CN101814193A - Real-time volume rendering method of three-dimensional heart data based on GPU (Graphic Processing Unit) acceleration - Google Patents

Abstract

The invention relates to a real-time volume rendering method of three-dimensional heart data based on GPU (Graphic Processing Unit) acceleration, belonging to the field of graph and image processing and solving the problems of low speed and high requirement on computer performance of the traditional volume rendering the three-dimensional heart data. The real-time volume rendering method of the three-dimensional heart data based on the GPU acceleration is realized based on a PC (Personal Computer) with a graphics card supporting Cg language. The volume rendering method concretely comprises the following steps of: firstly, reading and preprocessing data: organizing data of a plurality of slices of the heart to be drawn into a volume data file, packing the volume data file into a three-dimensional grain data packet and storing the three-dimensional grain data packet into the memory of the graphics card; secondly, converting data coordinates: converting the data coordinates in the there-dimensional grain packet obtained in the data preprocessing step into screen coordinates; and thirdly, outputting the volume data after the coordinate standardization by adopting a volume rendering method. The invention is also suitable for the volume rendering of other three-dimensional organs with complex structures.

Description

Real-time three-dimensional cardiac data object plotting method based on the GPU acceleration

Technical field

What the present invention relates to is a kind of based on GPU (Graphic Processing Unit) object plotting method, particularly a kind of with the method for baroque three-dimensional cardiac data with real-time, interactive, the demonstration of configurable form, belongs to the graph and image processing field.

Background technology

Heart is human recycle system's a power source as the vitals of human body.Heart disease such as coronary heart disease etc. have become one of maximum in the world cause of death.Reinforcement improves the diagnosis and the treatment of heart disease to heart working and pathogenetic understanding, has become a research focus of biomedical engineering field.Thereby relevant medical imaging and modeling and simulating technology are developed rapidly in recent years, have emerged in large numbers a large amount of as three-dimensional cardiac data such as medical section and electric physiological simulation.The complex structure of heart is made up of a plurality of organs and blood vessel.Outside organ has surrounded inner organ, makes people be difficult to understand the inner structure of heart, thereby is difficult to understand the one-piece construction of heart.Vassilios Hurmusiadis has developed the Virtual Heart heart emulation platform that is used for education and training, and this platform has been showed cell, tissue and the organ of heart from function and structure.Use Marching Cube method to draw three cardiac modules in the integrally-built demonstration part of heart, and use the inner structure of the method demonstration heart of shear plane.Because Marching Cube method can not be drawn out the internal information of heart, and shear plane has destroyed the effect of heart integral body, so this system can not draw out the one-piece construction of heart intuitively.People such as Yang design is applicable to the acceleration visualized algorithm of three-dimensional human heart data.This method has been showed the one-piece construction of heart well, but because render speed can't reach the purpose of good user interactions more slowly.

Volume drawing is also referred to as body and plays up (Volume Rendering), is a kind of important graphics process and visualization technique.It not only can the display model surface shape and texture information, and can show the ins and outs that is included in the data and the mutual relationship between data.Because the data volume of volume data is very big, on PC, use software approach to be difficult to realize Interactive Visualization, the researchist can only seek help from the powerful supercomputer of parallel processing capability.

Summary of the invention

The objective of the invention is to, the volume drawing speed that solves existing three-dimensional cardiac data is slow, to the demanding problem of computing power, so provide a kind of fast, can realize three-dimensional cardiac data volume method for drafting.

The real-time three-dimensional cardiac data object plotting method that quickens based on GPU of the present invention is based on that the PC that has the video card of supporting the Cg language realizes, specifically the comprising of described object plotting method:

The step of data read and pretreated step, data coordinates conversion and the step of image rendering, wherein:

Data read and pretreated step are that a plurality of slice of datas of heart to be drawn are organized into the volume data file, and after described volume data file is packaged into the three-D grain packet, store in the internal memory of video card; The sectioning image that the slice of data of heart to be drawn is represented wide and high respectively as the x of volume data, the y axle is with the quantity of all slice of datas of heart the to be drawn z axle as volume data;

The step of data coordinates conversion is to convert the data coordinates in the three-D grain packet of the pretreated step acquisition of data to screen coordinate;

The step of image rendering is to adopt the method for direct volume drawing to draw output the volume data after the coordinate standardization.

Innovative point of the present invention is the geometric manipulations ability of using GPU powerful, be applied to the interactive visual of complex three-dimensional cardiac data, make the data of heart can show the one-piece construction of heart or different organs in real time according to user's input, the present invention has avoided the method for using shear plane, make the user can observe each ingredient of heart, the while is not destroyed the structure of whole heart again.The present invention uses direct volume drawing (Direct Volume Rendering) method, makes the user can observe the one-piece construction of heart.

Advantage of the present invention has:

One, improved the image display speed: adopted the graphic hardware speed technology, promptly adopt existing high-rise painted language (with respect to the assembly language that only designs) to realize the acceleration of graphic presentation, and then improved the display speed of image greatly at specific video card based on the GPU architecture.

Two, improved the arithmetic speed of CPU: adopt graphic hardware to quicken to realize volume drawing, only the packing data that needs will be to be drawn becomes the form of texture to be transferred on the video card, then the complex calculations load also being transferred to video card carries out, make CPU be absorbed in other work, thereby improved the arithmetic speed of CPU greatly.

Three, realize the interactive visual of three-dimensional cardiac data: three-dimensional cardiac data volume method for drafting speed of the present invention is fast, can realize the Interactive Visualization of complex three-dimensional cardiac data, can make the user can alternatively dispose, observe, analyze visualization result.

Four, low cost.This method only needs a common PC and a video card of supporting the Cg language.Present household PC configuration can be used this method.

Five, Gao Du compatibility.The present invention use can be cross-platform development language and kit, so (or making an amendment slightly) may operate under the different operating system and hardware platform in theory.

Six, Gao Du ease for use.The present invention only needs mouse and keyboard just can finish adjustment and configuration to visualization result.Convenient, easy-to-use, user's medical knowledge and computation technical ability is not almost required.

Method of the present invention makes the visual observation analysis of cardiac data become possibility under the prerequisite of the render speed that guarantees cardiac module.The present invention can be applied to medic's intracardiac section foundation training and virtual operation training, also can be as the demo system of popularizing heart disease prevention knowledge.Along with the raising of people's health care consciousness, native system has enormous and latent market.

Method of the present invention still is used for the volume drawing of the three-dimensional internal organs of other labyrinth.

Description of drawings

Fig. 1 is the gray-scale value frequency statistics figure of the volume data described in the embodiment.Fig. 2 is the optical properties mapping graph of heart tissue gray-scale value and each gray-scale value correspondence.Fig. 3 adopts existing method to carry out the picture of volume drawing acquisition according to the slice of data of heart.Fig. 4 and Fig. 5 adopt method of the present invention to draw the picture that obtains, the two sight angle differences of attaching pictures according to the slice of data of heart.

Embodiment

The described real-time three-dimensional cardiac data object plotting method that quickens based on GPU of present embodiment is based on that the PC that has the video card of supporting the Cg language realizes.The graphics process acp chip of described video card is GPU (Graphic Processing Unit), specifically the comprising of this method:

The step of data read and pretreated step, data coordinates conversion and the step of image rendering, wherein:

Data read and pretreated step are that a plurality of slice of datas of heart to be drawn are organized into the volume data file, and after described volume data file is packaged into the three-D grain packet, store in the internal memory of video card; The sectioning image that the slice of data of heart to be drawn is represented wide and high respectively as the x of volume data, the y axle is with the quantity of all slice of datas of heart the to be drawn z axle as volume data; Detailed process is:

Header in each slice of data of heart to be drawn is deleted, to delete then all slice of datas after the header adjust, form satisfy the three-D grain packing require the volume data file, the GPU by the video card of supporting the Cg language is packaged into the three-D grain packet with described volume data file at last.

In the above-mentioned steps, the process of the processing of slice of data and organizator data file has the CPU of PC to realize, and the process that described volume data file is packaged into the three-D grain packet is finished by the GPU of the video card of supporting the Cg language, formed three data texturing bags are stored in the internal memory of video card, do not take the resource of CPU.

The step of data coordinates conversion is to convert the data coordinates in the three-D grain packet of the pretreated step acquisition of data to screen coordinate, and detailed process is:

The coordinate of three-dimensional data is carried out view transformation, projective transformation and normalized transformation successively, finally obtain screen coordinate.

The conversion process that data coordinates is carried out view is: with the setting coordinate of view origin P is coordinate P=(x under the world coordinate system ₀, y ₀, z ₀), then the coordinate translation transformation for mula is:

$T=\left[\begin{array}{cccc}1& 0& 0& -x_0\\ 0& 1& 0& -{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="HSA00000050579200011.png"\; state="\{\{state\}\}">y</figure-callout>}}_0\\ 0& 0& 1& -z_0\\ 0& 0& 0& 1\end{array}\right]---\left(1\right)$

The Rotating Transition of Coordinate formula is:

$R=\left[\begin{array}{cccc}{u}_x& u_y& u_z& 0\\ v_x& v_y& v_z& 0\\ n_x& n_{\mathrm{<figure-callout\; id="0"\; label="y\; n\; z"\; filenames="HSA00000050579200011.png"\; state="\{\{state\}\}">y</figure-callout>}}& n_z{}_{}& 0\\ 0& 0& 0& 1\end{array}\right]---\left(2\right)$

Wherein, vector of unit length u, v, n are three direction vectors of three coordinate axis of the reference frame under the world coordinate system:

Translation transformation formula and rotational transform formula are multiplied each other, obtain to be tied to view coordinate transform formula under the view coordinate system from world coordinates:

$M_{\mathrm{wc},\mathrm{vc}}=R\&CenterDot;T=\left[\begin{array}{cccc}{u}_x& u_y& u_z& -u\&CenterDot;P_0\\ v_x& v_y& v_z& -v\&CenterDot;P_0\\ n_x& n_y& n_z& -n\&CenterDot;{\mathrm{<figure-callout\; id="0"\; label="P"\; filenames="HSA00000050579200011.png"\; state="\{\{state\}\}">P</figure-callout>}}_0\\ 0& 0& 0& 1\end{array}\right]---\left(3\right)$

Data coordinates and view coordinate transform formula are multiplied each other, obtain the data coordinates under the view coordinate system.

Projective transformation is meant the coordinate transformation process of the data coordinates under the view coordinate system being carried out projective transformation, adopts orthogonal projection (Orthogonal Projection) method that the data coordinates under the view coordinate system is carried out conversion in the present embodiment, and detailed process is:

(x, y z) carry out orthogonal projection, and the coordinate that obtains after the projection is: x with the arbitrfary point in the data under the view coordinate system _p=x, y _p=y, former and later two shear surfaces are set on z, be respectively hither plane (near plane) and far plane (far plane), part between described hither plane (near plane) and far plane (far plane) is what comes into a driver's body (View Volume), and the coordinate that carries out all data after the projective transformation all is positioned at described what comes into a driver's body (View Volume).

The effect of above-mentioned projective transformation is that the object in the three-dimensional space is represented to be mapped on the view plane of two dimension.In the process of projective transformation, can adopt various projection patterns to carry out coordinate transform, through experimental test, the orthogonal projection method that present embodiment adopted can reach best visual effect.

Normalized transformation is meant that the data coordinates that will obtain after the projection carries out normalized conversion, and detailed process is:

To be positioned at the coordinate points (x in the hither plane lower left corner _Min, y _Min, z _Near) be mapped to (1 ,-1 ,-1), will be positioned at the coordinate points (x in the far plane upper right corner _Max, y _Max, z _Far) be mapped to (1,1,1), then according to formula

$\mathrm{Mortho},\mathrm{norm}=\left[\begin{array}{cccc}\frac{2}{{\mathrm{xw}}_{\max }-{\mathrm{<figure-callout\; id="0"\; label="xw\; min"\; filenames="HSA00000050579200011.png"\; state="\{\{state\}\}">xw</figure-callout>}}_{\min }}& 0& 0& -\frac{{\mathrm{xw}}_{\max }+{\mathrm{xw}}_{\min }}{{\mathrm{xw}}_{\max }-{\mathrm{<figure-callout\; id="0"\; label="xw\; min"\; filenames="HSA00000050579200011.png"\; state="\{\{state\}\}">xw</figure-callout>}}_{\min }}\\ 0& \frac{2}{{\mathrm{yw}}_{\max }-{\mathrm{yw}}_{\min }}& 0& -\frac{{\mathrm{yw}}_{\max }+{\mathrm{yw}}_{\min }}{{\mathrm{yw}}_{\max }-{\mathrm{yw}}_{\mathrm{<figure-callout\; id="0"\; label="min"\; filenames="HSA00000050579200011.png"\; state="\{\{state\}\}">min</figure-callout>}}}\\ 0& 0& \frac{-2}{z_{\mathrm{near}}-z_{\mathrm{far}}}& \frac{z_{\mathrm{near}}+z_{\mathrm{far}}}{z_{\mathrm{near}}-z_{\mathrm{far}}}\\ 0& 0& 0& 1\end{array}\right]---\left(5\right)$

To be positioned at hither plane z _NearWith far plane z _FarBetween the coordinate of all data all be mapped between-1 to 1;

With the coordinate of data after the standardization according to formula

$\left[\begin{array}{cccc}\frac{{\mathrm{xv}}_{\max }-{\mathrm{xv}}_{\min }}{2}& 0& 0& \frac{{\mathrm{xv}}_{\max }+{\mathrm{xv}}_{\min }}{2}\\ 0& \frac{{\mathrm{yv}}_{\max }-{\mathrm{yv}}_{\min }}{2}& 0& \frac{{\mathrm{yw}}_{\max }+{\mathrm{yw}}_{\min }}{2}\\ 0& 0& \frac{1}{2}& \frac{1}{2}\\ 0& 0& 0& 1\end{array}\right]---\left(6\right)$

Become screen coordinate, the z coordinate of being had a few on screen all is 0, and the coordinate in the lower left corner of the screen area of viewport (viewport) is (xv _Min, yv _Min, 0), the coordinate in the upper right corner is (xv _Max, yv _Max, 0).

The step of image rendering is to adopt the method for direct volume drawing to draw output the volume data after the coordinate standardization.

Direct volume drawing (Direct Volume Rendering) is a kind of existing object plotting method, and it can demonstrate the overall situation of volume data, especially can disclose the internal feature of data.Its basic thought is that the optical properties that calculates voxel comprises color and opacity, and the light that simulation is injected from specific direction is propagated volume data.

(or claim: optical properties voxel) is provided with by transport function (Transfer Function) volume data usually, and acquisition has the volume data of optical properties, described transport function realizes the gray-scale value of each voxel in the volume data and the corresponding relation of optical properties, and detailed process is:

The process of the direct volume drawing in the present embodiment is: according to the design of the intensity profile in volume data transport function, set up the corresponding relation between gray-scale value and the optical properties, and store into described transport function in the internal memory of video card as the one dimension data texturing; Simulate the light of injecting from assigned direction then and volume data, propagate, finish the volume drawing of three-dimensional cardiac.

Through after the above-mentioned steps, volume data and transport function have all suffered at the video card internal memory, and then, the GPU in the video card just can calculate and obtain the optical properties of each volume data, and image rendering is come out, and has made full use of the computing power that GPU uses like this.Improved speed greatly.

Described transport function has been set up the gray-scale value of volume data and the corresponding relation between the optical properties, converts the volume data that has optical properties to so that will have the volume data of gray-scale value.

Transport function reasonable in design can disclose the important internal information and the space structure of volume data, strengthens important information, suppresses the visual effect of inessential information.

According to the design of the intensity profile in volume data transport function, the design process that the process of setting up the corresponding relation between gray-scale value and the optical properties is is in the present embodiment:

At first, the distribution situation of the gray-scale value of statistics volume data.

Adopting the gray-scale value statistical graph to realize this process in the present embodiment, referring to shown in Figure 1, is the statistics of the volume data of a heart.Among the figure, transverse axis is represented different heart tissue corresponding gray, and the scope of the gray-scale value of the data relevant with heart tissue is in interval [30,70], and the gray-scale value that is not marked in this interval on the transverse axis represents that this gray-scale value does not appear in the volume data of heart.

Then, grey value profile situation according to volume data, determine the heart tissue of different gray scale correspondences, and its corresponding optical properties is set according to the locus of described heart tissue, be opacity and color value, the opacity that setting is finished and the set of color are organized into color lookup table ColorMap; Between gray-scale value and the optical properties corresponding relation set up and to finish.

In the process of image rendering, in color lookup table ColorMap, retrieve corresponding optical properties according to the gray-scale value of the different heart tissues of volume data, according to the corresponding formula of gray scale-optical properties:

F(g)→(R，G，B，A)????????????????????????????(7)

Realize that wherein g represents gray-scale value, (B A) represents the optical properties of this gray scale correspondence for R, G, and wherein R, G and B represent color, and A represents opacity (opacity).

When different gray-scale value optical properties was set, the main principle of following was the three-dimensional relative position by the comparison heart tissue, made the lacuna opacity of heart inside be higher than the tissue of heart outside, and the link portions branch of different lacunas is different from lacuna itself.

In the present embodiment, determine the tissue of different gray scale correspondences and its corresponding optical properties is set according to intensity profile shown in Figure 1 according to the locus of organizing, the heart tissue gray-scale value that forms and the optical properties mapping graph of each gray-scale value correspondence are referring to shown in Figure 2, and the longitudinal axis is represented heart volume data gray scale is added up the different gray-scale values that obtain among the figure.Square color lump among the figure represents to use the corresponding formula of gray scale-optical properties to retrieve this gray-scale value in color lookup table ColorMap, according to the optical properties of the data that this gray scale is arranged.Four-tuple (R, G, B, A) expression of the optical properties of each gray value data below square color lump.

Adopt radiation-absorption (the emission and absorption) spread state of modeling light in volume data then, and finally produce sampling to volume data along direction of visual lines, and along the color of direction of visual lines accumulative total voxel and the result of opacity, the volume drawing of the heart of finally completing.

Color during final drafting of the volume data that each is discrete is:

$C=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{C}_i\underset{j=1}{\overset{i-1}{\mathrm{\&Pi;}}}(1-A_i)---\left(8\right)$

Opacity during final drafting of the volume data that each is discrete is:

$A=1-\underset{j=1}{\overset{n}{\mathrm{\&Pi;}}}(1-A_j)---\left(9\right)$

Wherein, C _iAnd A _iBy the transport function setting, represent color and opacity respectively i sampling number certificate.A _iThe light that the approximate representation volume data absorbs.The color C of opacity weighting _iApproximate representation is along volume data radiation between light sampled point i and i+1 the sampling and the light that absorbs.The light that partly is illustrated in i sampled point for the product in the color component had been constantly to decay before intelligent's eyes.

Use OpenGL and Cg implementation algorithm, wherein OpenGL is responsible for the demonstration of processing graphics, and user interactions, and Cg is used for writing vertex program (vertex shader) and sheet metaprogram (fragment shader).

Three-dimensional cardiac rendering algorithm based on GPU can adopt following program to realize:

Input：segmented?3D?heart?volume?dataset，color?lookup?table

Output：heart?image

1.VolumeTexture←Volume?Dataset，ColorMap←Color?lookup?table；

2.while?v?in?VolumeTexture

3.????v＝T·v；

4.????clip(v)；

5.????v(color)＝ColorMap[v(greyLevel)]；

6.for?i←1?to?sliceNum

7.????drawSlice()；

8.????composite()；

The implication of said procedure section is:

Be input as: cut apart the volume data behind the tissue, and color lookup table;

Be output as: cardiac image.

Variable V olumeTexture represents the three-D grain of package data, Volume Dataset represents the set of volume data, ColorMap represents to deposit the one dimension texture of the volume data optical properties of different gray scales, Colorlookup table represents color lookup table (optical properties of different gray-scale value correspondences), and sliceNum represents the quantity of volume data along the section of the direction of sight line.

To each voxel v in three-D grain, adopt matrix T to carry out coordinate conversion, then adopting Clip (v) voxel v to be limited in the effective space, adopt ColorMap[v (greyLevel) again], with the gray-scale value v (greyLevel) of voxel v as index, in color lookup table, search corresponding one dimension data texturing, and the optical properties v (color) of this voxel is set according to result for retrieval.

SliceNum does following processing respectively until last is cut into slices to the 1st the section beginning of direction along sight line: adopt drawSlice () to draw current section, adopt composite () that current slice and previous section are mixed then.Handle after all sliceNum the sections, drawing result is presented on the view plane the most at last.

The described object plotting method of present embodiment can adopt Intel Core2 Quad CPU 2.41GHz, internal memory 4GB, video card NVIDIA GF8600, video memory 256MB hardware environment.Use C Plus Plus and OpenGL to realize the framework of program, the summit of GPU and sheet metaprogram adopt the Cg language compilation, use version to be Cg2.0.

In the read method of volume data, contrast OpenGL that acceleration ray cast (Accelerated Ray Casting) ARC method, OpenGL (Open Graphics Library) stationary pipes collimation method and present embodiment that people such as existing Yang proposes adopt in conjunction with Cg pipeline method able to programme processing time of totally three kinds of methods, through the data read time of the above-mentioned three kinds of methods of experiment acquisition, result's ginseng is shown in Table 1:

Table 1 algorithms of different reading of data time ratio

Method	Pretreatment time	Size of data
			The ARC method	150.562 second	?472×325×4878bit
The OpenGL method	9.671 second	?472×325×4878bit
			The OpenGL+Cg method	0.344 second	?472×325×4878bit

In drawing process, because the ARC method is difficult to realize real-time user interactions that only compared the drafting efficient of the OpenGL of OpenGL fixed pipelines method and present embodiment employing in conjunction with the method for Cg pipeline able to programme here, experimental result is as shown in table 2.

The average render speed of table 2 algorithms of different relatively

Method for drafting	Size of data	The average frame number of per second (f/s)
			The OpenGL method	??472×325×4878bit	?8.5
The OpenGL+Cg method	??472×325×4878bit	?14.6

Referring to Fig. 3-shown in Figure 5, be comparison to drawing result before and after the data classification.Use transport function to carry out sorted image and can obviously see the profile of heart, structure and inner lacuna.The function void glBlendFunc (GLenumsfactor, GLenum dfactor) that the color of classification back image uses OpenGL to provide by the color among Fig. 1 mixes.

The present invention uses GPU figure speed technology to realize the three-dimensional cardiac visualization of data, makes the user significantly reduce the response time of system with respect to non-GPU method with alternant way observer heart.This paper is made into three-D grain with the three-dimensional cardiac data set, has designed transport function according to the characteristics of cardiac data, uses the method for rectangle closure to design rendering algorithm at cardiac data.

Use OpenGL and Cg that algorithm is realized, and use three-dimensional cardiac data to test.Can see that according to experimental result method used herein is significantly increased with software approach reading in the process relatively of data, in the process of drafting, has realized real-time user interactions on efficient.

Claims (10)

1. the real-time three-dimensional cardiac data object plotting method that quickens based on GPU is characterized in that it is based on that the PC that has the video card of supporting the Cg language realizes, specifically the comprising of described object plotting method:

The step of data read and pretreated step, data coordinates conversion and the step of image rendering, wherein:

Data read and pretreated step are that a plurality of slice of datas of heart to be drawn are organized into the volume data file, and after described volume data file is packaged into the three-D grain packet, store in the internal memory of video card; The sectioning image that the slice of data of heart to be drawn is represented wide and high respectively as the x of volume data, the y axle is with the quantity of all slice of datas of heart the to be drawn z axle as volume data;

The step of data coordinates conversion is to convert the data coordinates in the three-D grain packet of the pretreated step acquisition of data to screen coordinate;

The step of image rendering is to adopt the method for direct volume drawing to draw output the volume data after the coordinate standardization.

2. the real-time three-dimensional cardiac data object plotting method that quickens based on GPU according to claim 1 is characterized in that the process that a plurality of slice of datas of heart to be drawn is organized into the volume data file is:

With the header in each slice of data of heart to be drawn deletion, will delete then all slice of datas after the header adjust, form satisfy that three-D grain packs require the volume data file.

3. the real-time three-dimensional cardiac data object plotting method that quickens based on GPU according to claim 1 is characterized in that the detailed process of the step of data coordinates conversion is:

The coordinate of three-dimensional data is carried out view transformation, projective transformation and normalized transformation successively, finally obtain screen coordinate.

4. the real-time three-dimensional cardiac data object plotting method that quickens based on GPU according to claim 3, it is characterized in that, view transformation is meant the conversion of data coordinates being carried out view, and detailed process is: with the setting coordinate of view origin P is coordinate P=(x under the world coordinate system ₀, y ₀, z ₀), then the coordinate translation transformation for mula is:

$T=\left[\begin{array}{cccc}1& 0& 0& {-x}_0\\ 0& 1& 0& {-y}_0\\ 0& 0& 1& {-z}_0\\ 0& 0& 0& 1\end{array}\right]$

The Rotating Transition of Coordinate formula is:

$R=\left[\begin{array}{cccc}{u}_x& u_y& u_z& 0\\ v_x& v_y& v_z& 0\\ n_x& n_y& n_z& 0\\ 0& 0& 0& 1\end{array}\right]$

Wherein, vector of unit length u, v, n are three direction vectors of three coordinate axis of the reference frame under the world coordinate system:

Translation transformation formula and rotational transform formula are multiplied each other, obtain to be tied to view coordinate transform formula under the view coordinate system from world coordinates:

$M_{\mathrm{wc},\mathrm{vc}}=R\&CenterDot;T=\left[\begin{array}{cccc}{u}_x& u_y& u_z& -u\&CenterDot;P_0\\ v_x& v_y& v_z& -v\&CenterDot;P_0\\ n_x& n_y& n_z& -n\&CenterDot;P_0\\ 0& 0& 0& 1\end{array}\right]$

Data coordinates and view coordinate transform formula are multiplied each other, obtain the data coordinates under the view coordinate system.

5. the real-time three-dimensional cardiac data object plotting method that quickens based on GPU according to claim 3, it is characterized in that, described projective transformation is meant the coordinate transformation process of the data coordinates under the view coordinate system being carried out projective transformation, adopt the orthogonal projection method that the data coordinates under the view coordinate system is carried out conversion, detailed process is:

(x, y z) carry out orthogonal projection, and the coordinate that obtains after the projection is: x with the arbitrfary point in the data under the view coordinate system _p=x, y _p=y is provided with former and later two shear surfaces on z, be respectively hither plane and far plane, and the part between described hither plane and far plane is the what comes into a driver's body, and the coordinate that carries out all data after the projective transformation all is positioned at described what comes into a driver's body.

6. the real-time three-dimensional cardiac data object plotting method that quickens based on GPU according to claim 3 is characterized in that, normalized transformation is meant that the data coordinates that will obtain after the projection carries out normalized conversion, and detailed process is:

To be positioned at the coordinate points (x in the hither plane lower left corner _Min, y _Min, z _Near) be mapped to (1 ,-1 ,-1), will be positioned at the coordinate points (x in the far plane upper right corner _Max, y _Max, z _Far) be mapped to (1,1,1), then according to formula

$\mathrm{Mortho},\mathrm{norm}=\left[\begin{array}{cccc}\frac{2}{{\mathrm{xw}}_{\max }-{\mathrm{xw}}_{\min }}& 0& 0& -\frac{{\mathrm{xw}}_{\max }+{\mathrm{xw}}_{\min }}{{\mathrm{xw}}_{\max }-{\mathrm{xw}}_{\min }}\\ 0& \frac{2}{{\mathrm{yw}}_{\max }-{\mathrm{yw}}_{\min }}& 0& -\frac{{\mathrm{yw}}_{\max }+{\mathrm{yw}}_{\min }}{{\mathrm{yw}}_{\max }-{\mathrm{yw}}_{\min }}\\ 0& 0& \frac{-2}{z_{\mathrm{near}}-z_{\mathrm{far}}}& \frac{z_{\mathrm{near}}+z_{\mathrm{far}}}{z_{\mathrm{near}}-z_{\mathrm{far}}}\\ 0& 0& 0& 1\end{array}\right]$

To be positioned at hither plane z _NearWith far plane z _FarBetween the coordinate of all data all be mapped between-1 to 1;

With the coordinate of data after the standardization according to formula

$\mathrm{Mnormviewvol},3\mathrm{Dscreen}=\left[\begin{array}{cccc}\frac{{\mathrm{xv}}_{\max }-{\mathrm{xv}}_{\min }}{2}& 0& 0& \frac{{\mathrm{xv}}_{\max }+{\mathrm{xv}}_{\min }}{2}\\ 0& \frac{{\mathrm{yv}}_{\max }-{\mathrm{yv}}_{\min }}{2}& 0& \frac{{\mathrm{yw}}_{\max }+{\mathrm{yw}}_{\min }}{2}\\ 0& 0& \frac{1}{2}& \frac{1}{2}\\ 0& 0& 0& 1\end{array}\right]$

Become screen coordinate, the z coordinate of being had a few on screen all is 0, and the coordinate in the lower left corner of the screen area of viewport is (xv _Min, yv _Min, 0), the coordinate in the upper right corner is (xv _Max, yv _Max, 0).

7. the real-time three-dimensional cardiac data object plotting method that quickens based on GPU according to claim 1, the step that it is characterized in that image rendering adopts the method for direct volume drawing to realize, the process of direct volume drawing is: according to the design of the intensity profile in volume data transport function, set up the corresponding relation between gray-scale value and the optical properties, and store into described transport function in the internal memory of video card as the one dimension data texturing; Simulate the light of injecting from assigned direction then and volume data, propagate, finish the volume drawing of three-dimensional cardiac.

8. the real-time three-dimensional cardiac data object plotting method that quickens based on GPU according to claim 7 is characterized in that, described according to the design of the intensity profile in volume data transport function, the process of setting up the corresponding relation between gray-scale value and the optical properties is:

At first, the distribution situation of the gray-scale value of statistics volume data;

Then, grey value profile situation according to volume data, determine the heart tissue of different gray scale correspondences, and its corresponding optical properties is set according to the locus of described heart tissue, be opacity and color value, the opacity that setting is finished and the set of color are organized into color lookup table ColorMap; Between gray-scale value and the optical properties corresponding relation set up and to finish.

9. the real-time three-dimensional cardiac data object plotting method that quickens based on GPU according to claim 7, it is characterized in that, the light that simulation is injected from assigned direction is propagated volume data, the process of finishing the volume drawing of cardiac data is: adopt the spread state of radiation-absorbing model simulated light in volume data, and finally produce sampling to volume data along direction of visual lines, and along the color of direction of visual lines accumulative total voxel and the result of opacity, obtain the final drafting color and the opacity of each voxel one by one, and then the volume drawing of the heart of finally completing.

10. the real-time three-dimensional cardiac data object plotting method that quickens based on GPU according to claim 9 is characterized in that the final drafting color of each voxel and the preparation method of opacity are: the color during final drafting of the volume data that each is discrete is:

$C=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{C}_i\underset{j=1}{\overset{i-1}{\mathrm{\&Pi;}}}(1-A_i)$

Opacity during final drafting of the volume data that each is discrete is:

$A=1-\underset{j=1}{\overset{n}{\mathrm{\&Pi;}}}(1-A_j)$

Wherein, C _iAnd A _iBy the transport function setting, represent color and opacity respectively, A i sampling number certificate _iThe light that the approximate representation volume data absorbs.

Images