CN106204679B - Projecting method, apparatus and system based on separable footprint function technology - Google Patents
Projecting method, apparatus and system based on separable footprint function technology Download PDFInfo
- Publication number
- CN106204679B CN106204679B CN201610566245.XA CN201610566245A CN106204679B CN 106204679 B CN106204679 B CN 106204679B CN 201610566245 A CN201610566245 A CN 201610566245A CN 106204679 B CN106204679 B CN 106204679B
- Authority
- CN
- China
- Prior art keywords
- detector
- image
- gpu
- projection
- value
- 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.)
- Expired - Fee Related
Links
Classifications
-
- 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
- G06T11/006—Inverse problem, transformation from projection-space into object-space, e.g. transform methods, back-projection, algebraic methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2200/00—Indexing scheme for image data processing or generation, in general
- G06T2200/28—Indexing scheme for image data processing or generation, in general involving image processing hardware
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2211/00—Image generation
- G06T2211/40—Computed tomography
- G06T2211/424—Iterative
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Algebra (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Mathematical Physics (AREA)
- Pure & Applied Mathematics (AREA)
- Image Analysis (AREA)
Abstract
Present invention implementation provides projecting method, apparatus and system based on separable footprint function technology, obtains the pixel value of the three-dimensional voxel of each coordinate position in 3-D image；Fuzzy footprint function in the horizontal direction that three-dimensional voxel described in parallel computation projects to detector on the cross section of the 3-D image；According to the projection value in the axis direction of detector described in pixel value parallel computation in the axis direction of the detector；The product of the projection value on fuzzy footprint function and the axis direction in the plane of the detector in horizontal direction described in parallel computation obtains the projection value of each detector cells of plane of the detector.Using the fuzzy footprint function in parallel computation horizontal direction, the projection value of projection value and each detector cells of parallel computation in parallel computation axis direction, improves the efficiency of projecting method at the time needed for shortening forward and backward projection.
Description
Technical field
The present invention relates to technical field of imaging, more particularly to a kind of projection side based on separable footprint function technology
Method, apparatus and system.
Background technique
Forward projection (Forward Projection) and back projection (Back Projection) are that imaging field is common
Projecting method.Forward projection is generated just using system model matrix and the product of the human body or object that are scanned by imaging system
To projection value, back projection is the conjugate operation of forward projection.
By taking X-CT (X-ray CT scan imaging technique) as an example, using the X-ray beam of Accurate collimation to human body
Organ makees profile scanning, and forward projection of the 3-D image in detector cells of scanning is received using the high detector of sensitivity
Value.According to each detector cells measured value collected of detector, each seat in 3-D image is obtained using back projection
The back projection value of cursor position.
But since the spatial resolution of X-CT imaging is very high, the dimension of system model matrix is up to 109*109Left and right, leads
Cause the time needed for obtaining 3-D image using existing forward projection and backprojection techniques very long, projecting method low efficiency.
Summary of the invention
Present invention solves the technical problem that being to provide a kind of projecting method, device based on separable footprint function technology
And system, so as to the time needed for shortening forward projection or back projection.
For this purpose, the technical solution that the present invention solves technical problem is:
A kind of projecting method based on separable footprint function technology, which comprises
Obtain the pixel value of the three-dimensional voxel of each coordinate position in 3-D image；
In the horizontal direction that three-dimensional voxel described in parallel computation projects to detector on the cross section of the 3-D image
Fuzzy footprint function, the horizontal direction is projecting direction of the cross section in the detector of the 3-D image；
According to the throwing in the axis direction of detector described in pixel value parallel computation in the axis direction of the detector
Shadow value, the projection value in the axis direction are the products of fuzzy the footprint function and the pixel value of the axis direction of the detector
In the sum of the axis direction of the 3-D image；
Fuzzy footprint function and the axis direction in the plane of the detector in horizontal direction described in parallel computation
On projection value product obtain the detector each detector cells of plane projection value.
Optionally, the projection value for calculating each detector cells includes:
Obtain projection model of the three-dimensional voxel of each coordinate position in the 3-D image in the horizontal direction of detector
It encloses；
Contributive drop shadow spreads all to detector cells are searched, the contributive drop shadow spread includes the detection
The horizontal coordinate value of device unit；
Three-dimensional voxel corresponding to all contributive drop shadow spreads is obtained as contributive three-dimensional voxel；
Utilize the projection value on the fuzzy footprint function and axis direction in the horizontal direction of the contributive three-dimensional voxel
Product obtain the projection values of the detector cells.
Optionally, the method also includes:
Obtain the measured value of each detector cells on detector plane；
Fuzzy footprint function and the spy according to the fuzzy footprint function in the horizontal direction, in the axis direction
The measured value of each detector cells is on each coordinate position of the 3-D image three described in parallel computation in survey device plane
Tie up the back projection value of each coordinate position in image.
Optionally, the measured value according to detector plane is counted parallel on each coordinate position of the 3-D image
The back projection value for calculating each coordinate position in the 3-D image includes:
According to each detector cells on the fuzzy footprint function and the detector plane in the horizontal direction
The projection value of measured value three-dimensional voxel of 3-D image described in parallel computation on the cross section of the 3-D image, the three-dimensional
The fuzzy footprint function that the projection value of the three-dimensional voxel of image is horizontally oriented multiplies with the measured value of the detector plane
Sum of the product on the three-dimensional voxel of the cross section of the 3-D image；
According to the projection value of the fuzzy footprint function in the axis direction and three-dimensional voxel on the cross section described
The projection value of each coordinate position, the projection value are in 3-D image described in each coordinate position parallel computation of 3-D image
The product of the projection value of three-dimensional voxel and the fuzzy footprint function in the axis direction is in the 3-D image on the cross section
Coordinate position sum.
Optionally,
Fuzzy footprint function in the horizontal direction is trapezoidal function or rectangular function.
Optionally,
Fuzzy footprint function in the axis direction is trapezoidal function or rectangular function.
A kind of projecting method that more GPU are realized, which comprises
Projection angle is divided equally into multiple projections according to the number of the GPU by the number of GPU in GPU acquisition system
Angular range, the number of the projection angle range are identical as GPU number；
GPU obtains the device numbering of the GPU, and the projection angle model for belonging to the GPU is obtained according to the device numbering
It encloses；
GPU obtains the pixel value of each coordinate position in 3-D image；
GPU is according to the pixel value according to above-mentioned forward projection's method parallel computation based on separable footprint function technology
The projection value of the planar detector unit of detector under the projection angle range.
Optionally, the method also includes:
The abscissa of the 3-D image or ordinate are divided equally into multiple coordinates according to the number of the GPU by GPU
Range, the number of the coordinate range and the number of GPU are identical；
GPU obtains the device numbering of the GPU, and the coordinate range for belonging to the GPU is obtained according to the device numbering；
GPU obtains the measured value of each detector cells in the plane of detector；
GPU is according to belonging to all measured values of the coordinate range according to above-mentioned based on the anti-of separable footprint function technology
Under to coordinate range described in projecting method parallel computation in the 3-D image three-dimensional voxel back projection value.
A kind of projection arrangement based on separable footprint function technology, described device include:
First obtains module, for obtaining the pixel value of the three-dimensional voxel of each coordinate position in 3-D image；
First computing module projects to spy for three-dimensional voxel described in the parallel computation on the cross section of the 3-D image
The fuzzy footprint function in the horizontal direction of device is surveyed, the horizontal direction is the cross section of the 3-D image in the detector
Projecting direction；
Second computing module, for being detected according to pixel value parallel computation in the axis direction of the detector
Projection value in the axis direction of device, the projection value in the axis direction be the axis direction of the detector fuzzy footprint function with
Sum of the product of the pixel value in the axis direction of the 3-D image；
Third computing module, the fuzzy footprint in horizontal direction described in the parallel computation in the plane of the detector
The product of projection value on function and the axis direction obtains the projection value of each detector cells of plane of the detector.
Optionally,
The third module includes multiple Unit the 4th, is visited in the number and the plane of the detector of Unit the 4th
The number for surveying device unit is identical；Each Unit the 4th includes:
First obtains subelement, for obtaining the three-dimensional voxel of each coordinate position in the 3-D image in detector
Drop shadow spread in horizontal direction；
Subelement is searched, for searching contributive drop shadow spreads all to detector cells, the contributive projection
Range includes the horizontal coordinate value of the detector cells；
Second obtains subelement, contributes for obtaining the conduct of three-dimensional voxel corresponding to all contributive drop shadow spreads
Three-dimensional voxel；
First computation subunit, for utilizing the fuzzy footprint function in the horizontal direction of the contributive three-dimensional voxel
The projection value of the detector cells is calculated with the product of the projection value in axis direction.
Optionally, described device further include:
Second obtains module, for obtaining the measured value of each detector cells on detector plane；
4th computing module is fuzzy in the axis direction for according to the fuzzy footprint function in the horizontal direction
Each coordinate bit of the measured value of each detector cells in the 3-D image on footprint function and the detector plane
Set the back projection value of each coordinate position in 3-D image described in parallel computation.
Optionally,
4th computing module includes multiple second computation subunits and multiple third computation subunits, second son
The number of unit is identical as the number of three-dimensional voxel on the cross section of the 3-D image, the third computation subunit with it is described
The number of the three-dimensional voxel of 3-D image is identical；
Second computation subunit, for according to the fuzzy footprint function and the detector in the horizontal direction
In plane the measured value of each detector cells on the cross section of the 3-D image 3-D image described in parallel computation three
Tie up voxel projection value, the fuzzy footprint function that the projection value of the three-dimensional voxel of the 3-D image is horizontally oriented with it is described
Sum of the product of the measured value of detector plane on the three-dimensional voxel of the cross section of the 3-D image；
The third subelement, for according in the axis direction fuzzy footprint function and the cross section on it is three-dimensional
The projection value of voxel each coordinate position in the 3-D image described in each coordinate position parallel computation of the 3-D image
Projection value, the projection value are the projection value of three-dimensional voxel and the fuzzy footprint function in the axis direction on the cross section
Sum of the product in the coordinate position of the 3-D image.
A kind of GPU, the GPU include:
Projection angle is averaged for the number of GPU in acquisition system according to the number of the GPU by the first division module
Multiple projection angle ranges are divided into, the number of the projection angle range is identical as GPU number；
First obtains module, for obtaining the device numbering of the GPU, is belonged to according to device numbering acquisition described
The projection angle range of GPU；
Second obtains module, for obtaining the pixel value of each coordinate position in 3-D image；
First computing module, for according to the pixel value according to the forward projection side based on separable footprint function technology
The projection value of the planar detector unit of detector under projection angle range described in method parallel computation.
Optionally, the GPU further include:
Second division module, for the abscissa of the 3-D image or ordinate to be averaged according to the number of the GPU
Multiple coordinate ranges are divided into, the number of the coordinate range and the number of GPU are identical；
Third obtains module, for obtaining the device numbering of the GPU, is belonged to according to device numbering acquisition described
The coordinate range of GPU；
4th obtains module, the measured value of each detector cells in the plane for obtaining detector；
Second computing module belongs to all measured values of the coordinate range according to based on separable footprint letter for basis
Described in the back projection method parallel computation of number technologies under coordinate range in the 3-D image three-dimensional voxel projection value.
A kind of system, the system comprises:
Multiple above-mentioned GPU.
According to the above-mentioned technical solution, the method have the advantages that:
Present invention implementation provides projecting method, apparatus and system based on separable footprint function technology, obtains three-dimensional
The pixel value of the three-dimensional voxel of each coordinate position in image；The three-dimensional described in parallel computation on the cross section of the 3-D image
Voxel projects to the fuzzy footprint function in the horizontal direction of detector；According to the pixel value the detector axis direction
Projection value in the axis direction of detector described in upper parallel computation；The level side described in parallel computation in the plane of the detector
The product of projection value on upward fuzzy footprint function and the axis direction obtains each detector of plane of the detector
The projection value of unit.Using the fuzzy footprint function in parallel computation horizontal direction, projection value in parallel computation axis direction with
And the projection value of each detector cells of parallel computation, time needed for shortening forward projection, improve forward projection and reversed throwing
The efficiency of shadow.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is the projecting method flow chart provided in an embodiment of the present invention based on separable footprint function technology；
Fig. 2 is the schematic diagram that scanning device provided in an embodiment of the present invention irradiates that three-dimensional imaging body projects on the detector；
Fig. 3 is the projecting method flow chart that more GPU provided in an embodiment of the present invention are realized；
Fig. 4 is the projection arrangement structural schematic diagram provided in an embodiment of the present invention based on separable footprint function technology；
Fig. 5 is a kind of GPU architecture schematic diagram provided in an embodiment of the present invention；
Fig. 6 is the projection system architecture schematic diagram provided in an embodiment of the present invention based on separable footprint function technology；
Fig. 7 is the projection arrangement hardware configuration signal provided in an embodiment of the present invention based on separable footprint function technology
Figure；
Fig. 8 is a kind of GPU hardware structural schematic diagram provided in an embodiment of the present invention；
Fig. 9 is the optical projection system hardware configuration signal provided in an embodiment of the present invention based on separable footprint function technology
Figure.
Specific embodiment
In order to provide shorten projecting method needed for the time implementation, the embodiment of the invention provides one kind based on can
Projecting method, the apparatus and system for separating footprint function technology, below in conjunction with Figure of description to the preferred embodiment of the present invention
It is illustrated.
Embodiment one
Forward projection can be indicated using formula (1):
Wherein, f (x, y, z) is that coordinate position is the pixel value of the three-dimensional voxel of (x, y, z) in 3-D image；g(s,t,β)
For the projection value of detector cells on detector plane, β is the projection angle of 3-D image, and s is the horizontal seat on detector plane
Mark, t are the ordinate on detector plane, and the direction t is parallel with the direction z (axis direction of 3-D image), a (s, t, β；x,y,z)
To obscure footprint function.
a(s,t,β；X, y, z) it can be indicated with formula (2):
a(s,t,β；X, y, z)=u (β；x,y)v(s,t,β)F1(s,β；x,y)F2(t,β；x,y,z) (2)；
Wherein, u is the first amplitude function, and v is the second amplitude function, F1For the fuzzy footprint function in horizontal direction, F2For
The fuzzy footprint function of axis direction.
By formula (1) and formula (2) it is found that forward projection is indicated using formula (3):
Fig. 1 is the projecting method flow chart provided in an embodiment of the present invention based on separable footprint function technology, comprising:
101: obtaining the pixel value of the three-dimensional voxel of each coordinate position in 3-D image.
Fig. 2 is the schematic diagram that scanning device irradiates that three-dimensional imaging body projects on the detector.Wherein, the direction z is three-dimensional figure
The axis direction of picture, the i.e. axis direction of scanning device；(x, y) plane is the irradiated plane in flying spot source in scanning device, i.e.,
Cross section of the three-dimensional imaging body perpendicular to axis direction；Dotted line is the motion profile in flying spot source in detecting devices；The direction s is
Projection of (x, the y) plane in detector plane.
Scanning device irradiates three-dimensional imaging body and obtains a 3-D image, can obtain each coordinate position of 3-D image
Three-dimensional voxel pixel value, that is, obtain the pixel value of the three-dimensional voxel of each (x, y, z) coordinate position.
102: three-dimensional voxel described in parallel computation projects to the level side of detector on the cross section of the 3-D image
Upward fuzzy footprint function, the horizontal direction are projecting direction of the cross section in the detector of the 3-D image.
Fuzzy footprint function in parallel computation horizontal direction is that x × y thread is utilized on the cross section of 3-D image
Carry out parallel computation.For example, as shown in Fig. 2, being counted simultaneously in (x, y) plane of 3-D image using x × y thread
Calculate the fuzzy footprint function F in the horizontal direction of each three-dimensional voxel1(s,β；x,y).
For per thread, a column three-dimensional voxel in a thread corresponding three-dimensional image, i.e., the same thread institute
(x, y) of the coordinate position for the three-dimensional voxel to be calculated is identical, but z is different.Identical (x, y), the coordinate position of different z
Fuzzy footprint function in the horizontal direction of three-dimensional voxel is identical.
Wherein, the fuzzy footprint function F in horizontal direction1(s,β；X, y) it can be trapezoidal function, it is also possible to rectangle letter
Number.Specific calculation is not discussed here referring to the patent of Patent No. US8913805 B2.
103: according in the axis direction of detector described in pixel value parallel computation in the axis direction of the detector
Projection value, the projection value in the axis direction is the fuzzy footprint function and the pixel value of the axis direction of the detector
Sum of the product in the axis direction of the 3-D image.
Projection value in the axis direction of parallel computation detector is carried out simultaneously in the axis direction of detector using t thread
Row calculates.For example, as shown in Fig. 2, on the direction t of detector, using in t thread parallel calculating detector axis direction
Projection value.
For a thread, a thread corresponds to a t value in the axis direction of detector, i.e. a thread institute
It is identical that the z coordinate for the three-dimensional voxel to be calculated projects to the t value on detector axis direction.It precalculates each three in 3-D image
Tie up the fuzzy footprint function F of voxel2(t,β；X, y, z), per thread calculates the fuzzy footprint letter of all three-dimensional voxels of the thread
Number F2(t,β；X, y, z) with the sum of pixel value f (x, y, z) product of the three-dimensional voxel as in thread axis direction calculated
Projection value p1(t,β；x,y).
Wherein, the fuzzy footprint function F in axis direction2(t,β；X, y, z) it can be trapezoidal function, it is also possible to rectangle letter
Number.Specific calculation is referring to the patent of Patent No. US8913805 B2, and which is not described herein again.
For example, calculating the throwing in the axis direction that the thread to be calculated using formula (4) for per thread
Shadow value p1(t,β；X, y):
In another example, if projection value p in axis direction1(t,β；X, y) when being projection value without amplitude, using public affairs
Formula (5) calculates the projection value p for having amplitude in axis direction2(t,β；X, y):
p2(t,β；X, y)=p1(t,β；x,y)u(β；x,y) (5)；
Wherein, u (β；X, y) it is the first amplitude function.
104: fuzzy footprint function and the axis in the plane of the detector in horizontal direction described in parallel computation
The product of projection value on direction obtains the projection value of each detector cells of plane of the detector.
The projection value of each detector cells of parallel computation is to utilize multithreads computing, institute on detector plane
The number of the thread of use is identical as the number of detector cells on detector plane.For example, as shown in Fig. 2, detecting
Parallel computation is carried out using s × t thread in device plane, the throwing of a detector cells in per thread calculating detector plane
Shadow value g (s, t, β).
In one example, for a thread, per thread is by the following method one in calculating detector plane
The projection value g (s, t, β) of a detector cells:
Obtain projection model of the three-dimensional voxel of each coordinate position in the 3-D image in the horizontal direction of detector
It encloses；
Contributive drop shadow spreads all to detector cells are searched, the contributive drop shadow spread includes the detection
The horizontal coordinate value of device unit；
Three-dimensional voxel corresponding to all contributive drop shadow spreads is obtained as contributive three-dimensional voxel；
Utilize the projection value on the fuzzy footprint function and axis direction in the horizontal direction of the contributive three-dimensional voxel
Product obtain the projection values of the detector cells.
As shown in Fig. 2, three-dimensional image projection is on the horizontal direction s of detector plane, for the same of 3-D image
For three-dimensional voxel, the projection angle β of 3-D image is different, projection position of the three-dimensional voxel on the direction s of detector plane
Set difference.Each three-dimensional voxel Dou Youyige drop shadow spread (s on the direction s of detector plane in 3-D imagemin,smax)。
In one example, drop shadow spread of each three-dimensional voxel on the direction s of detector plane is good in advance,
Used calculation method are as follows:
The drop shadow spread of each three-dimensional voxel of parallel computation calculates each three-dimensional voxel in the projected angle of different 3-D images
The position coordinates for spending the projection under β on detector plane, search maximum abscissa s in resulting position coordinatesmaxAnd minimum
Abscissa smin, then drop shadow spread of the three-dimensional voxel in detector horizontal direction is (smin,smax)。
For detector cells each on detector plane, if the abscissa s of the detector cells is located at three-dimensional figure
Drop shadow spread (the s of a three-dimensional voxel as inmin,smax) in, then the three-dimensional voxel is contributive to the detector cells three
Tie up voxel.For example, if the abscissa of detector cells is 5, if the drop shadow spread of a three-dimensional voxel of 3-D image is
(3,7), then the three-dimensional voxel is to the contributive three-dimensional voxel of the detector cells；If a three-dimensional voxel of 3-D image
Drop shadow spread is (6,10), then the three-dimensional voxel is that do not have contributive three-dimensional voxel to the detector cells.
All three-dimensional voxels contributive to the detector cells on thread detector plane calculated are obtained, benefit
It is obtained with the product of the projection value on the fuzzy footprint function and axis direction in the horizontal direction of the contributive three-dimensional voxel
The projection value of the detector cells.
In one example, per thread calculates the projection value of the corresponding detector cells of the thread according to formula (6):
Wherein, g'(s, t, β) be detector cells projection value, F1(s,β；X, y) it is to the corresponding detector of the thread
Fuzzy footprint function in the horizontal direction of the contributive three-dimensional voxel of unit, p2(t,β；X, y) it is to the corresponding spy of the thread
The projection value in the axis direction of the contributive three-dimensional voxel of device unit is surveyed, the value range of s is s ∈ [smin,smax]。
In one example, g'(s, t, β) be have amplitude projection value when, can be directly as the projection of detector cells
Value g (s, t, β).
In another example, g'(s, t, β) when being projection value without amplitude, the throwing for having amplitude is calculated using formula (7)
Shadow value g (s, t, β):
G (s, t, β)=v (s, t, β) g'(s, t, β) (7)；
Wherein, v (s, t, β) is the second amplitude function.
For example: forward projection's method of the above-mentioned forward projection's method of the present invention and the prior art is applied respectively in axle position
The test result of the circle spiral CT of CT and 8.Radiation and nuclear medicine 3-D image of the prior art referring to 2011 in the application
56-59 pages of international conference collection of thesis is rebuild, paper is entitled " based on separable footprint function method in X ray CT image reconstruction
Three-dimensional forward and reverse projection GPU accelerating algorithm ", author is Wu's Mongolian, the paper of Geoffrey A Fei Sile.The paper describes S-
The GPU accelerating algorithm of F projector.(prior art reference: Wu Meng, and Jeffrey A.Fessler. " GPU
acceleration of 3D forward and backward projection using separable footprints
for X-ray CT image reconstruction.”Proc.Intl.Mtg.on Fully 3D Image Recon.in
Rad.and Nuc.Med.2011)。
The runing time and accuracy rate of 1 axle position CT of table compares
| Runing time (s) | NRMS error (%) | |
| The prior art | 11.64 | 0.000074 |
| The present invention | 7.92 | 0.000055 |
Table 28 encloses the runing time of spiral CT and accuracy rate compares
| Runing time (s) | NRMS error (%) | |
| The prior art | 71.27 | 0.000205 |
| The present invention | 57.23 | 0.000199 |
From the above data, it can be seen that forward direction used by forward projection's method of the present invention and the prior art is thrown
The method of shadow is compared, and time-consuming short, accuracy is high.
As shown in the above, the method have the advantages that:
In the horizontal direction that three-dimensional voxel described in parallel computation projects to detector on the cross section of the 3-D image
Fuzzy footprint function；According to the axis direction of detector described in pixel value parallel computation in the axis direction of the detector
On projection value；Fuzzy footprint function and the axis side in the plane of the detector in horizontal direction described in parallel computation
The product of upward projection value obtains the projection value of each detector cells of plane of the detector.Using parallel computation level
Fuzzy footprint function on direction, the projection of projection value and each detector cells of parallel computation in parallel computation axis direction
Value, improves the efficiency of projecting method at the time needed for shortening forward projection.
Above-mentioned is the specific descriptions to forward projection in projecting method.Projecting method further includes backprojection techniques.Reversely
Projection can be indicated with formula (8):
Wherein, b (x, y, z) is the back projection value of each coordinate position in 3-D image.
Due to a (s, t, β；X, y, z) it can be indicated with formula (2), according to formula (2) and formula (8) it is found that back projection
It can be indicated with formula (9):
Then in one example, the method also includes:
Obtain the measured value of each detector cells on detector plane；
Fuzzy footprint function and the spy according to the fuzzy footprint function in the horizontal direction, in the axis direction
The measured value of each detector cells is on each coordinate position of the 3-D image three described in parallel computation in survey device plane
Tie up the back projection value of each coordinate position in image.
It can only include forward projection's mistake of the transparent value of each detector cells of above-mentioned calculating during projecting method
Journey can also only include the back projection process for calculating the back projection value of each coordinate position in 3-D image, can also be both
It also include back projection process comprising forward projection's process.Back projection process is described in detail below.
Each coordinate position in 3-D image described in parallel computation on each coordinate position of the 3-D image
Back projection value include two parallel computations process, it is as follows:
According to each detector cells on the fuzzy footprint function and the detector plane in the horizontal direction
The projection value of measured value three-dimensional voxel of 3-D image described in parallel computation on the cross section of the 3-D image, the three-dimensional
The fuzzy footprint function that the projection value of the three-dimensional voxel of image is horizontally oriented multiplies with the measured value of the detector plane
Sum of the product on the three-dimensional voxel of the cross section of the 3-D image；
According to the projection value of the fuzzy footprint function in the axis direction and three-dimensional voxel on the cross section described
The projection value of each coordinate position, the projection value are in 3-D image described in each coordinate position parallel computation of 3-D image
The product of the projection value of three-dimensional voxel and the fuzzy footprint function in the axis direction is in the 3-D image on the cross section
Coordinate position sum.
The projection value of the three-dimensional voxel of 3-D image described in parallel computation is parallel on the cross section of the 3-D image
It calculates, carries out parallel computation using x × y × t thread, the z coordinate for the three-dimensional voxel in 3-D image that a thread calculates
The t value projected on detector axis direction is identical.For example: as shown in Fig. 2, calculating three simultaneously using x × y × t thread
The projection value of voxel is tieed up, the z coordinate of the coordinate position of the same thread three-dimensional voxel calculated projects in detection axis direction
T value it is identical, the projection value of same thread all three-dimensional voxels calculated is all identical.
In one example, per thread calculates the projection value of each three-dimensional voxel using formula (10):
Wherein, the value range of s is s ∈ [smin,smax]。
In another example, b1(t,β；X, y) if there is the projection value of amplitude, then b1(t,β；X, y) it is the three-dimensional
The projection value of the three-dimensional voxel of image, if b1(t,β；X, y) it is the projection value without amplitude, then described three are calculated using formula (11)
Tie up the projection value b of the three-dimensional voxel of image2(t,β；X, y):
b2(t,β；X, y)=b1(t,β；x,y)u(β；x,y) (11).
The projection value of each coordinate position in parallel computation 3-D image, be on each three-dimensional voxel of 3-D image simultaneously
Row calculates, and carries out parallel computation using multiple threads, the number of thread is identical as the number of the three-dimensional voxel in 3-D image, often
A thread calculates the back projection value of a three-dimensional voxel.For example: as shown in Fig. 2, calculating three using x × y × z thread
The back projection value of voxel is tieed up, per thread calculates the back projection value of the three-dimensional voxel of a coordinate position (x, y, z).
Per thread is calculated when calculating the back projection value of a three-dimensional voxel using formula (12):
Wherein, b (x, y, z) is the back projection value of three-dimensional voxel in the 3-D image, and the value range of z is z ∈
[zmin(t,β；x,y,z),zmax(t,β；x,y,z)].
As shown in the above, the present invention there are also following the utility model has the advantages that
The method of back projection provided by the present invention, the back projection of each coordinate position in parallel computation 3-D image
Value, improves the speed of projecting method at the time needed for shortening back projection.
For example: back projection method of the present invention and the back projection method of the prior art apply respectively axle position CT with
And 8 circle spiral CT test result:
The runing time and accuracy rate of 3 axle position CT of table compares
| Runing time (s) | NRMS error (%) | |
| The prior art | 16.34 | 0.457756 |
| The present invention | 5.62 | 0.0246058 |
Table 48 encloses the runing time of spiral CT and accuracy rate compares
| Runing time (s) | NRMS error (%) | |
| The prior art | 129.68 | 3.08 |
| The present invention | 38.67 | 0.0264 |
From the above data, it can be seen that back projection method of the present invention with reversely thrown used by the prior art
The method of shadow is compared, and time-consuming short, accuracy is high.
Embodiment two
Fig. 3 is the projecting method flow chart that more GPU provided in an embodiment of the present invention are realized, comprising:
Projection angle is divided equally into multiple by the number of GPU in 301:GPU acquisition system according to the number of the GPU
Projection angle range, the number of the projection angle range are identical as GPU number.
302:GPU obtains the device numbering of the GPU, and the projected angle for belonging to the GPU is obtained according to the device numbering
Spend range.
303:GPU obtains the pixel value of each coordinate position in 3-D image.
304:GPU is projected according to method parallel computation of the pixel value according to forward projection described in embodiment one
The projection value of the planar detector unit of detector under angular range.
In order to further increase the speed of forward projection in projecting method, multiple graphics processors can be used
(Graphics Processing Unit, GPU) parallel processing simultaneously.By the projection angle β of 3-D image according to the number of GPU
Multiple projection angle ranges are divided into, the number of projection angle range and the number of GPU are identical.Each GPU handles this GPU pairs
Three-dimensional voxel in the 3-D image for the projection angle range answered calculates the planar detector of the detector under the projection angle range
The projection value of device unit.
For example: according to 3 GPU parallel computations, the projection angle β of 3-D image is 360 °, and projection angle β is put down
It is divided into three parts, wherein the three-dimensional voxel for the coordinate position that a GPU processing projection angle β is 0~120 °, a GPU processing
Projection angle β is the three-dimensional voxel of 121 °~240 ° coordinate positions, and a GPU processing projection angle β is 241 °~360 ° coordinates
The three-dimensional voxel of position.
What needs to be explained here is that examples detailed above merely to better understand embodiments of the present invention, is not limited only to
The description of examples detailed above can also carry out parallel processing using the GPU of other numbers when specific implementation, and implementation method is similar, this
In repeat no more.
The method (method of forward projection) and Fig. 1 of the projection value of the planar detector unit of each GPU calculating detector
Shown in forward projection in method method it is similar, only each GPU handles the three of the coordinate position of different projection angle β
Voxel is tieed up, with reference to the method for the projection value of the planar detector unit of calculating detector shown in FIG. 1, which is not described herein again.
In one example, further includes:
The abscissa of the 3-D image or ordinate are divided equally into multiple coordinates according to the number of the GPU by GPU
Range, the number of the coordinate range and the number of GPU are identical；
GPU obtains the device numbering of the GPU, and the coordinate range for belonging to the GPU is obtained according to the device numbering；
GPU obtains the measured value of each detector cells in the plane of detector；
GPU is according to belonging to described in method parallel computation of all measured values of the coordinate range according to above-mentioned back projection
Under coordinate range in the 3-D image three-dimensional voxel back projection value.
It, can be using multiple GPU parallel processing simultaneously in order to further increase the speed of back projection in projecting method.It will
The abscissa x or ordinate y of 3-D image are divided equally into multiple coordinate ranges, the number of the coordinate range and of GPU
Number is identical.Each GPU handles the three-dimensional voxel in the 3-D image of the corresponding coordinate range of the GPU, calculates under the coordinate range
3-D image in three-dimensional voxel back projection value.
For example: according to 3 GPU parallel computations, the range of the abscissa x of 3-D image is (- 6,6), by horizontal seat
Mark x is divided into three parts, wherein a GPU processing abscissa x is the three-dimensional voxel of the coordinate position of (- 6, -2), a GPU
The three-dimensional voxel that abscissa x is (- 2,2) coordinate position is handled, a GPU processing abscissa x is the three-dimensional of (2,6) coordinate position
Voxel.
For example: according to 2 GPU parallel computations, the range of the abscissa y of 3-D image is (- 2,2), by horizontal seat
Mark y is divided into two parts, wherein a GPU processing abscissa x is the three-dimensional voxel of the coordinate position of (- 2,0), another GPU
Handle the three-dimensional voxel that abscissa x is (0,2) coordinate position.
What needs to be explained here is that examples detailed above merely to better understand embodiments of the present invention, is not limited only to
The description of examples detailed above can also carry out parallel processing using the GPU of other numbers when specific implementation, and implementation method is similar, this
In repeat no more.
Each GPU calculates the method (method of back projection) and Fig. 1 of the back projection value of three-dimensional voxel in 3-D image
Shown in method back projection method it is similar, only each GPU handles the three-dimensional voxel of different coordinate positions, with reference to figure
The method of the back projection value of three-dimensional voxel in calculating 3-D image shown in 1, which is not described herein again.
For example: when carrying out forward projection and back projection calculating using the GPU of different numbers, the ratio of required time
Compared with as shown in table 5 and table 6.
5 axle position CT runing time of table compares
| Forward projection's time (s) | The back projection time (s) | |
| One GPU | 7.92 | 5.62 |
| Two GPU | 4.03 | 2.89 |
| Three GPU | 2.78 | 1.94 |
Table 68 encloses spiral CT runing time and compares
| Forward projection's time (s) | The back projection time (s) | |
| One GPU | 57.23 | 38.67 |
| Two GPU | 28.98 | 19.52 |
| Three GPU | 20.10 | 13.42 |
From the above data, it can be seen that used GPU number is more, carry out needed for forward projection and rear orientation projection's calculating
Time it is shorter.
Embodiment three
Fig. 4 is the projection arrangement structural schematic diagram provided in an embodiment of the present invention based on separable footprint function technology, packet
It includes:
First obtains module 401, for obtaining the pixel value of the three-dimensional voxel of each coordinate position in 3-D image.
First computing module 402 is projected for three-dimensional voxel described in the parallel computation on the cross section of the 3-D image
Fuzzy footprint function on to the horizontal direction of detector, the horizontal direction are the cross sections of the 3-D image in the spy
Survey the projecting direction of device.
Second computing module 403, for according to pixel value parallel computation in the axis direction of the detector
Projection value in the axis direction of detector, the projection value in the axis direction are the fuzzy footprint letters of the axis direction of the detector
Several products with the pixel value the 3-D image axis direction and.
Third computing module 404 is fuzzy in horizontal direction described in the parallel computation in the plane of the detector
The product of projection value on footprint function and the axis direction obtains the projection of each detector cells of plane of the detector
Value.
In one example, the third module includes multiple Unit the 4th, the number and the spy of Unit the 4th
The number for surveying detector cells in the plane of device is identical；Each Unit the 4th includes:
First obtains subelement, for obtaining the three-dimensional voxel of each coordinate position in the 3-D image in detector
Drop shadow spread in horizontal direction；
Subelement is searched, for searching contributive drop shadow spreads all to detector cells, the contributive projection
Range includes the horizontal coordinate value of the detector cells；
Second obtains subelement, contributes for obtaining the conduct of three-dimensional voxel corresponding to all contributive drop shadow spreads
Three-dimensional voxel；
First computation subunit, for utilizing the fuzzy footprint function in the horizontal direction of the contributive three-dimensional voxel
The projection value of the detector cells is calculated with the product of the projection value in axis direction.
In one example, described device further include:
Second obtains module, for obtaining the measured value of each detector cells on detector plane；
4th computing module is fuzzy in the axis direction for according to the fuzzy footprint function in the horizontal direction
Each coordinate bit of the measured value of each detector cells in the 3-D image on footprint function and the detector plane
Set the back projection value of each coordinate position in 3-D image described in parallel computation.
In one example, the 4th computing module includes that multiple second computation subunits and multiple thirds calculate son list
The number of member, second subelement is identical as the number of three-dimensional voxel on the cross section of the 3-D image, the third meter
Operator unit is identical as the number of the three-dimensional voxel of the 3-D image；
Second computation subunit, for according to the fuzzy footprint function and the detector in the horizontal direction
In plane the measured value of each detector cells on the cross section of the 3-D image 3-D image described in parallel computation three
Tie up voxel projection value, the fuzzy footprint function that the projection value of the three-dimensional voxel of the 3-D image is horizontally oriented with it is described
Sum of the product of the measured value of detector plane on the three-dimensional voxel of the cross section of the 3-D image；
The third subelement, for according in the axis direction fuzzy footprint function and the cross section on it is three-dimensional
The projection value of voxel each coordinate position in the 3-D image described in each coordinate position parallel computation of the 3-D image
Projection value, the projection value are the projection value of three-dimensional voxel and the fuzzy footprint function in the axis direction on the cross section
Sum of the product in the coordinate position of the 3-D image.
Fig. 5 is a kind of GPU architecture schematic diagram provided in an embodiment of the present invention, and GPU shown in fig. 5 is applied to more GPU's
In system, each GPU is identical as GPU shown in fig. 5 in the system, comprising:
First division module 501 puts down projection angle according to the number of the GPU for the number of GPU in acquisition system
Multiple projection angle ranges are divided into, the number of the projection angle range is identical as GPU number.
First obtains module 502, for obtaining the device numbering of the GPU, belongs to institute according to device numbering acquisition
State the projection angle range of GPU.
Second obtains module 503, for obtaining the pixel value of each coordinate position in 3-D image.
First computing module 504, for the method according to the pixel value according to forward projection described in embodiment one
The projection value of the planar detector unit of detector under projection angle range described in parallel computation.
In one example, the GPU further include:
Second division module, for the abscissa of the 3-D image or ordinate to be averaged according to the number of the GPU
Multiple coordinate ranges are divided into, the number of the coordinate range and the number of GPU are identical；
Third obtains module, for obtaining the device numbering of the GPU, is belonged to according to device numbering acquisition described
The coordinate range of GPU；
4th obtains module, the measured value of each detector cells in the plane for obtaining detector；
Second computing module, for according to belonging to all measured values of the coordinate range according to reversely throwing in embodiment one
Under coordinate range described in method parallel computation described in shadow in the 3-D image three-dimensional voxel projection value.
Fig. 6 is the projection system architecture schematic diagram provided in an embodiment of the present invention based on separable footprint function technology, packet
It includes:
At least two GPU shown in fig. 5.
Example IV
Fig. 7 is the projection arrangement hardware configuration signal provided in an embodiment of the present invention based on separable footprint function technology
Figure, comprising:
Memory 701, and the processor 702 being connected with the memory 701；
The memory 701 is configured in store instruction, and the processor 702 is configured in from the memory 701
Described instruction is called to execute, described instruction includes:
Obtain the pixel value of the three-dimensional voxel of each coordinate position in 3-D image；
In the horizontal direction that three-dimensional voxel described in parallel computation projects to detector on the cross section of the 3-D image
Fuzzy footprint function, the horizontal direction is projecting direction of the cross section in the detector of the 3-D image；
According to the throwing in the axis direction of detector described in pixel value parallel computation in the axis direction of the detector
Shadow value, the projection value in the axis direction are the products of fuzzy the footprint function and the pixel value of the axis direction of the detector
In the sum of the axis direction of the 3-D image；
Fuzzy footprint function and the axis direction in the plane of the detector in horizontal direction described in parallel computation
On projection value product obtain the detector each detector cells of plane projection value.
In one example, the instruction of the projection value for calculating each detector cells includes:
Obtain projection model of the three-dimensional voxel of each coordinate position in the 3-D image in the horizontal direction of detector
It encloses；
Contributive drop shadow spreads all to detector cells are searched, the contributive drop shadow spread includes the detection
The horizontal coordinate value of device unit；
Three-dimensional voxel corresponding to all contributive drop shadow spreads is obtained as contributive three-dimensional voxel；
Utilize the projection value on the fuzzy footprint function and axis direction in the horizontal direction of the contributive three-dimensional voxel
Product obtain the projection values of the detector cells.
In one example, described instruction further include:
Obtain the measured value of each detector cells on detector plane；
Fuzzy footprint function and the spy according to the fuzzy footprint function in the horizontal direction, in the axis direction
The measured value of each detector cells is on each coordinate position of the 3-D image three described in parallel computation in survey device plane
Tie up the back projection value of each coordinate position in image.
In the above example, the measured value according to detector plane is on each coordinate position of the 3-D image
The instruction of the back projection value of each coordinate position includes: in 3-D image described in parallel computation
According to each detector cells on the fuzzy footprint function and the detector plane in the horizontal direction
The projection value of measured value three-dimensional voxel of 3-D image described in parallel computation on the cross section of the 3-D image, the three-dimensional
The fuzzy footprint function that the projection value of the three-dimensional voxel of image is horizontally oriented multiplies with the measured value of the detector plane
Sum of the product on the three-dimensional voxel of the cross section of the 3-D image；
According to the projection value of the fuzzy footprint function in the axis direction and three-dimensional voxel on the cross section described
The projection value of each coordinate position, the projection value are in 3-D image described in each coordinate position parallel computation of 3-D image
The product of the projection value of three-dimensional voxel and the fuzzy footprint function in the axis direction is in the 3-D image on the cross section
Coordinate position sum.
Wherein, the fuzzy footprint function in the horizontal direction is trapezoidal function or rectangular function.
Wherein, the fuzzy footprint function in the axis direction is trapezoidal function or rectangular function.
Fig. 8 is a kind of GPU hardware structural schematic diagram provided in an embodiment of the present invention, and GPU shown in Fig. 8 is applied to more GPU
System in, each GPU is identical as GPU shown in Fig. 8 in the system, comprising:
Memory 801, and the processor 802 being connected with the memory 801；
The memory 801 is configured in store instruction, and the processor 802 is configured in from the memory 801
Call instruction executes, and described instruction includes:
Projection angle is divided equally into multiple projection angles according to the number of the GPU by the number of GPU in acquisition system
Range, the number of the projection angle range are identical as GPU number；
The device numbering for obtaining the GPU obtains the projection angle range for belonging to the GPU according to the device numbering；
Obtain the pixel value of each coordinate position in 3-D image；
According to projection angle model described in method parallel computation of the pixel value according to forward projection described in embodiment one
Enclose the projection value of the planar detector unit of lower detector.
In one example, described instruction further include:
The abscissa of the 3-D image or ordinate are divided equally into multiple coordinate models according to the number of the GPU
It encloses, the number of the coordinate range and the number of GPU are identical；
The device numbering for obtaining the GPU obtains the coordinate range for belonging to the GPU according to the device numbering；
Obtain the measured value of each detector cells in the plane of detector；
It is parallel according to the method for back projection described in embodiment one according to all measured values for belonging to the coordinate range
Calculate the back projection value of three-dimensional voxel in the 3-D image under the coordinate range.
Fig. 9 is the optical projection system hardware configuration signal provided in an embodiment of the present invention based on separable footprint function technology
Figure, comprising:
At least two GPU shown in Fig. 8.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, without departing from the principle of the present invention, it can also make several improvements and retouch, these improvements and modifications are also answered
It is considered as protection scope of the present invention.
Claims (15)
1. a kind of projecting method based on separable footprint function technology, which is characterized in that the described method includes:
Obtain the pixel value of the three-dimensional voxel of each coordinate position in 3-D image；
The water that the three-dimensional voxel projects to detector is calculated using multiple thread parallels on the cross section of the 3-D image
Square upward fuzzy footprint function, the horizontal direction is the cross section of the 3-D image in the projection side of the detector
To；
Calculate the axis side of the detector using multiple thread parallels in the axis direction of the detector according to the pixel value
Upward projection value, the projection value in the axis direction be the axis direction of the detector fuzzy footprint function and the pixel
Sum of the product of value in the axis direction of the 3-D image；
Fuzzy footprint function and institute in the horizontal direction are calculated using multiple thread parallels in the plane of the detector
State the projection value in axis direction product obtain the detector each detector cells of plane projection value.
2. the method according to claim 1, wherein the projection value for calculating each detector cells includes:
Obtain drop shadow spread of the three-dimensional voxel of each coordinate position in the 3-D image in the horizontal direction of detector；
Contributive drop shadow spreads all to detector cells are searched, the contributive drop shadow spread includes the detector list
The horizontal coordinate value of member；
Three-dimensional voxel corresponding to all contributive drop shadow spreads is obtained as contributive three-dimensional voxel；
Utilize multiplying for the projection value on the fuzzy footprint function and axis direction in the horizontal direction of the contributive three-dimensional voxel
Product obtains the projection value of the detector cells.
3. method described in -2 any one according to claim 1, which is characterized in that the method also includes:
Obtain the measured value of each detector cells on detector plane；
Fuzzy footprint function and the detector according to the fuzzy footprint function in the horizontal direction, in the axis direction
The measured value of each detector cells three-dimensional figure described in parallel computation on each coordinate position of the 3-D image in plane
The back projection value of each coordinate position as in.
4. according to the method described in claim 3, it is characterized in that, the measured value according to detector plane is in the three-dimensional
The back projection value of each coordinate position includes: in 3-D image described in parallel computation on each coordinate position of image
According to the measurement of each detector cells on the fuzzy footprint function and the detector plane in the horizontal direction
It is worth the projection value of the three-dimensional voxel of 3-D image described in parallel computation on the cross section of the 3-D image, the 3-D image
The fuzzy footprint function that is horizontally oriented of projection value of three-dimensional voxel and the product of measured value of the detector plane exist
Sum on the three-dimensional voxel of the cross section of the 3-D image；
According to the projection value of the fuzzy footprint function in the axis direction and three-dimensional voxel on the cross section in the three-dimensional
The projection value of each coordinate position in 3-D image described in each coordinate position parallel computation of image, the projection value is described
Seat of the product of the projection value of three-dimensional voxel and the fuzzy footprint function in the axis direction in the 3-D image on cross section
The sum of cursor position.
5. the method according to claim 1, wherein
Fuzzy footprint function in the horizontal direction is trapezoidal function or rectangular function.
6. the method according to claim 1, wherein
Fuzzy footprint function in the axis direction is trapezoidal function or rectangular function.
7. a kind of projecting method that more GPU are realized, which is characterized in that the described method includes:
Projection angle is divided equally into multiple projection angles according to the number of the GPU by the number of GPU in GPU acquisition system
Range, the number of the projection angle range are identical as GPU number；
GPU obtains the device numbering of the GPU, and the projection angle range for belonging to the GPU is obtained according to the device numbering；
GPU obtains the pixel value of each coordinate position in 3-D image；
GPU is according to the pixel value according to projection angle range described in claim 1-2 any one the method parallel computation
The projection value of the planar detector unit of lower detector.
8. the method according to the description of claim 7 is characterized in that the method also includes:
The abscissa of the 3-D image or ordinate are divided equally into multiple coordinate ranges according to the number of the GPU by GPU,
The number of the coordinate range and the number of GPU are identical；
GPU obtains the device numbering of the GPU, and the coordinate range for belonging to the GPU is obtained according to the device numbering；
GPU obtains the measured value of each detector cells in the plane of detector；
GPU is counted according to all measured values for belonging to the coordinate range according to claim 3-4 any one the method parallel
Calculate the back projection value of three-dimensional voxel in the 3-D image under the coordinate range.
9. a kind of projection arrangement based on separable footprint function technology, which is characterized in that described device includes:
First obtains module, for obtaining the pixel value of the three-dimensional voxel of each coordinate position in 3-D image；
First computing module, for utilizing multiple thread parallels to calculate the three-dimensional voxel on the cross section of the 3-D image
The fuzzy footprint function in the horizontal direction of detector is projected to, the horizontal direction is the cross section of the 3-D image in institute
State the projecting direction of detector；
Second computing module, for being calculated in the axis direction of the detector using multiple thread parallels according to the pixel value
Projection value in the axis direction of the detector, the projection value in the axis direction are the fuzzy foots of the axis direction of the detector
The product of trace function and the pixel value the 3-D image axis direction and；
Third computing module, for utilizing multiple thread parallels to calculate in the horizontal direction in the plane of the detector
The product of projection value on fuzzy footprint function and the axis direction obtains each detector cells of plane of the detector
Projection value.
10. device according to claim 9, which is characterized in that
The third computing module includes multiple Unit the 4th, is visited in the number and the plane of the detector of Unit the 4th
The number for surveying device unit is identical；Each Unit the 4th includes:
First obtains subelement, for obtaining level of the three-dimensional voxel in detector of each coordinate position in the 3-D image
Drop shadow spread on direction；
Subelement is searched, for searching contributive drop shadow spreads all to detector cells, the contributive drop shadow spread
Horizontal coordinate value including the detector cells；
Second obtains subelement, for obtaining three-dimensional voxel corresponding to all contributive drop shadow spreads as contributive three
Tie up voxel；
First computation subunit, for utilizing the fuzzy footprint function and axis in the horizontal direction of the contributive three-dimensional voxel
The product of projection value on direction calculates the projection value of the detector cells.
11. according to device described in claim 9-10 any one, which is characterized in that described device further include:
Second obtains module, for obtaining the measured value of each detector cells on detector plane；
4th computing module, for the fuzzy footprint according to the fuzzy footprint function in the horizontal direction, in the axis direction
The measured value of each detector cells is on each coordinate position of the 3-D image on function and the detector plane
The back projection value of each coordinate position in 3-D image described in parallel computation.
12. device according to claim 11, which is characterized in that
4th computing module includes multiple second computation subunits and multiple third computation subunits, and described second calculates son
The number of unit is identical as the number of three-dimensional voxel on the cross section of the 3-D image, the third computation subunit with it is described
The number of the three-dimensional voxel of 3-D image is identical；
Second computation subunit, for according to the fuzzy footprint function and the detector plane in the horizontal direction
The said three-dimensional body of measured value 3-D image described in parallel computation on the cross section of the 3-D image of upper each detector cells
The projection value of element, the fuzzy footprint function and the detection that the projection value of the three-dimensional voxel of the 3-D image is horizontally oriented
Sum of the product of the measured value of device plane on the three-dimensional voxel of the cross section of the 3-D image；
The third computation subunit, for according in the axis direction fuzzy footprint function and the cross section on it is three-dimensional
The projection value of voxel each coordinate position in the 3-D image described in each coordinate position parallel computation of the 3-D image
Projection value, the projection value are the projection value of three-dimensional voxel and the fuzzy footprint function in the axis direction on the cross section
Sum of the product in the coordinate position of the 3-D image.
13. a kind of GPU, which is characterized in that the GPU includes:
First division module averagely divides projection angle according to the number of the GPU for the number of GPU in acquisition system
At multiple projection angle ranges, the number of the projection angle range is identical as GPU number；
First obtains module, for obtaining the device numbering of the GPU, belongs to the GPU's according to device numbering acquisition
Projection angle range；
Second obtains module, for obtaining the pixel value of each coordinate position in 3-D image；
First computing module, for according to the pixel value according to method parallel computation described in claim 1-2 any one
The projection value of the planar detector unit of detector under the projection angle range.
14. GPU according to claim 13, which is characterized in that the GPU further include:
Second division module, for averagely being divided the abscissa of the 3-D image or ordinate according to the number of the GPU
At multiple coordinate ranges, the number of the coordinate range and the number of GPU are identical；
Third obtains module, for obtaining the device numbering of the GPU, belongs to the GPU's according to device numbering acquisition
Coordinate range；
4th obtains module, the measured value of each detector cells in the plane for obtaining detector；
Second computing module belongs to all measured values of the coordinate range according to claim 3-4 any one for basis
Under coordinate range described in the method parallel computation in the 3-D image three-dimensional voxel projection value.
15. a kind of optical projection system based on separable footprint function technology, which is characterized in that the system comprises:
GPU described at least two claim 13-14 any one.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610566245.XA CN106204679B (en) | 2016-07-18 | 2016-07-18 | Projecting method, apparatus and system based on separable footprint function technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610566245.XA CN106204679B (en) | 2016-07-18 | 2016-07-18 | Projecting method, apparatus and system based on separable footprint function technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106204679A CN106204679A (en) | 2016-12-07 |
| CN106204679B true CN106204679B (en) | 2019-04-09 |
Family
ID=57493927
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610566245.XA Expired - Fee Related CN106204679B (en) | 2016-07-18 | 2016-07-18 | Projecting method, apparatus and system based on separable footprint function technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106204679B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112215953B (en) * | 2020-11-10 | 2023-11-17 | 中国科学院高能物理研究所 | Image reconstruction method and device and electronic equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101370430A (en) * | 2006-01-24 | 2009-02-18 | 株式会社岛津制作所 | X-ray imager |
| CN102456227B (en) * | 2010-10-28 | 2015-05-27 | 清华大学 | Reconstruction method and device for CT (computerized tomography) image |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7227982B2 (en) * | 2002-04-15 | 2007-06-05 | General Electric Company | Three-dimensional reprojection and backprojection methods and algorithms for implementation thereof |
| US8913805B2 (en) * | 2010-08-30 | 2014-12-16 | The Regents Of The University Of Michigan | Three-dimensional forward and back projection methods |
-
2016
- 2016-07-18 CN CN201610566245.XA patent/CN106204679B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101370430A (en) * | 2006-01-24 | 2009-02-18 | 株式会社岛津制作所 | X-ray imager |
| CN102456227B (en) * | 2010-10-28 | 2015-05-27 | 清华大学 | Reconstruction method and device for CT (computerized tomography) image |
Non-Patent Citations (1)
| Title |
|---|
| GPU Acceleration of 3D Forward and Backward Projection Using Separable Footprints for X-ray CT Image Reconstruction;Meng Wu 等;《3rd Workshop on High Performance Image Reconstruction》;20111231;第58-59页第Ⅲ部分 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106204679A (en) | 2016-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8000513B2 (en) | System and method for 3D time of flight PET forward projection based on an exact axial inverse rebinning relation in fourier space | |
| CN107111867B (en) | Multi-modality imaging system and method | |
| US7417231B2 (en) | Fourier re-binning of time-of-flight positron emission tomography data | |
| US6339652B1 (en) | Source-assisted attenuation correction for emission computed tomography | |
| CN106466188B (en) | System and method for emission tomography quantification | |
| US9645261B2 (en) | Normalization coefficients in PET continuous bed motion acquisition | |
| CN103315760B (en) | Systems and methods for attenuation compensation in nuclear medicine imaging based on emission data | |
| CN100587391C (en) | Projection rotating center measurement method suitable for 2D-CT scanning system | |
| US9058658B2 (en) | Methods and devices for locating object in CT imaging | |
| CN101911119B (en) | Image registration alignment metric | |
| CN104567758B (en) | Stereo imaging system and its method | |
| CN107110988A (en) | Pet detector timing alignment | |
| CN101646957B (en) | Isotropic resolution image reconstruction | |
| CN107106109A (en) | Computed tomograph scanner system | |
| CN100573580C (en) | Multiple focus acquisition | |
| CN103430048A (en) | Positron CT device and reconstruction method | |
| Feng et al. | Use of three-dimensional Gaussian interpolation in the projector/backprojector pair of iterative reconstruction for compensation of known rigid-body motion in SPECT | |
| CN106204679B (en) | Projecting method, apparatus and system based on separable footprint function technology | |
| TW201616448A (en) | Projection method of three-dimensional imaging | |
| Białas et al. | System response kernel calculation for list-mode reconstruction in strip PET detector | |
| Grangeat et al. | Cone-beam spelt with a tilted detector | |
| Santos | Preconditioning conjugate gradient with symmetric algebraic reconstruction technique (ART) in computerized tomography | |
| Liu | Improve industrial cone-beam computed tomography by integrating prior information | |
| JPWO2020059135A1 (en) | Data processing methods, programs, data processing equipment and positron emission tomographic imaging equipment | |
| US9279887B2 (en) | Self-similar, transversely shift-invariant point response function in single photon emission computed tomography |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190409 Termination date: 20210718 |