723B CN108537723B (en) | 2021-09-28 |
Family
ID=63483358
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810307540.2A Active CN108537723B (en) | 2018-04-08 | 2018-04-08 | Three-dimensional nonlinear registration method and system for massive brain image data sets |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108537723B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112132877A (en) * | 2020-09-30 | 2020-12-25 | 东南大学 | Image registration improvement algorithm based on SIFT (scale invariant feature transform) features |
TWI737404B (en) * | 2020-07-15 | 2021-08-21 | 臺北醫學大學 | Medical image processing system and method thereof |
CN113516693A (en) * | 2021-05-21 | 2021-10-19 | 郑健青 | Rapid and universal image registration method |
US11238591B1 (en) | 2020-07-15 | 2022-02-01 | Taipei Medical University (Tmu) | Medical image processing system and method thereof |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1883386A (en) * | 2006-07-06 | 2006-12-27 | 上海交通大学 | Elastic registration method of stereo MRI brain image based on machine learning |
CN101238993A (en) * | 2008-02-01 | 2008-08-13 | 哈尔滨工业大学 | Medical ultrasound image registration method based on integer lifting wavelet multiresolution analysis |
CN104933716A (en) * | 2015-06-16 | 2015-09-23 | 山东大学(威海) | Non-rigid registration method applied to medical image |
CN105931184A (en) * | 2016-04-18 | 2016-09-07 | 西安电子科技大学 | SAR image super-resolution method based on combined optimization |
CN106803267A (en) * | 2017-01-10 | 2017-06-06 | 西安电子科技大学 | Indoor scene three-dimensional rebuilding method based on Kinect |
US9858640B1 (en) * | 2015-07-15 | 2018-01-02 | Hrl Laboratories, Llc | Device and method for merging 3D point clouds from sparsely distributed viewpoints |
-
2018
- 2018-04-08 CN CN201810307540.2A patent/CN108537723B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1883386A (en) * | 2006-07-06 | 2006-12-27 | 上海交通大学 | Elastic registration method of stereo MRI brain image based on machine learning |
CN101238993A (en) * | 2008-02-01 | 2008-08-13 | 哈尔滨工业大学 | Medical ultrasound image registration method based on integer lifting wavelet multiresolution analysis |
CN104933716A (en) * | 2015-06-16 | 2015-09-23 | 山东大学(威海) | Non-rigid registration method applied to medical image |
US9858640B1 (en) * | 2015-07-15 | 2018-01-02 | Hrl Laboratories, Llc | Device and method for merging 3D point clouds from sparsely distributed viewpoints |
CN105931184A (en) * | 2016-04-18 | 2016-09-07 | 西安电子科技大学 | SAR image super-resolution method based on combined optimization |
CN106803267A (en) * | 2017-01-10 | 2017-06-06 | 西安电子科技大学 | Indoor scene three-dimensional rebuilding method based on Kinect |
Non-Patent Citations (4)
Title |
---|
ARNO KLEIN 等: "Evaluation of 14 nonlinear deformation algorithms applied to human brain MRI registration", 《NEUROIMAGE》 * |
CHAOZHEN TAN 等: "3D Affine Registration of Large Image Stack for High-Resolution Brain Imaging Techniques", 《14TH INTERNATIONAL CONFERENCE ON PHOTONICS AND IMAGING IN BIOLOGY AND MEDICINE》 * |
TERRENCE CHEN 等: "A New Coarse-to-Fine Framework for 3D Brain MR Image Registration", 《INTERNATIONAL WORKSHOP ON COMPUTER VISION FOR BIOMEDICAL IMAGE APPLICATIONS》 * |
冯兆美 等: "基于B样条自由形变三维医学图像非刚性配准研究", 《影像科学与光化学》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI737404B (en) * | 2020-07-15 | 2021-08-21 | 臺北醫學大學 | Medical image processing system and method thereof |
US11238591B1 (en) | 2020-07-15 | 2022-02-01 | Taipei Medical University (Tmu) | Medical image processing system and method thereof |
CN112132877A (en) * | 2020-09-30 | 2020-12-25 | 东南大学 | Image registration improvement algorithm based on SIFT (scale invariant feature transform) features |
CN113516693A (en) * | 2021-05-21 | 2021-10-19 | 郑健青 | Rapid and universal image registration method |
Also Published As
Publication number | Publication date |
---|---|
CN108537723B (en) | 2021-09-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108537723A (en) | The three dimensional non-linear method for registering and system of magnanimity brain image data collection | |
Deng et al. | A novel camera calibration technique based on differential evolution particle swarm optimization algorithm | |
CN103369466B (en) | A kind of map match assists indoor orientation method | |
CN107798685A (en) | Pedestrian's height determines method, apparatus and system | |
CN103220777A (en) | Mobile device positioning system | |
CN102110284B (en) | Information processing apparatus and information processing method | |
CN101281648A (en) | Method for tracking dimension self-adaption video target with low complex degree | |
CN109186461A (en) | A kind of measurement method and measuring device of cabinet size | |
CN110570449A (en) | positioning and mapping method based on millimeter wave radar and visual SLAM | |
CN109816698A (en) | Unmanned plane visual target tracking method based on dimension self-adaption core correlation filtering | |
CN104034269A (en) | Monocular vision measuring method and monocular vision measuring device | |
CN115796712B (en) | Regional land ecosystem carbon reserve estimation method and device and electronic equipment | |
CN109580137A (en) | A kind of bridge structure displacement influence line measurement method based on computer vision technique | |
CN111914615A (en) | Fire-fighting area passability analysis system based on stereoscopic vision | |
CN110651475B (en) | Hierarchical data organization for compact optical streaming | |
CN108734148A (en) | A kind of public arena image information collecting unmanned aerial vehicle control system based on cloud computing | |
CN106778925B (en) | Face recognition pose over-complete face automatic registration method and device | |
CN103955687B (en) | A kind of method for rapidly positioning of the light spot image center based on centroid method | |
CN111161330B (en) | Non-rigid image registration method, device, system, electronic equipment and storage medium | |
CN113870327A (en) | Medical image registration method based on multi-level deformation field prediction | |
CN111885703A (en) | Indoor positioning method | |
CN107330934A (en) | The boundling Adjustable calculation method and system of low dimensional | |
CN106920213A (en) | The acquisition methods and system of a kind of high-definition picture | |
CN106775596A (en) | A kind of infrared image linear interpolation expands hardware processing method | |
CN107423513A (en) | A kind of method, equipment and its storage device of four-dimensional interpolation data processing |
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 |