CN110458950A - A kind of method for reconstructing three-dimensional model, mobile terminal, storage medium and electronic equipment - Google Patents
A kind of method for reconstructing three-dimensional model, mobile terminal, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN110458950A CN110458950A CN201910750239.3A CN201910750239A CN110458950A CN 110458950 A CN110458950 A CN 110458950A CN 201910750239 A CN201910750239 A CN 201910750239A CN 110458950 A CN110458950 A CN 110458950A
- Authority
- CN
- China
- Prior art keywords
- voxel
- boundary
- intersection point
- dimensional surface
- gpu
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/20—Processor architectures; Processor configuration, e.g. pipelining
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
- G06T11/005—Specific pre-processing for tomographic reconstruction, e.g. calibration, source positioning, rebinning, scatter correction, retrospective gating
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
Abstract
A kind of method for reconstructing three-dimensional model, mobile terminal, storage medium and electronic equipment, this method are applied to mobile terminal, comprising: obtain two-dimensional surface tomographic sequence；Adjacent two two-dimensional surface faultage images are extracted, form one layer；Multiple voxels are divided by every layer；All voxels are sent to GPU；GPU is used to calculate the seamed edge of boundary voxel and the intersection point of default contour surface in the case where voxel is boundary voxel；All intersection points that GPU is sent are received, and by the intersection point line in same boundary voxel, construct polygonal patch corresponding with the boundary voxel；Layernumber information based on all boundary voxels, in the location coordinate information of this layer, combines all polygonal patch to obtain threedimensional model with it.Method provided by the invention can construct threedimensional model based on two-dimensional surface faultage image, improve the accuracy and efficiency of image procossing, reduce image procossing cost, be suitable for mobile terminal, improve universality.
Description
Technical field
The present invention relates to technical field of medical image processing more particularly to a kind of method for reconstructing three-dimensional model, mobile terminal,
Storage medium and electronic equipment.
Background technique
The development of medical image has benefited from the discovery of X-ray at the end of the 19th century first, the research of subsequent medical imaging field at
For the focus of numerous scientific workers research, new research achievement and invention come one after another.In recent years, new medical imaging techniques
Constantly improvement upgrading, the derived techniques such as CT, CR, DR, MRI, US, PET, SPECT are gradually developed and are examined in medical conditions
It is widely applied in disconnected, greatly improves the correct diagnosis of disease, also established Medical Imaging in medical diagnosis
Irreplaceable role and status.
Research of the foreign countries in the three-dimensional reconstruction field of medical image starts from nineteen seventies.These developed countries
Research starting in this field is relatively early and relatively gos deep into, and has been achieved in the three-dimensional reconstruction field of medical image quite significant
Scientific achievement.They have put into very more manpower and material resources to carry out the research in terms of medical image three-dimensional reconstruction, and have
There are many superior resources.It is mainly reflected in two aspects of medical image software platform system and visual human.Currently, in the world more
Popular magic magiscan is relatively more.For example, VolVis system, 3D Slicer system and performance are very superior
Business visualization system Volulne Pro and Vitrea2 etc..In addition, also having emerged in large numbers one in three-dimensional visualization field in the world
A little outstanding development kits: VTK (visualization Toolkit) open source visual post process, ITK (Insight
Segmentation and Registration Toolkit) Medical Image Processing kit.There are also the U.S. to start in 1989
The human visualization planning item of starting.Researcher uses MRI and CT to do the scanning of human body, and utilizes the weight of computer
Structure technology builds virtual human body to obtain data set, so that virtual reality to have been introduced to the application category of medical domain.
The nineties of eighties of last century or so, the country just started to spread out three-dimensional reconstruction medicine neighborhood using with
Research, starting is than later relatively for foreign countries.As domestic more and more scientific research scholars are to reconstruction of medical images technology
Research and discussion, with regard at present for, we also achieve certain research achievement.For example, the scientific research in the case where Tian Jie is led is small
The Medical Image Processing of group research and development and the developing instrument MITK (Medical Imaging Toolkit) of analysis and the Chinese Academy of Sciences push away
3DMed visualization system out is considered as the software of the first systematization of China's medical visualization software view, can be realized figure
As information control, store and have access to, three-dimensional reconstruction, and simple model analysis with the sequence of operations such as interact.Then east
Soft (Neusoft) group has issued its PACS/RIS plateform system voluntarily studied, and platform covering is examined from point calls out the numbers to evolution
Take medicine almost entire medical cycle information transmission and management, wherein being the pipe of medical image information in terms of core the most
Reason, although the platform can not directly carry out three-dimensional visualization processing, it provides a variety of including three-dimensional reconstruction software
The interface of the poster processing soft plays certain impetus for the development of domestic medical visualization software.Zhejiang University is autonomous
The Med Vis Medical Image Three Dimensional Visualization System of research and development, important feature are the reality with good interactivity and drafting
It is fast to draw speed, but is realized on high performance graphics computer by Shi Xing, higher to hardware requirement.The country is also 2001
It has been put forward for the first time Performance Computers from Digitized Virtual Human project when year, after 2 years, first domestic successful structure of women virtual human dataset
At.The acquisition of human body slice of data also in what is completed under particular medical device, be all in data set have it is high-precision
Data, the subsequent equipment using the data set should have biggish graphics calculations ability.
With the development of science and technology, successively to produce many outstanding medical image three-dimensional reconstructions soft for many medical device corporations
Part, such as the Mimics medical image control of Simpleware 3D rendering processing and finite element analysis software, Materialise company
System processed, 3D doctor etc..And many scholars also are attempting to develop self-designed medical image three-dimensional reconstruction system.
But the prior art have the defects that it is as follows:
1, existing three-dimensional image system carries out three-dimensional reconstruction just for the appearance of object, lacks to object
The stereo reconstruction of inherent anatomical structure is not suitable for medical consultations needs.
2, at high cost.Existing numerous medical image three-dimensional reconstruction systems require reconstruction precision when, in addition to rebuild calculate
The accuracy and reasonability of method necessarily use the hardware device of superior performance.Such as Mimics, 3DMed medical image control system
System is all that simultaneously O&M is designed and developed by company, and the expense of the equipment is apparently not that each individual is afforded.
3, lack universality.Computer end is all developed and be suitable for most three-dimensional reconstruction system, needs electronic material
Support can just reconstruct the threedimensional model at a certain position.Firstly, few people carry computer.Secondly, existing hospital is logical
Patient's picture material is often only provided.Patient can not take electronic material, also will be unable to using the system, this makes the demand of patient
It cannot meet in time.
Summary of the invention
(1) goal of the invention
The object of the present invention is to provide a kind of method for reconstructing three-dimensional model, mobile terminal, storage medium and electronic equipments, should
Method for reconstructing three-dimensional model is applied on mobile terminal, can construct threedimensional model based on two-dimensional surface faultage image, pass through shifting
CPU and GPU in dynamic terminal cooperate, and improve the accuracy and efficiency of image procossing, reduce image procossing cost, be suitable for
Mobile terminal improves universality.
(2) technical solution
To solve the above problems, the first aspect of the present invention provides a kind of method for reconstructing three-dimensional model, it is applied to movement
Terminal, comprising: obtain two-dimensional surface tomographic sequence；Adjacent two two-dimensional surface faultage images are extracted, form one layer；
Multiple voxels are divided by every layer；The voxel have layernumber information and its this layer location coordinate information；It will be all described
Voxel is sent to GPU；The GPU is used to calculate the seamed edge of the boundary voxel in the case where the voxel is boundary voxel
With the intersection point of default contour surface；All intersection points that GPU is sent are received, and by the intersection point line in same boundary voxel,
Construct polygonal patch corresponding with the boundary voxel；Layernumber information based on all boundary voxels is with it in this layer
Location coordinate information combines all polygonal patch to obtain threedimensional model.
According to another aspect of the present invention, a kind of mobile terminal for reconstructing three-dimensional model is provided, including CPU and
GPU；The CPU includes: image collection module, for obtaining two-dimensional surface tomographic sequence；Layer building module, for extracting
Adjacent two two-dimensional surface faultage images form one layer；Voxel constructs module, for being divided into multiple voxels for every layer；It is described
Voxel have layernumber information and its this layer location coordinate information；Voxel sending module, for sending out all voxels
It send to GPU；Polygonal patch constructs module, for receiving all intersection points of GPU transmission, and will be in same boundary voxel
With all intersection point lines, polygonal patch corresponding with the boundary voxel is constructed；Polygonal patch composite module, for being based on institute
The layernumber information for the boundary voxel having, in the location coordinate information of this layer, combines all polygonal patch with it
Obtain threedimensional model；The GPU includes: intersecting point coordinate computing module, for counting in the case where the voxel is boundary voxel
Calculate the seamed edge of the boundary voxel and the intersection point of default contour surface.
It in one embodiment, further include intersection point rendering module, for being rendered based on default spatial cue antinode.
According to another aspect of the invention, a kind of storage medium is provided, is stored with computer program on the storage medium,
The step of method for reconstructing three-dimensional model described above is realized when described program is executed by processor.
According to another aspect of the invention, a kind of electronic equipment is provided, including memory, processor and is stored in described deposit
On reservoir and the computer program that can run on the processor, the processor are realized described above when executing described program
The step of method for reconstructing three-dimensional model.
(3) beneficial effect
Above-mentioned technical proposal of the invention has following beneficial technical effect:
1, method for reconstructing three-dimensional model provided by the invention can obtain three-dimensional mould based on two-dimensional surface tomographic image reconstructing
Type is cooperated by CPU in mobile terminal and GPU, improves the accuracy and efficiency of image procossing, reduce image procossing at
This.
2, method for reconstructing three-dimensional model provided by the invention can be applicable on this portable device of mobile terminal, be convenient for
It carries, universality with higher.
3, provided by the present invention for the mobile terminal of reconstructing three-dimensional model, by using threedimensional model provided by the invention
Method for reconstructing improves the accuracy and efficiency of image procossing, reduces image procossing cost, while easy to carry, have compared with
High universality.
Detailed description of the invention
Fig. 1 is provided in an embodiment of the present invention to the pretreated method flow diagram of two-dimensional surface faultage image；
Fig. 2 is the flow chart for the method for reconstructing three-dimensional model that one embodiment of the invention provides；
Fig. 3 is the intersection figure of voxel and default contour surface in the default look-up table of one embodiment of the invention offer；
Fig. 4 be another embodiment of the present invention provides method for reconstructing three-dimensional model flow chart；
Fig. 5 is the module composition schematic diagram of the mobile terminal provided in an embodiment of the present invention for reconstructing three-dimensional model.
Appended drawing reference:
100, CPU, 101, image collection module, 102, layer building module, 103, voxel building module, 104, voxel transmission
Module, 105, polygonal patch building module, 106, polygonal patch composite module, 200, GPU, 201, boundary voxel judgement mould
Block, 2011, density value comparing unit, 2012, voxel lookup of state unit, 202, intersecting point coordinate computing module, 203, rendering mould
Block, 2031, radiance computing unit, 2032, RGB component conversion unit, 2033, coloring units.
Specific embodiment
In order to make the objectives, technical solutions and advantages of the present invention clearer, With reference to embodiment and join
According to attached drawing, the present invention is described in more detail.It should be understood that these descriptions are merely illustrative, and it is not intended to limit this hair
Bright range.In addition, in the following description, descriptions of well-known structures and technologies are omitted, to avoid this is unnecessarily obscured
The concept of invention.
Before describing in detail to the embodiment of the present invention, term explanation is first carried out, under occurring hereinafter and in figure
Column term concrete meaning is as follows:
CPU (Centre Processing Unit): central processing unit；
GPU (Graphic Processing Unit): graphics processor；
Two-dimensional surface tomographic sequence: refer to the series sequentially continuously acquired in different time, different direction to target
Two-dimensional surface faultage image.
Local low-rank matrix: there is the matrix of low-rank characteristic in the regional area of image.
Due to the image that two-dimensional surface faultage image is medical domain, wherein the black of the tissue in image or white are come table
Show.When two-dimensional surface faultage image is shot or is scanned, it is possible to have unclear problem, and then influence subsequent builds
The accuracy of threedimensional model, therefore, in a preferred embodiment, before constructing threedimensional model, first to each Zhang Erwei
Plane fault image is pre-processed, so that two-dimensional surface faultage image becomes apparent from.
Fig. 1 is provided in an embodiment of the present invention to the pretreated method flow diagram of two-dimensional surface faultage image.
Optionally, in the present embodiment, two-dimensional surface faultage image can pass through the mobile terminals such as mobile phone or plate by user
Shooting is obtained by the software scans in mobile phone, is sent in mobile phone after can also being scanned by scanning device.
Fig. 1 is please referred to, in embodiments of the present invention, which includes step S101-S103.
S101 formats two-dimensional surface faultage image, obtains the two-dimensional surface faultage image of preset format.
Usual hospital only provides the non-electronic version two-dimensional surface faultage image of patient, even if being to provide electronic edition material generally also
It is with the storage of dicom standard format.Therefore subsequent operation for convenience, in this step, first by every two-dimensional surface tomography
Image formats, and is unified format.Optionally, one of preset format tif, png or jpg.Due to png lattice
Formula is a kind of bitmap piece shape format of lossless compression, has the advantages such as compression ratio is high, and generation file size is small, therefore in this implementation
Png format is preferably converted thereof into example.
Optionally, preset format may be the CT/MRI/Micro CT/Micro of the reading DICOM format of interactive
MRI/ INDUSTRIAL CT IMAGE or the Common image formats BMP/TIFF of non-DICOM of sequence etc..
Requirement for input picture is: the tomography thickness of input image sequence gets over Bao Yuehao, i.e. the number of plies is The more the better.
Preferably, shooting angle needs to shoot perpendicular to the front of image as far as possible.
S102 carries out enhancing processing to the two-dimensional surface faultage image of preset format, including denoising and resolution ratio increase
By force.
Wherein, step S102 includes step S1021-S1026.
S1021 is split the two-dimensional surface faultage image of preset format, obtains multiple images block.For the ease of meter
It calculates, block is often based in image procossing and carries out operation to entire image.
S1022 calculates the local low-rank matrix M1 of current image block.
Wherein, local low-rank matrix refers in the matrix that image local has low-rank characteristic, reflects image local pixel
Between correlation.
Specifically, the calculation formula of local low-rank matrix M1 are as follows:
The part M1 low-rank matrix is one and is made of four submatrixs, whereinIt is the high quality graphic part rebuild in generation
Area pixel constitute, wherein think in denoising noise position pixel be it is unknown, resolution ratio enhancing processing in, will differentiate
Rate enhancing can be used as interpolation and be handled, and the pixel being inserted into is similarly unknown.It isIn each pixel adjacent pixels
The matrix that the row vector of (eight directions) composition is formed.BecauseWithIn there is the partial pixel to be unknown (such as noise position
The correct pixel value set, is inserted into the pixel value of pixel), these unknown pixel values are our places to be solved.SimilarlyIt is that low-quality image (is noisy image in Denoising Algorithm, is low resolution figure in resolution ratio enhancing processing
Picture) in two matrixes constituting of pixel, they collectively constitute a local low-rank matrix.(part is relative to non local
For the position of currently processed image block or pixel)
S1023 searches similar image block similar with current image block in all image blocks.
Specifically, the method for searching similar image block is: calculate separately current image block and other image blocks it is European away from
From Euclidean distance represents the similarity degree of two image blocks, and apart from smaller, two image blocks are more similar.By every other figure
Picture block sorts by similarity degree, n image block before taking, this n image block is similar image block similar with current image block.
S1024 is based on similar image block, the non-local low rank matrix M2 of current image block is calculated.
Specifically, it is based on similar image block, splices to obtain non-local low rank matrix by the vector form of similar image block
M2。
S1025 is based on part low-rank matrix M1 and non-local low rank matrix M2, rebuilds to current image block.
Specifically, in conjunction with local low-rank matrix M1 and non-local low rank matrix M2, image reconstruction Optimized model is established, is solved
Reconstruction model rebuilds image block, obtains the value of unknown pixel in high quality graphic matrix.
S1026 traverses all image blocks, the two-dimensional surface faultage image after being optimized.
In the present embodiment, to the method that the two-dimensional surface faultage image of preset format carries out enhancing processing, it can make an uproar to band
The denoising of two-dimensional surface faultage image can also carry out resolution ratio enhancing to low resolution two-dimensional surface faultage image, can specifically lead to
It crosses and the different above-mentioned functions of parameter realization is set.
S103 carries out Threshold segmentation processing to enhanced two-dimensional surface faultage image, and it is flat to obtain pretreated two dimension
Bedding fault image.
Fig. 2 is the flow chart for the method for reconstructing three-dimensional model that one embodiment of the invention provides.
Referring to figure 2., in an embodiment of the present invention, a kind of method for reconstructing three-dimensional model provided, including step S201-
S206。
S201 obtains two-dimensional surface tomographic sequence.
Optionally, after obtaining two-dimensional surface tomographic sequence, using step S101-S103 to two-dimensional surface tomography
Image preprocessing constructs threedimensional model using pretreated two-dimensional surface faultage image.
S202 extracts adjacent two two-dimensional surface faultage images, forms one layer.
S203 is divided into multiple voxels for every layer；Wherein, there is voxel layernumber information to believe with its position coordinates in this layer
Breath.
Wherein, being divided into multiple voxels for every layer includes:
Based on corresponding 4 pixels on two two-dimensional surface faultage images in every layer, a voxel is constructed, thus
The layer is divided into multiple voxels；4 pixels are adjacent in same two-dimensional surface faultage image and constitute square.
Specifically, based on corresponding 4 pixels on two two-dimensional surface faultage images in every layer, one individual of building
Element refers to, based in every layer wherein in a two-dimensional surface faultage image 4 pixels that are adjacent and constituting square with it is another
It opens corresponding 4 pixels in two-dimensional surface faultage image and constructs a voxel.
Optionally, voxel is cube structure.
All voxels are sent to GPU by S204；GPU is used to calculate boundary body in the case where voxel is boundary voxel
The seamed edge of element and the intersection point of default contour surface.Wherein, intersecting point coordinate is corresponding with boundary voxel, and has the level number of boundary voxel
With its this layer location coordinate information.
Wherein, GPU determines whether voxel is that boundary voxel includes:
The density value on 8 vertex of voxel and the threshold value of default contour surface are compared by GPU respectively, obtain the rope of voxel
Draw value, the index value is between 0-255.
The state of voxel corresponding with index value is searched in default look-up table.
The state of voxel includes: whether voxel is boundary voxel, and boundary voxel and the intersection point of default contour surface are located at boundary
On which seamed edge of voxel and the connection type of intersection point.
Specifically, the threshold value for presetting contour surface is set according to the density value at the corresponding position of the threedimensional model to be constructed,
By can be derived that the voxel is boundary for the threshold value comparison of the density value on 8 vertex of voxel and the default contour surface
Voxel finds out the equivalent point that gray value on the seamed edge of boundary voxel is the threshold value if it is boundary voxel, this equivalent point is exactly
The intersection point of the boundary voxel and default contour surface.
Fig. 3 is the intersection figure of voxel and default contour surface in the default look-up table of one embodiment of the invention offer.
The building process of default look-up table is simply introduced below:
The density value on 8 vertex in cube is compared with the size relation of the threshold value of default contour surface.If 8 tops
The density value of point be both greater than or less than default contour surface threshold value, illustrate the voxel and default contour surface without intersection point, the voxel is not
It is boundary voxel.Conversely, then illustrating that the voxel and default contour surface have intersection point, and there are many different intersection modes.
If every case to be all considered as to a kind of intersection (including the case where not having intersection point), 256 kinds of states are shared,
According to symmetry, it can simplify into 15 kinds of states, as shown in Figure 3.
According to this 15 kinds of states, a default look-up table can be constructed, the length of the default look-up table is 256, is had recorded
The connection type of every kind of state intersection point, every kind of state correspond an index value, each index in 256 kinds of states under 15 classes
Value is between 0-255, that is, first state, index value 0, and second element with state index is the 1, the 256th state index
Value is 255.The purpose that index value is arranged is easy for comparing, so that search speed is faster.
When whether judge voxel is boundary voxel, it is only necessary to by the density value on 8 vertex of voxel and default contour surface
Threshold value be compared, the index value between a 0-255 can be obtained, looked into default look-up table according to index value
Which look for, so that it may determine whether the voxel is boundary voxel, can also obtain having equivalent point (i.e. body on a seamed edge of boundary voxel
The plain intersection point with default contour surface) and the equivalent point connection type.
Wherein, the intersecting point coordinate for calculating the seamed edge and the intersection point of default contour surface of boundary voxel includes:
Voxel-based state calculates the seamed edge of boundary voxel and the intersection point of default contour surface using linear interpolation method
Intersecting point coordinate.
The whole voxels obtained by all two-dimensional surface faultage images are sent to GPU by the present embodiment in batches, are carried out parallel
Processing.Process serial number is marked with the level number Z where voxel and in the position (X, Y) of this layer.It should be noted that the application is sharp
With GPU in the calculating advantage in terms of batch processing and the independence between each voxel, parallel computation is realized, so that calculation be greatly improved
The calculated performance of method.
S205 receives all intersecting point coordinates that GPU is sent, and by the intersection point line in same boundary voxel, building with
The corresponding polygonal patch of the boundary voxel.
Optionally, polygonal patch is triangle surface.But system that invention is not limited thereto, polygonal patch are also possible to
The polygonal patch of other shapes.
S206, layernumber information based on all boundary voxels and its in the location coordinate information of this layer, will be all more
Side shape dough sheet combines to obtain threedimensional model.
Above-mentioned method for reconstructing three-dimensional model provided in an embodiment of the present invention, can be based on two-dimensional surface faultage image building three
Dimension module is cooperated by CPU and GPU, improves the accuracy and efficiency of image procossing, reduce image procossing cost；And it should
Method for reconstructing three-dimensional model can be applicable on this portable device of mobile terminal, easy to carry, universality with higher.
Fig. 4 be another embodiment of the present invention provides method for reconstructing three-dimensional model flow chart.
As shown in figure 4, in an alternative embodiment of the invention, in order to enable threedimensional model is closer to real-world object, in step
After rapid S204 obtains the seamed edge of boundary voxel and the intersection point of default contour surface, in step S205, CPU processor is receiving GPU
Before the intersection point for all boundary voxel that processor is sent, further include the steps that rendering the intersection point, and then
Threedimensional model after to rendering.Specific steps are as follows:
Step S204-1, the friendship according to the information of preset rendering, to the seamed edge and default contour surface of the boundary voxel
Point is rendered.
It should be noted that in this step, calculating advantage and each polygon facet of the GPU in terms of batch processing is utilized
Independence between piece realizes parallel computation, so that the calculated performance of algorithm be greatly improved.
Specifically, GPU is used to calculate object boundary voxel using reflectivity equation based on preset spatial cue
The radiance of the intersection point (i.e. threedimensional model surface vertices) of seamed edge and default contour surface；Specifically, GPU is believed based on preset rendering
Breath simulates actual physical condition, calculates the seamed edge of object boundary voxel and the intersection point of default contour surface using reflectivity equation
The radiance of (i.e. threedimensional model surface vertices), radiance refers to the radiation intensity of luminous energy herein.
Optionally, spatial cue includes Lighting information, observation viewpoint position information and object dimensional model surface material letter
Breath.
GPU is also used to for radiance being converted to RGB component, and is based on RGB component, seamed edge to each boundary voxel and pre-
If the intersection point of contour surface colours, the intersection point after being rendered, and sends CPU for the intersection point after rendering.
S205 step becomes, and CPU receives the intersection point all renderings for sending of GPU in batches after, and by same boundary voxel
In intersection point line, the polygonal patch after constructing corresponding with boundary voxel rendering.
S206 step becomes, and the layernumber information based on all boundary voxels, in the location coordinate information of this layer, is incited somebody to action with it
All polygonal patch by after the rendering of same boundary voxel after rendering combine, the threedimensional model after being rendered.
In the present embodiment, GPU in batches renders the seamed edge of boundary voxel and the intersection point of default contour surface, and CPU will
Line between the seamed edge of the same boundary voxel and the intersection point of default contour surface, the polygonal patch after being rendered, and to wash with watercolours
Polygonal patch after dye combines the threedimensional model after being rendered, and realizes the rendering to threedimensional model, so that threedimensional model
It is more intuitive, closer to real-world object, lay a good foundation for subsequent multiplanar reconstruction and threedimensional model measurement.
Fig. 5 is the module composition schematic diagram of the mobile terminal provided in an embodiment of the present invention for reconstructing three-dimensional model.
Referring to figure 5., the embodiment of the present invention also provides a kind of mobile terminal for reconstructing three-dimensional model, comprising:
CPU100 and GPU200.
Wherein, CPU100 includes: image collection module 101, layer building module 102, voxel building module 103, voxel hair
Send module 104, polygonal patch building module 105 and polygonal patch composite module 106.
Image collection module 101, for obtaining two-dimensional surface tomographic sequence.
Layer building module 102 forms one layer for extracting adjacent two two-dimensional surface faultage images.
Voxel constructs module 103, for being divided into multiple voxels for every layer；Wherein, voxel have layernumber information and its
The location coordinate information of this layer.
Voxel sending module 104, for all voxels to be sent to GPU200.
Polygonal patch constructs module 105, for receiving multiple intersecting point coordinates of GPU200 transmission, and by same boundary body
Intersection point line corresponding with intersecting point coordinate in element constructs polygonal patch corresponding with the boundary voxel.
Polygonal patch composite module 106, for the layernumber information based on all boundary voxels with it in the position of this layer
Coordinate information is set, combines all polygonal patch to obtain threedimensional model.
Wherein, GPU200 includes: boundary voxel determination module 201 and intersecting point coordinate computing module 202.
Boundary voxel determination module 201, for determining whether voxel is boundary voxel.
Wherein, boundary voxel determination module 201 includes: density value comparing unit 2011, voxel lookup of state unit 2012.
Density value comparing unit 2011, for distinguishing the threshold value of the density value on 8 vertex of voxel and default contour surface
It is compared, obtains the index value of voxel.
Voxel lookup of state unit 2012, for searching the state of voxel corresponding with index value in default look-up table.
The state of voxel includes: whether voxel is boundary voxel, and boundary voxel and the intersection point of default contour surface are located at boundary
On which seamed edge of voxel and the connection type of intersection point.
Intersecting point coordinate computing module 202, in the case where voxel is boundary voxel, calculate the seamed edge of boundary voxel with
The intersecting point coordinate of default contour surface.
In an embodiment of the present embodiment, GPU200 further include: rendering module 203.
Rendering module 203 includes: radiance computing unit 2031, RGB component conversion unit 2032 and coloring units 2033.
Radiance computing unit 2031, for calculating boundary body using reflectivity equation based on preset spatial cue
The radiance of the intersection point of the seamed edge and default contour surface of element, radiance refers to the radiation intensity of luminous energy herein.
Optionally, spatial cue includes Lighting information, observation viewpoint position information and threedimensional model Facing material information.
RGB component conversion unit 2032, for radiance to be converted to RGB component.
Coloring units 2033, for being based on RGB component, the intersection point of seamed edge and default contour surface to each boundary voxel
After color, the intersection point after coloring is sent to polygonal patch building module 105.
After the polygonal patch building module 105 of CPU100 is also used to receive whole renderings that the GPU processor is sent
The intersection point (data transfer path in Fig. 5 shown in dotted line in the present embodiment), and based on all boundary voxels
Layernumber information constructs the intersection point line in same boundary voxel and the boundary voxel with its location coordinate information in this layer
Polygonal patch after corresponding rendering.
The polygonal patch composite module 106 of CPU100 combines all polygonal patch after rendering after obtaining rendering
Threedimensional model.
In an embodiment of the present embodiment, CPU100 further includes image pre-processing module.
Specifically, image pre-processing module pre-processes two-dimensional surface faultage image for executing in above method embodiment
The step of, the concrete composition of the image pre-processing module can appropriate adjustment according to actual needs, be not specifically limited herein.
Mobile terminal provided in an embodiment of the present invention for reconstructing three-dimensional model, by using three-dimensional provided by the invention
Model reconstruction method improves the accuracy and efficiency of image procossing, reduces image procossing cost, while easy to carry, tool
There is higher universality.
Further embodiment of this invention additionally provides a kind of storage medium, is stored with computer program on the storage medium, should
The step of method for reconstructing three-dimensional model in above-described embodiment is realized when program is executed by processor.
Further embodiment of this invention additionally provides a kind of electronic equipment, including memory, processor and is stored in described deposit
On reservoir and the computer program that can run on the processor, the processor realize above-mentioned implementation when executing described program
In example the step of method for reconstructing three-dimensional model.
The present invention is directed to protect a kind of method for reconstructing three-dimensional model, mobile terminal, storage medium and electronic equipment, have such as
Beneficial technical effect down:
1, method for reconstructing three-dimensional model provided by the invention can construct threedimensional model based on two-dimensional surface faultage image,
Cooperated by the CPU and GPU in mobile terminal, improves the accuracy and efficiency of image procossing, reduce image procossing cost.
2, method for reconstructing three-dimensional model provided by the invention can be applicable on this portable device of mobile terminal, be convenient for
It carries, universality with higher.
3, provided by the present invention for the mobile terminal of reconstructing three-dimensional model, by using threedimensional model provided by the invention
Method for reconstructing improves the accuracy and efficiency of image procossing, reduces image procossing cost, while easy to carry, have compared with
High universality.
It should be understood that above-mentioned specific embodiment of the invention is used only for exemplary illustration or explains of the invention
Principle, but not to limit the present invention.Therefore, that is done without departing from the spirit and scope of the present invention is any
Modification, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.In addition, appended claims purport of the present invention
Covering the whole variations fallen into attached claim scope and boundary or this range and the equivalent form on boundary and is repairing
Change example.
Claims (10)
1. a kind of method for reconstructing three-dimensional model, which is characterized in that be applied to mobile terminal, comprising:
Obtain two-dimensional surface tomographic sequence；
Adjacent two two-dimensional surface faultage images are extracted, form one layer；
Multiple voxels are divided by every layer；The voxel have layernumber information and its this layer location coordinate information；
All voxels are sent to GPU；The GPU is used to calculate institute in the case where the voxel is boundary voxel
State the seamed edge of boundary voxel and the intersection point of default contour surface；
All intersection points that GPU is sent are received, and by intersection point lines all in same boundary voxel, are constructed and the boundary body
The corresponding polygonal patch of element；
Layernumber information based on all boundary voxels and its in the location coordinate information of this layer, will be all described polygon
Shape dough sheet combines to obtain threedimensional model.
2. being divided into multiple voxels the method according to claim 1, wherein described for every layer and including:
Based on corresponding 4 pixels on two two-dimensional surface faultage images in every layer, a voxel is constructed, thus
The layer is divided into multiple voxels；4 adjacent pixels constitute square in same Zhang Suoshu two-dimensional surface faultage image.
3. the method according to claim 1, wherein the GPU determines whether voxel is that boundary voxel includes:
The density value on 8 vertex of the voxel and the threshold value of the default contour surface are compared by GPU respectively, are obtained described
The index value of voxel；
The state of voxel corresponding with the index value is searched in default look-up table；The state of the voxel includes: the body
Whether element is boundary voxel, and the intersection point of the boundary voxel and the default contour surface is located at which rib of the boundary voxel
The connection type of Bian Shang and the intersection point.
4. according to the method described in claim 3, it is characterized in that, the seamed edge for calculating the boundary voxel and default equivalence
The intersection point in face includes:
The seamed edge and the default equivalence of the boundary voxel are calculated using linear interpolation method based on the state of the voxel
The intersection point in face.
5. the method according to claim 1, wherein in the seamed edge and default contour surface that calculate the boundary voxel
Intersection point after, further include being also used to carry out wash with watercolours to the intersection point of the seamed edge of the boundary voxel and default contour surface to the GPU
Dye:
The seamed edge and default contour surface of the boundary voxel are calculated using reflectivity equation based on preset spatial cue
The radiance of intersection point；The GPU processor is also used to for the radiance being converted to RGB component, and is based on the RGB component,
Each intersection point is coloured, the intersection point after being rendered；
The intersection point after receiving whole renderings that the GPU processor is sent, by the intersection point line in same boundary voxel, structure
Polygonal patch after building rendering corresponding with the boundary voxel；
Layernumber information based on the boundary voxel and its this layer location coordinate information, by the polygon facet after the rendering
Piece combines the threedimensional model after being rendered.
6. method according to any of claims 1-4, which is characterized in that obtaining two-dimensional surface tomographic sequence
Later, adjacent two two-dimensional surface faultage images are being extracted, is being formed before one layer, further includes to the two-dimensional surface tomography
The step of image preprocessing:
The two-dimensional surface faultage image is formatted, the two-dimensional surface faultage image of preset format is obtained；
Enhancing processing is carried out to the two-dimensional surface faultage image of the preset format；
Threshold segmentation processing is carried out to the enhanced two-dimensional surface faultage image, obtains pretreated two-dimensional surface tomography
Image.
7. according to the method described in claim 6, it is characterized in that, the two-dimensional surface faultage image to preset format carries out
Enhancing is handled
The two-dimensional surface faultage image of the preset format is split, multiple images block is obtained；
Calculate the local low-rank matrix of current image block；
Based on the local low-rank matrix, similar image block similar with the current image block is searched in all image blocks；
Based on the similar image block, the non-local low rank matrix of the current image block is calculated；
Based on the local low-rank matrix and non-local low rank matrix, the current image block is rebuild；
All image blocks are traversed, the enhanced two-dimensional surface faultage image is obtained.
8. a kind of mobile terminal for reconstructing three-dimensional model, which is characterized in that including CPU and GPU；
The CPU includes:
Image collection module, for obtaining two-dimensional surface tomographic sequence；
Layer building module forms one layer for extracting adjacent two two-dimensional surface faultage images；
Voxel constructs module, for being divided into multiple voxels for every layer；The voxel has layernumber information with it in the position of this layer
Set coordinate information；
Voxel sending module, for all voxels to be sent to GPU；
Polygonal patch constructs module, connects for receiving all intersection points of GPU transmission, and by the intersection point in same boundary voxel
Line constructs polygonal patch corresponding with the boundary voxel；
Polygonal patch composite module is sat with it in the position of this layer for the layernumber information based on all boundary voxels
Information is marked, combines all polygonal patch to obtain threedimensional model；
The GPU includes:
Intersecting point coordinate computing module, for calculating the seamed edge of the boundary voxel in the case where the voxel is boundary voxel
With the intersection point of default contour surface.
9. a kind of storage medium, which is characterized in that be stored with computer program on the storage medium, described program is by processor
The step of method for reconstructing three-dimensional model described in any one of claim 1-7 is realized when execution.
10. a kind of electronic equipment, which is characterized in that including memory, processor and be stored on the memory and can be in institute
The computer program run on processor is stated, the processor realizes any one of claim 1-7 institute when executing described program
The step of stating method for reconstructing three-dimensional model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910750239.3A CN110458950A (en) | 2019-08-14 | 2019-08-14 | A kind of method for reconstructing three-dimensional model, mobile terminal, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910750239.3A CN110458950A (en) | 2019-08-14 | 2019-08-14 | A kind of method for reconstructing three-dimensional model, mobile terminal, storage medium and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110458950A true CN110458950A (en) | 2019-11-15 |
Family
ID=68486522
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910750239.3A Pending CN110458950A (en) | 2019-08-14 | 2019-08-14 | A kind of method for reconstructing three-dimensional model, mobile terminal, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110458950A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111053574A (en) * | 2019-12-05 | 2020-04-24 | 深圳开立生物医疗科技股份有限公司 | Three-dimensional surface reconstruction method and device, ultrasonic equipment and storage medium |
| CN111325821A (en) * | 2020-02-18 | 2020-06-23 | Oppo广东移动通信有限公司 | Grid model processing method and device, equipment and storage medium |
| CN111583391A (en) * | 2020-04-29 | 2020-08-25 | 北京深测科技有限公司 | Object three-dimensional reconstruction method and system |
| CN112465802A (en) * | 2020-12-10 | 2021-03-09 | 上海联影医疗科技股份有限公司 | Method, system, terminal and computer readable storage medium for processing tomographic image |
| CN112767532A (en) * | 2020-12-30 | 2021-05-07 | 华东师范大学 | Ultrasonic or CT medical image three-dimensional reconstruction method based on transfer learning |
| CN112819950A (en) * | 2021-02-07 | 2021-05-18 | 清华-伯克利深圳学院筹备办公室 | Preparation method, device and equipment of brain tissue section die and storage medium |
| CN113035306A (en) * | 2021-03-17 | 2021-06-25 | 广州华端科技有限公司 | Method, system, equipment and medium for remotely browsing images |
| CN113470180A (en) * | 2021-05-25 | 2021-10-01 | 杭州思看科技有限公司 | Three-dimensional mesh reconstruction method, device, electronic device and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0884697A2 (en) * | 1997-06-13 | 1998-12-16 | Hitachi, Ltd. | Image processing method and apparatus for radiotherapy using computed tomography |
| CN101271574A (en) * | 2008-03-20 | 2008-09-24 | 华南师范大学 | Three-dimensional visualization method and device |
| CN101593345A (en) * | 2009-07-01 | 2009-12-02 | 电子科技大学 | Three-dimensional medical image display method based on the GPU acceleration |
| CN103136785A (en) * | 2013-02-01 | 2013-06-05 | 上海交通大学医学院附属第九人民医院 | Human body medical model three-dimensional visualization system used for mobile terminal and realizing method thereof |
| CN103295269A (en) * | 2013-06-26 | 2013-09-11 | 电子科技大学 | Iso-surface extraction method of volume data of electromagnetic environment |
| CN104008538A (en) * | 2014-05-12 | 2014-08-27 | 清华大学 | Super-resolution method based on single image |
| CN107895364A (en) * | 2017-10-31 | 2018-04-10 | 哈尔滨理工大学 | A kind of three-dimensional reconstruction system for the preoperative planning of virtual operation |
-
2019
- 2019-08-14 CN CN201910750239.3A patent/CN110458950A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0884697A2 (en) * | 1997-06-13 | 1998-12-16 | Hitachi, Ltd. | Image processing method and apparatus for radiotherapy using computed tomography |
| CN101271574A (en) * | 2008-03-20 | 2008-09-24 | 华南师范大学 | Three-dimensional visualization method and device |
| CN101593345A (en) * | 2009-07-01 | 2009-12-02 | 电子科技大学 | Three-dimensional medical image display method based on the GPU acceleration |
| CN103136785A (en) * | 2013-02-01 | 2013-06-05 | 上海交通大学医学院附属第九人民医院 | Human body medical model three-dimensional visualization system used for mobile terminal and realizing method thereof |
| CN103295269A (en) * | 2013-06-26 | 2013-09-11 | 电子科技大学 | Iso-surface extraction method of volume data of electromagnetic environment |
| CN104008538A (en) * | 2014-05-12 | 2014-08-27 | 清华大学 | Super-resolution method based on single image |
| CN107895364A (en) * | 2017-10-31 | 2018-04-10 | 哈尔滨理工大学 | A kind of three-dimensional reconstruction system for the preoperative planning of virtual operation |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111053574A (en) * | 2019-12-05 | 2020-04-24 | 深圳开立生物医疗科技股份有限公司 | Three-dimensional surface reconstruction method and device, ultrasonic equipment and storage medium |
| CN111325821A (en) * | 2020-02-18 | 2020-06-23 | Oppo广东移动通信有限公司 | Grid model processing method and device, equipment and storage medium |
| CN111325821B (en) * | 2020-02-18 | 2023-06-06 | Oppo广东移动通信有限公司 | Grid model processing method and device, equipment and storage medium |
| CN111583391A (en) * | 2020-04-29 | 2020-08-25 | 北京深测科技有限公司 | Object three-dimensional reconstruction method and system |
| CN112465802A (en) * | 2020-12-10 | 2021-03-09 | 上海联影医疗科技股份有限公司 | Method, system, terminal and computer readable storage medium for processing tomographic image |
| CN112767532A (en) * | 2020-12-30 | 2021-05-07 | 华东师范大学 | Ultrasonic or CT medical image three-dimensional reconstruction method based on transfer learning |
| CN112767532B (en) * | 2020-12-30 | 2022-07-08 | 华东师范大学 | Ultrasonic or CT medical image three-dimensional reconstruction method based on transfer learning |
| CN112819950A (en) * | 2021-02-07 | 2021-05-18 | 清华-伯克利深圳学院筹备办公室 | Preparation method, device and equipment of brain tissue section die and storage medium |
| CN113035306A (en) * | 2021-03-17 | 2021-06-25 | 广州华端科技有限公司 | Method, system, equipment and medium for remotely browsing images |
| CN113035306B (en) * | 2021-03-17 | 2024-03-29 | 广州开云影像科技有限公司 | Method, system, equipment and medium for remotely browsing images |
| CN113470180A (en) * | 2021-05-25 | 2021-10-01 | 杭州思看科技有限公司 | Three-dimensional mesh reconstruction method, device, electronic device and storage medium |
| CN113470180B (en) * | 2021-05-25 | 2022-11-29 | 思看科技(杭州)股份有限公司 | Three-dimensional mesh reconstruction method, device, electronic device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110458950A (en) | A kind of method for reconstructing three-dimensional model, mobile terminal, storage medium and electronic equipment | |
| US11710242B2 (en) | Methods and systems for image segmentation | |
| CN110458949A (en) | Method for reconstructing, mobile terminal and the electronic equipment of the two-dimentional tangent plane of threedimensional model | |
| Yan et al. | Variational mesh segmentation via quadric surface fitting | |
| Yang | Dealing with textureless regions and specular highlights-a progressive space carving scheme using a novel photo-consistency measure | |
| EP2074499B1 (en) | 3d connected shadow mouse pointer | |
| WO2017114366A1 (en) | Data rendering method and system thereof | |
| US9218542B2 (en) | Localization of anatomical structures using learning-based regression and efficient searching or deformation strategy | |
| CN109285215A (en) | A kind of human 3d model method for reconstructing, device and storage medium | |
| US10692272B2 (en) | System and method for removing voxel image data from being rendered according to a cutting region | |
| CN109191510B (en) | 3D reconstruction method and device for pathological section | |
| Näf et al. | 3D Voronoi skeletons and their usage for the characterization and recognition of 3D organ shape | |
| US9047660B2 (en) | Network cycle features in relative neighborhood graphs | |
| US20200057778A1 (en) | Depth image pose search with a bootstrapped-created database | |
| Livnat et al. | Interactive point-based isosurface extraction | |
| US20220343589A1 (en) | System and method for image processing | |
| CN105825471A (en) | Unity-3D-based three-dimensional surface reconstruction and rendering method | |
| US11955246B2 (en) | Unified anisotropic volume and surface mesh storage | |
| CN114445328A (en) | Medical image brain tumor detection method and system based on improved Faster R-CNN | |
| US20210044788A1 (en) | System and method for performing 3d imaging of an object | |
| Song et al. | Breast tissue 3D segmentation and visualization on MRI | |
| Arikan et al. | Multi-depth-map raytracing for efficient large-scene reconstruction | |
| Law et al. | Automatic centerline extraction for 3D virtual bronchoscopy | |
| CN109872353B (en) | White light data and CT data registration method based on improved iterative closest point algorithm | |
| CN107316025B (en) | Hand gesture recognition method and system |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191115 |
|
| RJ01 | Rejection of invention patent application after publication |