CN104954715A - GPU (graphics processing unit) acceleration based video display method adopting multi-projector splicing fusion on special-shaped screens - Google Patents

Abstract

The invention discloses a GPU (graphics processing unit) acceleration based video display method adopting multi-projector splicing fusion on special-shaped screens. The method comprises steps as follows: step (1), arranging at least two projectors, generating control meshes according to the number of the projectors and adjusting the control meshes on the special-shaped screens; step (2), reading a video file; step (3), reading one frame of image of the video file as a source image, adjusting the control meshes and performing geometric correction on the source image to obtain a target image, and getting a mapping matrix of pixel points in the source image and the target image; step (4), reading a next frame of image of the video file; step (5), generating an intermediate image; step (6), performing brightness processing and image splicing fusion on the intermediate image through GPU acceleration; step (7), obtaining a target image corresponding to the image after splicing fusion; step (8),projecting the target image corresponding to the image after splicing fusion on the special-shaped screens; step (9), repeating the step (4) to step (8) until the video display is finished.

Description

Based on the method for the special-shaped screen multi-projection system fusion video playback that GPU accelerates

Technical field

The invention belongs to projection control field, particularly relate to a kind of method that special-shaped screen multi-projection system accelerated based on GPU merges video playback.

Background technology

Large scale special-shaped screen prjection can strengthen the feeling of immersion of user, brings better experience, is widely used in fields such as virtual reality, education, amusement, simulation trainings.There is various projection screen in actual applications, even if for a specific application scenarios, As time goes on its parameters such as the position of projecting apparatus, screen shape also can change, so wish that optical projection system will be easy to build and safeguard, and the operation often more complicated in reality.

When splicing with multiple projecting apparatus, use hard-edge splicing, namely allow the picture of two projecting apparatus simply be close together, Problems existing is: usually have obvious cut-off rule; When using overlap, if do not processed the brightness of image, the brightness of overlapping region can be higher, affects visual effect, so brightness will be used to merge.

For video playback, based on the shape of projection screen, launch again after needs correct each frame wanted in projection video, because image procossing is more time-consuming, a lot of application is play after video is descended preliminary treatment online again, if scene conversion, need again to carry out preliminary treatment to video, just need the wait one long period to watch after such shift scene, and each scene will preserve a corresponding video file.

Summary of the invention

In order to solve the shortcoming of prior art, the invention provides a kind of method that special-shaped screen multi-projection system accelerated based on GPU merges video playback, the method achieve multi-projector to project on special-shaped screen splicing, when projecting apparatus and screen position constant, carry out correction to video in real time to play, save and pretreated operation is carried out to video.

For achieving the above object, the present invention is by the following technical solutions:

Based on a method for the special-shaped screen multi-projection system fusion video playback that GPU accelerates, comprising:

Step (1): at least two projecting apparatus are set, the number according to projecting apparatus generates control mesh, and special-shaped screen adjusts control mesh, the corresponding projecting apparatus of each group control mesh;

Step (2): read video file, obtains video file attributes; Read a two field picture of video file as source images; Adjustment control mesh carries out geometric correction to source images, and after the correction obtained, image is as target image, and asks for the mapping matrix of pixel in source images and target image;

Step (3): the next frame image reading video file; Shared by the overlapping region of adjacent projectors, the ratio of territory, whole projecting apparatus frame buffer zone size, generates a width intermediate image;

Step (4): the method adopting GPU to accelerate carries out brightness processed to intermediate image and image mosaic merges; Carrying out geometric correction according to the method that mapping matrix and the GPU of pixel in source images and target image accelerate to splicing the image after merging, obtaining the target image that splicing fused image is corresponding;

Step (5): the target image that projection splicing fused image is corresponding on special-shaped screen;

Step (6): repeat step (3) ~ step (5), until video playback completes.

The detailed process of the adjustment control mesh in described step (1) is:

Generation control mesh is in full screen on special-shaped screen, then control mesh is increased or reduces one group of control mesh, the number increasing and decreasing grid line or the operation of mobile control point, make the control mesh that is projected on screen indeformable, and the alignment of the grid line of different projecting apparatus.

Video file attributes in described step (1), comprising: the size of video frame rate, each frame and the total duration of video.

The detailed process asking for the mapping matrix of pixel in source images and target image in described step (2) is:

Step (2.1): the mapping matrix of definition source images and target image is two-dimensional matrix M _xand M _y, two-dimensional matrix M _xand M _yall identical with the size of target image;

Step (2.2): the arbitrary pixel (i, j) in source images, obtains target image respective pixel (i', j') by the method adjusting control mesh, and at M _xmiddle preservation i', at M _ymiddle preservation j'; Wherein, i and j represents integer variable.

The detailed process generating intermediate image in described step (3) is:

Suppose there is N number of projecting apparatus, the overlapping region of adjacent two projecting apparatus is single view field after correcting so first, source images is divided in the horizontal direction N (m-1)+1 part, every a width is d, and they are numbered 0,1 ..., N (m-1); Generate a panel height and wide intermediate image for d × m × N identical with source images, intermediate image is divided into m × N part;

K (m-1) will be numbered in source images, k (m-1)+1 ... (k+1) region of (m-1) is added in intermediate image and is numbered km, km+1, (k+1) region of m-1, each overlapping region is launched by two projecting apparatus respectively; Wherein, N represents the integer being more than or equal to 2; M represents the integer being more than or equal to 2; D represents constant; K represents the arbitrary integer from 0 to N.

Image after merging splicing in described step (4) carries out the detailed process of geometric correction, comprising:

Step (4.1): the fade factor determining image overlapping region;

Step (4.2): the method adopting GPU to accelerate obtains each pixel of image overlapping region, after the rgb value of each pixel obtaining image overlapping region is multiplied by its each self-corresponding fade factor, obtains new value and replaces original pixel value.

In described step (4.1), if the brightness of image overlapping region is constant, then there is not overlapping region, fade factor is 1; Otherwise there is overlapping region, the fade factor of image overlapping region comprises right image overlapping region fade factor and left image overlapping region fade factor.

The expression formula of described right image overlapping region fade factor is:

$f\left(x\right)=\left\{\begin{array}{c}\mathrm{\&omega;}{\left(2x\right)}^{\mathrm{\&upsi;}},0<x<0.5\\ 1-(1-\mathrm{\&omega;}){(2-2x)}^{\mathrm{\&upsi;}+2},0.5\&le;x\&le;1\end{array}\right.$

The expression formula of described left image overlapping region fade factor is:

$f\left(x\right)=\left\{\begin{array}{c}1-(1-\mathrm{\&omega;}){\left(2x\right)}^{\mathrm{\&upsi;}+2},0<x<0.5\\ \mathrm{\&omega;}{(2-2x)}^{\mathrm{\&upsi;}},0.5\&le;x\&le;1\end{array}\right.$

Wherein, ω=0.7, υ=2.5, x represents current in left image or right image and is listed in overlapping region row number and the ratio of total columns of single overlapping region.

In described step (4.2), the detailed process of the method that GPU accelerates is:

In GPU kernel function, first thread number is obtained, calculate the pixel (i corresponding to this thread, j) fade factor and corresponding to pixel (i, j), by pixel (i, j) fade factor corresponding to and pixel (i, j) rgb value is multiplied, as the pixel value that pixel (i, j) is final.

Special-shaped screen in described step (1) and step (5) comprises arcuate screens, annular screen and spherical screen.

Beneficial effect of the present invention is:

(1) the present invention is when the multiple projecting apparatus of special-shaped screen splices, the method adopting GPU to accelerate carries out brightness processed to intermediate image and image mosaic merges, reduce the cut-off rule between two images, and brightness fusion is carried out to two superimposed images, enhance visual effect;

(2) the present invention adopts the method for splicing fusion and adjustment control mesh to improve the efficiency of video playback to image rectification and the specific demand of scene conversion, and the video real time correction realized on special-shaped screen is play;

(3) the method cost of the present invention is low, simple to operate, is easy to build and safeguard.

Accompanying drawing explanation

Fig. 1 (a) is original control mesh Control point schematic diagram;

Fig. 1 (b) is the control Control point schematic diagram after adjustment;

Fig. 2 (a) is original control mesh corresponding pixel points schematic diagram;

Fig. 2 (b) is the control corresponding pixel points schematic diagram after adjustment;

Fig. 3 is control mesh adjustment flow chart;

Fig. 4 (a) is original control mesh image;

Fig. 4 (b) is the control mesh image after adjustment;

Fig. 5 (a) is uncorrected image;

Fig. 5 (b) is that projection splicing corrects result;

Fig. 6 is video playback flow chart.

Embodiment

Below in conjunction with accompanying drawing and embodiment, the present invention will be further described:

This program is with the Visual Studio 2010 in Windows 7 for development platform, and its embodiment is as follows:

Multiple stage projecting apparatus is connected on same computer by video card or deconcentrator by the present invention, projecting apparatus is arbitrarily put, but make the view field of adjacent projectors have certain overlap, multiple display is set to mode of extension, set the view field of projecting apparatus.

The special-shaped screen multi-projection system accelerated based on GPU of the present invention merges the method for video playback, manually adjusts control mesh, completes multi-projector splicing fast and corrects, and accelerates to realize the system that video processes broadcasting in real time by using GPU.

System of the present invention comprises:

Projecting apparatus (2 and more than): projected image;

Arc curtain: image that displaying projector is thrown;

Special-shaped screen prjection splicing system: generate control mesh, receives and manually adjusts, process and show, process displaying video in real time to image.

Image mosaic in the present invention is manually splicing, and its principle is:

By projecting apparatus, control mesh is projected on screen, the intersection point of grid line is called control point, adjusts control point, and screen is determined a desirable target projection region, make it fill in this region to correct image, also just reach the effect that we want.Have several projecting apparatus just to have several groups of control meshs, one group of control mesh can only be launched by a projecting apparatus, separate between each group control mesh.On screen, grid line corresponding to adjacent projectors overlapping region, control point will align.

First manually adjust control point, make the visual effect that Grid Projection image reaches desirable on screen, the size of each small grid and being evenly distributed, the grid line of overlapping region and control point alignment, can save the control mesh adjusted.As long as piece image is carried out the distortion same with control mesh to be launched by the same projecting apparatus of same position, just undeformed projected image can be obtained.

Below the process of carrying out image rectification:

For the source image regions I that certain projecting apparatus is corresponding _s, suppose that the control mesh of its correspondence has the capable v row of u, we call target image the image after correction, generate the width size all black picture identical with grid image as target image, and the region of the corresponding target image of this projecting apparatus is I _d, by source image regions I _sevenly be divided into (u-1) OK, (v-1) arrange, angle point in image is one to one according to ranks order and control point, so just obtains these angle points position in the target image, namely obtains the corresponding relation of part point in source images and target image.The pixel at these some places can find original pixels easily, and the pixel for other will go out corresponding original pixels by interpolation calculation:

Suppose I _son have original control mesh C ₀, I _don have the control mesh C after adjustment ₁:

For I _din each pixel P ₁(x ₁, y ₁), according to C ₁, judge whether it is positioned at control mesh inside, if in outside, this pixel is set to black.If inner at grid, then this pixel is found to be by any four control point A ₁, B ₁, C ₁, D ₁in the quadrangle formed, C can be found accordingly ₀four somes A, B, C, D of middle correspondence, as shown in Fig. 1 (a) He Fig. 1 (b).

Further, find the triangle at place, in order to try to achieve P ₁color value, need to find source images I _sin corresponding points P.As shown in Fig. 2 (a) He Fig. 2 (b), control point A, B, C, D and A ₁, B ₁, C ₁, D ₁coordinate can obtain, so according to Di Luoni trigonometric interpolation:

$\frac{S_{\mathrm{\&Delta;}ABP}}{S_{\mathrm{\&Delta;}ABC}}=\frac{S_{{\mathrm{\&Delta;A}}_1{B}_1{P}_1}}{S_{{\mathrm{\&Delta;A}}_1{B}_1{C}_1}},\frac{S_{\mathrm{\&Delta;}BCP}}{S_{\mathrm{\&Delta;}ABC}}=\frac{S_{{\mathrm{\&Delta;B}}_1{C}_1{P}_1}}{S_{{\mathrm{\&Delta;A}}_1{B}_1{C}_1}},\frac{S_{\mathrm{\&Delta;}CAP}}{S_{\mathrm{\&Delta;}ABC}}=\frac{S_{{\mathrm{\&Delta;C}}_1{A}_1{P}_1}}{S_{{\mathrm{\&Delta;A}}_1{B}_1{C}_1}},$

Leg-of-mutton area is represented with the coordinate on three summits, such as:

Can solve:

$\begin{array}{c}x=\frac{(S_{{\mathrm{\&Delta;B}}_1{C}_1{P}_1}k+x_b{y}_c-y_b{x}_c)(x_a-x_b)-(S_{{\mathrm{\&Delta;A}}_1{B}_1{P}_1}k+x_a{y}_a-y_a{x}_b)(x_b-x_c)}{S_{\mathrm{\&Delta;}ABC}},\\ y=\frac{(S_{{\mathrm{\&Delta;B}}_1{C}_1{P}_1}k+x_b{y}_c-y_b{x}_c)(y_a-y_b)-(S_{{\mathrm{\&Delta;A}}_1{B}_1{P}_1}k+x_a{y}_a-y_a{x}_b)(y_c-y_b)}{S_{\mathrm{\&Delta;}ABC}},\end{array}$

Wherein $k=\frac{S_{\mathrm{\&Delta;}ABC}}{S_{{\mathrm{\&Delta;A}}_1{B}_1{C}_1}}.$

Calculate the approximate coordinate of P point thus, according to the attribute of image, the color value of neighbor pixel has Maximum Possibility to be identical, with the coordinate of this approximate coordinate as P.Then the pixel value of point (x, y) is assigned to P ₁, this completes the generation of target image.

Fig. 3 is control mesh adjustment flow chart.Below each step in figure is elaborated:

Step 1: the control mesh preserved before generating new control mesh or loading;

Step 2: control mesh is displayed in full screen;

Step 3: adjustment control mesh.Comprise and increase or reduce by one group of control mesh, the number of increase and decrease grid line, mobile control point, returns to initial condition etc.; Fig. 4 (a) is original control mesh, and Fig. 4 (b) is the control mesh after adjustment;

Step 4: preserve control mesh.

As shown in Figure 6, a kind of special-shaped screen multi-projection system accelerated based on GPU merges the method for video playback, comprising:

Step (1): at least two projecting apparatus are set, the number according to projecting apparatus generates control mesh, and special-shaped screen adjusts control mesh, the corresponding projecting apparatus of each group control mesh;

Step (2): read video file, obtains video file attributes; Read a two field picture of video file as source images; Adjustment control mesh carries out geometric correction to source images, and after the correction obtained, image is as target image, and asks for the mapping matrix of pixel in source images and target image;

Step (3): the next frame image reading video file; Shared by the overlapping region of adjacent projectors, the ratio of territory, whole projecting apparatus frame buffer zone size, generates a width intermediate image;

Step (4): the method adopting GPU to accelerate carries out brightness processed to intermediate image and image mosaic merges; Carrying out geometric correction according to the method that mapping matrix and the GPU of pixel in source images and target image accelerate to splicing the image after merging, obtaining the target image that splicing fused image is corresponding;

Step (5): the target image that projection splicing fused image is corresponding on special-shaped screen;

Step (6): repeat step (3) ~ step (5), until video playback completes.

The detailed process of the adjustment control mesh in step (1) is:

Generation control mesh is in full screen on special-shaped screen, then control mesh is increased or reduces one group of control mesh, the number increasing and decreasing grid line or the operation of mobile control point, make the control mesh that is projected on screen indeformable, and the alignment of the grid line of different projecting apparatus.

Video file attributes in step (1), comprising: the size of video frame rate, each frame and the total duration of video.

The detailed process asking for the mapping matrix of pixel in source images and target image in step (2) is:

Step (2.1): the mapping matrix of definition source images and target image is two-dimensional matrix M _xand M _y, two-dimensional matrix M _xand M _yall identical with the size of target image;

Step (2.2): the arbitrary pixel (i, j) in source images, obtains target image respective pixel (i', j') by the method adjusting control mesh, and at M _xmiddle preservation i', at M _ymiddle preservation j'; Wherein, i and j represents integer variable.

The detailed process generating intermediate image in step (3) is:

Suppose there is N number of projecting apparatus, the overlapping region of adjacent two projecting apparatus is single view field after correcting so first, source images is divided in the horizontal direction N (m-1)+1 part, every a width is d, and they are numbered 0,1 ..., N (m-1); Generate a panel height and wide intermediate image for d × m × N identical with source images, intermediate image is divided into m × N part;

K (m-1) will be numbered in source images, k (m-1)+1 ... (k+1) region of (m-1) is added in intermediate image and is numbered km, km+1, (k+1) region of m-1, each overlapping region is launched by two projecting apparatus respectively; Wherein, N represents the integer being more than or equal to 2; M represents the integer being more than or equal to 2; D represents constant; K represents the arbitrary integer from 0 to N.

Image after merging splicing in step (4) carries out the detailed process of geometric correction, comprising:

Step (4.1): the fade factor determining image overlapping region;

Step (4.2): the method adopting GPU to accelerate obtains each pixel of image overlapping region, after the rgb value of each pixel obtaining image overlapping region is multiplied by its each self-corresponding fade factor, obtains new value and replaces original pixel value.

In step (4.1), if the brightness of image overlapping region is constant, then there is not overlapping region, fade factor is 1; Otherwise there is overlapping region, the fade factor of image overlapping region comprises right image overlapping region fade factor and left image overlapping region fade factor.

The expression formula of right image overlapping region fade factor is:

$f\left(x\right)=\left\{\begin{array}{c}\mathrm{\&omega;}{\left(2x\right)}^{\mathrm{\&upsi;}},0<x<0.5\\ 1-(1-\mathrm{\&omega;}){(2-2x)}^{\mathrm{\&upsi;}+2},0.5\&le;x\&le;1\end{array}\right.$

The expression formula of left image overlapping region fade factor is:

$f\left(x\right)=\left\{\begin{array}{c}1-(1-\mathrm{\&omega;}){\left(2x\right)}^{\mathrm{\&upsi;}+2},0<x<0.5\\ \mathrm{\&omega;}{(2-2x)}^{\mathrm{\&upsi;}},0.5\&le;x\&le;1\end{array}\right.$

Wherein, ω=0.7, υ=2.5, x represents current in left image or right image and is listed in overlapping region row number and the ratio of total columns of single overlapping region.

In step (4.2), the detailed process of the method that GPU accelerates is:

In GPU kernel function, first thread number is obtained, calculate the pixel (i corresponding to this thread, j) fade factor and corresponding to pixel (i, j), by pixel (i, j) fade factor corresponding to and pixel (i, j) rgb value is multiplied, as the pixel value that pixel (i, j) is final.

Special-shaped screen in step (1) and step (5) comprises arcuate screens, annular screen and spherical screen.

The special-shaped screen multi-projection system accelerated based on GPU of the present invention merges the displaying result of the method for video playback, as shown in Fig. 5 (b), by with correct before image graph 5 (a), can find out: the image overlapping region brightness before correction is comparatively large, and the brightness of the image overlapping region after correction is consistent with other regions; The image that image before correction two projecting apparatus are thrown does not align, and image has geometric deformation, and the texture of the image left images after correction achieves alignment, does not visually have geometric deformation.

The embodiment of step (5) is: the target image of Mat type is converted to the target image of CImage type to show in dialog box, complete fast at the time standby memcpy function changed, then use the draw function of MFC to show.

By reference to the accompanying drawings the specific embodiment of the present invention is described although above-mentioned; but not limiting the scope of the invention; one of ordinary skill in the art should be understood that; on the basis of technical scheme of the present invention, those skilled in the art do not need to pay various amendment or distortion that creative work can make still within protection scope of the present invention.

Claims (10)

1. the special-shaped screen multi-projection system accelerated based on GPU merges a method for video playback, it is characterized in that, comprising:

Step (1): at least two projecting apparatus are set, the number according to projecting apparatus generates control mesh, and special-shaped screen adjusts control mesh, the corresponding projecting apparatus of each group control mesh;

Step (2): read video file, obtains video file attributes; Read a two field picture of video file as source images; Adjustment control mesh carries out geometric correction to source images, and after the correction obtained, image is as target image, and asks for the mapping matrix of pixel in source images and target image;

Step (3): the next frame image reading video file; Shared by the overlapping region of adjacent projectors, the ratio of territory, whole projecting apparatus frame buffer zone size, generates a width intermediate image;

Step (4): the method adopting GPU to accelerate carries out brightness processed to intermediate image and image mosaic merges; Carrying out geometric correction according to the method that mapping matrix and the GPU of pixel in source images and target image accelerate to splicing the image after merging, obtaining the target image that splicing fused image is corresponding;

Step (5): the target image that projection splicing fused image is corresponding on special-shaped screen;

Step (6): repeat step (3) ~ step (5), until video playback completes.

2. a kind of special-shaped screen multi-projection system accelerated based on GPU merges the method for video playback as claimed in claim 1, and it is characterized in that, the detailed process of the adjustment control mesh in described step (1) is:

Generation control mesh is in full screen on special-shaped screen, then control mesh is increased or reduces one group of control mesh, the number increasing and decreasing grid line or the operation of mobile control point, make the control mesh that is projected on screen indeformable, and the alignment of the grid line of different projecting apparatus.

3. a kind of special-shaped screen multi-projection system accelerated based on GPU merges the method for video playback as claimed in claim 1, it is characterized in that, video file attributes in described step (1), comprising: the size of video frame rate, each frame and the total duration of video.

4. a kind of special-shaped screen multi-projection system accelerated based on GPU merges the method for video playback as claimed in claim 1, and it is characterized in that, the detailed process asking for the mapping matrix of pixel in source images and target image in described step (2) is:

Step (2.1): the mapping matrix of definition source images and target image is two-dimensional matrix M _xand M _y, two-dimensional matrix M _xand M _yall identical with the size of target image;

Step (2.2): the arbitrary pixel (i, j) in source images, obtains target image respective pixel (i', j') by the method adjusting control mesh, and at M _xmiddle preservation i', at M _ymiddle preservation j'; Wherein, i and j represents integer variable.

5. a kind of special-shaped screen multi-projection system accelerated based on GPU merges the method for video playback as claimed in claim 1, and it is characterized in that, the detailed process generating intermediate image in described step (3) is:

Suppose there is N number of projecting apparatus, the overlapping region of adjacent two projecting apparatus is single view field after correcting so first, source images is divided in the horizontal direction N (m-1)+1 part, every a width is d, and they are numbered 0,1 ..., N (m-1); Generate a panel height and wide intermediate image for d × m × N identical with source images, intermediate image is divided into m × N part;

K (m-1) will be numbered in source images, k (m-1)+1 ... (k+1) region of (m-1) is added in intermediate image and is numbered km, km+1, (k+1) region of m-1, each overlapping region is launched by two projecting apparatus respectively; Wherein, N represents the integer being more than or equal to 2; M represents the integer being more than or equal to 2; D represents constant; K represents the arbitrary integer from 0 to N.

6. a kind of special-shaped screen multi-projection system accelerated based on GPU merges the method for video playback as claimed in claim 1, it is characterized in that, the image after merging splicing in described step (4) carries out the detailed process of geometric correction, comprising:

Step (4.1): the fade factor determining image overlapping region;

Step (4.2): the method adopting GPU to accelerate obtains each pixel of image overlapping region, after the rgb value of each pixel obtaining image overlapping region is multiplied by its each self-corresponding fade factor, obtains new value and replaces original pixel value.

7. a kind of special-shaped screen multi-projection system accelerated based on GPU merges the method for video playback as claimed in claim 6, it is characterized in that, in described step (4.1), if the brightness of image overlapping region is constant, then there is not overlapping region, fade factor is 1; Otherwise there is overlapping region, the fade factor of image overlapping region comprises right image overlapping region fade factor and left image overlapping region fade factor.

8. a kind of special-shaped screen multi-projection system accelerated based on GPU merges the method for video playback as claimed in claim 7, and it is characterized in that, the expression formula of described right image overlapping region fade factor is:

$f\left(x\right)=\left\{\begin{array}{c}\mathrm{\&omega;}{\left(2x\right)}^{\mathrm{\&upsi;}},0<x<0.5\\ 1-(1-\mathrm{\&omega;}){(2-2x)}^{\mathrm{\&upsi;}+2},0.5\&le;x\&le;1\end{array}\right.$

The expression formula of described left image overlapping region fade factor is:

$f\left(x\right)=\left\{\begin{array}{c}1-(1-\mathrm{\&omega;}){\left(2x\right)}^{\mathrm{\&upsi;}+2},0<x<0.5\\ \mathrm{\&omega;}{(2-2x)}^{\mathrm{\&upsi;}},0.5\&le;x\&le;1\end{array}\right.$

Wherein, ω=0.7, υ=2.5, x represents current in left image or right image and is listed in overlapping region row number and the ratio of total columns of single overlapping region.

9. a kind of special-shaped screen multi-projection system accelerated based on GPU merges the method for video playback as claimed in claim 6, and it is characterized in that, in described step (4.2), the detailed process of GPU accelerated method is:

In GPU kernel function, first thread number is obtained, calculate the pixel (i corresponding to this thread, j) fade factor and corresponding to pixel (i, j), by pixel (i, j) fade factor corresponding to and pixel (i, j) rgb value is multiplied, as the pixel value that pixel (i, j) is final.

10. a kind of special-shaped screen multi-projection system accelerated based on GPU merges the method for video playback as claimed in claim 1, it is characterized in that, the special-shaped screen in described step (1) and step (5) comprises arcuate screens, annular screen and spherical screen.