JPH0385688A - Stereographic image display system - Google Patents

Abstract

PURPOSE:To accelerate an image generation processing by performing the asynchronous decentralized processing of a processing unit stereographic part with each processing part, and the centralized processing of common image area partial image data obtained at each processing part with a correspondence processing part. CONSTITUTION:The decentralized processing of respective processing unit stereographic part processed at each processing part 6i by using the initial data of a three-dimensional body inputted from an input part 2 to each processing part 6i is performed between each processing part 6i via a network 10. The common image area partial image data obtained by being processed at each processing part is written on a corresponding image data storage part 8j. In parallel with the write of the data, output with prescribed display time series from each image data storage part 8i can be taken out by supplying output control with an output control circuit 14 to a multiplexer 12. A picture element data string outputted via the multiplexer 12 is supplied to a display device 4, then, the three-dimensional body can be displayed on a screen. In such a way, image data can be generated at high speed.

Description

[Detailed Description of the Invention] [Table of Contents] [Overview] [Prior Art] Fig. 13 A diagram showing the processing flow in the system (Fig. 14)
[Problem to be solved by the invention] [Means for solving the problem] [Operation] [Example] A diagram showing an embodiment of the present invention (Fig. 2), a configuration diagram of a processor (Fig. 3) ), Network configuration diagram (Figure 4), Diagram showing the image output of each PE and control circuit system of the multiplexer (Figure 5), Diagram showing the system processing flow (Figure 6), Processing of each PE Diagram showing the flow (Fig. 7), [Effects of the invention] [Summary] A three-dimensional object that asynchronously uses distributed processing of each processing unit three-dimensional part of a three-dimensional object and intensive processing of a common image area part for display control. Regarding the image display system, for the purpose of high-speed image generation by asynchronous distributed processing etc. for the processing unit solid part of the three-dimensional object, an input section for inputting initial data of the three-dimensional object, a display device, and an input section connected to the same unit are used. a plurality of processing units, an image data storage unit corresponding to the plurality of processing units;
A network provided for asynchronous data transfer between a plurality of processing units, a multiplexer, and a plurality of image data storage units that output images from the plurality of image data storage units to the display device via the multiplexer in synchronization with the display of the display device at a predetermined time. It is equipped with an output control circuit that outputs the output in series, and the processing for each processing unit object part of the three-dimensional object to be displayed on the screen of the display device is distributed to each processing unit via a network, and the processed For image area parts that are common in terms of display control between processing unit three-dimensional parts, common image area partial image data obtained by transfer processing via a network to a processing unit allocated in advance to the common image area part is handled. The image data is written in an image data storage section and displayed on the screen of a display device via a multiplexer.

[Industrial application field]

The present invention provides asynchronous distributed processing by a plurality of processors for each processing unit object portion of a three-dimensional object, and processing by a processor assigned in advance to the common image region portion for display control. The present invention relates to a three-dimensional image display system using a three-dimensional image display system.

The image actually displayed on the display screen of a display device in conventional CAD or computer animation is created by applying coordinate transformation processing, clipping processing, hidden surface processing, and filling processing (shading processing) to model data of a three-dimensional object. The obtained image data is displayed on the display screen by being applied to the display device under the control of the display control device. Therefore, since the image displayed on the screen requires various types of processing as described above, it is unavoidable that the processing time is included in the display.

In addition, the image display involves the processing of thousands of times per second, at time intervals that are imperceptible to the naked eye. Therefore, in order to avoid any resistance to conversational operations in an image generation system, it is important for the system to immediately return a processing response to the conversational operations when building this type of system. This is an important matter.

[Conventional technology]

In order to meet the above-mentioned requirements, three-dimensional image generation systems that satisfy the above-mentioned ideas have been developed in the past. The system has a configuration as shown in FIG. 13, and handles graphic processing such as coordinate transformation and clipping,
A dedicated processor is used for pixel processing such as hidden surface processing and fill-in processing to improve the responsiveness of the aforementioned processes. The outline of image generation in this system is explained as follows. The main processor 50 generates a display list (see 104 in FIG. 14) determined by a graphics program (see 102 in FIG. 14) installed therein according to a drawing request from an application program (see 100 in FIG. 14). In addition to editing,
It also transfers data to the graphics processor 54, communicates with the outside, and controls the entire system. Although the main processor 50 has complicated processing, it is not required to be as fast as the graphics processor 54 and drawing processor 56 described later, so a general-purpose microprocessor is used. The display list is often a hierarchical data structure and may be stored in main memory 52 of main processor 50 or in dedicated memory. The reason why it is a hierarchical data structure is that it allows the description of complex objects and its shape can be easily changed. A three-dimensional object in that description is a large number of simple figures (
expressed by combining parts). Each part is represented by expression data completed in a coordinate system unique to each part.

The graph ink processor 54 performs coordinate conversion processing (see 106 in FIG. 14) and clipping processing (see 108 in FIG. 14) for data in the display list given from the main processor 50 via the bus 51.
A character 1 figure is generated (see 110 in FIG. 14).

Coordinate transformations include modeling transformation, viewing transformation, perspective transformation, and the like. As mentioned above, modeling transformation involves expressing a three-dimensional object with the many parts necessary for it, and by scaling and rotating each of those parts, the three-dimensional object is expressed within the world coordinate system. This is the coordinate transformation required to assemble the original object.

A three-dimensional object assembled within that world coordinate system,
so that it can be viewed on the screen as a three-dimensional object in an at-a-glance display mode that would be visible if the three-dimensional object were viewed from any position and direction in the world coordinate system;
Viewing transformation converts the coordinates of object data described in the world coordinate system to coordinates in the viewpoint coordinate system.

In this viewing transformation, a viewpoint and a line-of-sight direction within the world coordinate system are set, and the coordinates of each component are coordinate-transformed into the viewpoint coordinate system using these as a reference. Perspective transformation is the projection of a three-dimensional object in this viewpoint coordinate system onto an appropriate projection plane viewed from the viewpoint and viewpoint direction used for the transformation.

Through this perspective transformation, three-dimensional object data expressed in the viewpoint coordinate system is transformed into two-dimensional graphic data on the projection plane.

The clipping process is a process of cutting out only figure data included within a designated display area from within the two-dimensional figure of the three-dimensional object data projected onto the projection plane.

After the clipping process in the graphics processor 54, the graphic data is taken into the drawing processor 56 via the bus 51 etc. in order to perform hidden surface processing and shading processing on the three-dimensional object region. Hidden surface processing is a process that uses the graphic data to avoid displaying hidden surface areas on the surface of a three-dimensional object when viewed from the above-mentioned viewpoint and line-of-sight direction. This is a process of shading a surface using the brightness determined by the orientation of the object surface and the direction of the light beam. Shading processing includes constant shading, which fills one surface with the same brightness, and smooth shading, which fills one surface by interpolating brightness and color. Both the drawing processor 56 and the graphic processor 54 have relatively simple processing calculations, but the amount of data involved in the processing increases and the time required for the calculation increases. Since a lack of responsiveness must be avoided, high-speed processing is required, and both of these processors are dedicated processors controlled by a microprogram.

Each pixel data of the display area within the projection plane, which has been generated through the hidden surface processing and filling processing as described above, is written into the frame buffer 58.

As mentioned above, the fact that both the graphics processor 54 and the drawing processor 56 are configured in a tightly coupled type in which they are directly connected to the bus 51 of the main processor 50 is that the transfer time between the processors acts as a negative factor on the processing responsiveness mentioned above. This is one way to eliminate this problem. As another such device, the main processor 50 and the graphics processor 54 and the graphics processor 54 and the drawing processor 56 are directly connected by high-speed hubs 53, 55, and 57.

The graphic data written in the frame buffer 58 is read out by the video controller 60 and is read out by the conversion device (
LUT) 62 converts the signal into an analog video signal and displays it on a display device 64.

Data is read from frame buffer 58 at a rate of 30 to 70 times per second to eliminate flicker on both sides. Because display on high resolution display device 64 requires extremely high data transfer rates, video controller 60. Various improvements have been made to the frame buffer 58 as well.

[Problem to be solved by the invention]

In the conventional system mentioned above, the system cannot be constructed unless it includes dedicated hardware.
Not only does it take a long time to develop, but the cost of developing it is also high. Various hardware (processors) in the system developed in this way
Since the performance of the processors differs, it is difficult to balance the performance among these hardwares, and data transfer between each processor tends to be a factor that inhibits performance improvement. In a system that includes these various technical challenges, if you want to build the system by giving all processors parallel processing capability by some means to further improve the performance, then Further complication cannot be avoided.

Furthermore, in the aforementioned coordinate transformation for displaying curved surfaces (circles, ellipses) and brightness calculations at vertices, a method of approximating the curved surface with a large number of minute polygons is adopted as a method suitable for the system configuration described above. , if you want to display a smooth surface, it is necessary to approximate it with at least several dozen micro polygons. Therefore, in the design of molecular models that require the display of dozens of spheres, simulations of molecular dynamics, etc., the inefficiency of the calculation processing becomes apparent in the system configuration described above. .

The present invention was created in view of such technical problems, and provides a three-dimensional image display system that can produce image data at high speed through asynchronous distributed processing for each processing unit object part in image data generation. Its purpose is to provide.

[Means to solve the problem]

FIG. 1 shows a block diagram of the principle of the present invention. As shown in this figure, the present invention includes an input section 2 for inputting initial data of a three-dimensional object, a display device 4, connected to the input section 2,
A plurality of processing units Elz having the same configuration perform processing for a processing unit solid portion of a three-dimensional object to be displayed on the screen of the display device 4, and processing for a common image area portion for display control.
(i=12, . . . , n) and the plurality of processing units 6.
Corresponding image data storage unit 8. and a network 10 for asynchronous data transfer between the plurality of processing units 6.
and a multiplexer 12, and output control for outputting the image outputs of the plurality of image data storage units 381 to the display device 4 in a predetermined display time series via the multiplexer 12 in synchronization with the display on the display device 4. A circuit 14 is provided. Then, processing for each processing unit solid portion of the three-dimensional object to be displayed on the screen of the display device 4 is sent to each processing unit 68 through the network 1.
0, and the processing of an image area portion that is common in terms of display control between the processed processing unit three-dimensional portions is sent via the network 10 to a processing unit that has been allocated in advance to the common image area portion. Common image area partial image data obtained by processing with or without transfer is written into the corresponding image data storage section and displayed on the screen of the display device 4 via the multiplexer 12. configured.

[For production]

For each of the processing unit three-dimensional parts processed in each processing section 6i using the initial data of the three-dimensional object input from the input section 2 to each processing section 68, the processing is carried out by the network 8 between each processing section 68. Distributed processing via . Processing for image area parts that are common in terms of display control between the processing unit three-dimensional parts is carried out by the network 1.
0 (j is one of i, which represents the processing unit previously allocated to the common image area).
If the processing for the common image area is performed by the own processing section, processing is performed without transferring.

In this way, the common image area partial image data obtained by processing in each processing section is written into the corresponding image data storage section 8J.

In parallel with such writing to the image data storage section 8j, each image data storage section 8. The output in a predetermined display time series is produced by applying output control to the multiplexer 12 by the output control circuit 14, and the pixel data string outputted via the multiplexer 12 is supplied to the display device 10. By doing so, a three-dimensional object is displayed on both sides.

As mentioned above, each processing unit 4 performs asynchronous distributed processing on the processing unit solid part of the three-dimensional object to be displayed, and the common image area partial image data obtained by processing in each processing unit is As for processing, since the corresponding processing unit performs centralized processing, the image generation process becomes faster without the use of dedicated hardware, improving processing responsiveness on the system side and improving system constructability. Become.

〔Example〕

FIG. 2 shows an embodiment of the invention. In this figure, host computer 2. is connected to the processor (PE) via bus 22.
I)24. to processor (PEn) 24. connected to. The processors 24+ to 24I are connected to each other via a network 10, and via a multiplexer 12 and a video controller 26 to a raster format color CRT display device (hereinafter simply referred to as a display device).

) 41.

Processor 24. The processors 241+ are configured as shown in FIG. The processors 24 to 24I all have a CPU 30 and an internal bus 32.
, memory 34, host-PE communication interface 36
i-network communication interface 3 days, image memory 4 days
0 and an address counter 42. In FIG. 3, only the reference numbers are used without adding subscripts to distinguish the processors. The host-PE communication interface 36 is connected to the host computer 21 via the bus 22. The memory 34 stores a program that executes the processing flow shown in FIG. Network communication interface 38 is connected to network 10 .

The output of image memory 40 is connected to multiplexer 12, and address counter 42 receives synchronization signals and a clock from video controller 26 (see also FIG. 5). or,
The image memory 40 stores pixel data corresponding to the screen.

The network 10 may be configured in any format shown in (1) and (2) in FIG. The requirement is to be able to provide asynchronous communication between each processor.

Multiplexer 12, in response to a selection signal from video controller 26, connects processors 24 . to processor 24. , and outputs the pixel data from the image memory 401 to the video controller 26 as a pixel data string.

2 to 6, the host computer 2I corresponds to the input section 2 of FIG. 1, the processor 24° corresponds to the processing section 68 of FIG. 1, and the image memory 40 stores the image data of FIG. It corresponds to the storage section 88.

Video controller 26 corresponds to output control circuit 14 in FIG.

The operation of the above-mentioned system of the present invention will be explained below.

When the system starts operating, initial data is transferred from the host computer 2 to each PE (see S1 in Figure 6).
. The initial data transferred to each PE is transferred to the host computer 21
It is decided in. The initial data are constants for the object 8 to be displayed on the screen of the display device 4 (object definition data, various parameters such as the line of sight direction for viewing the object, and the position of the line of sight). Assume that the object to be displayed is, for example, a molecular model represented by spherical steric bonds.

When the initial data is written to the memory 34 via the host-PE interface 36, the application program of each PE performs calculation processing on the initial data (see 31 in FIG. 7) and generates object data. (See 32 in Figure 6 and 32 in Figure 7). Assuming that the object to be displayed on the screen is a molecular model represented by three-dimensional bonds of spheres (atoms) as described above, this object data is three-dimensional data representing each sphere. This three-dimensional data includes data such as the types and positions of atoms in the steric bonds of the molecular model. The application programs of each PE that generate this object data proceed with their processing independently.

The generated object data is distributed to the span decomposition tasks of all PEs via the network 10. This distribution is performed in order, for example, from PEI to PEn.

The application tasks within each PE and between PEs, the span decomposition task that receives the distributed object data, and the filling task described below all perform their task processing asynchronously. Object data that remains unprocessed in the decomposition task is distributed again after distribution. If it is assumed that the amount of object data generated by the application program of each PE is almost the same for all PEs, the distribution is performed only for the span decomposition task of the own PE.

(Margins below this page) 0 The object data distributed to the span decomposition tasks of each PE (see Sll in FIG. 7) is coordinate-transformed according to the viewpoint data contained therein. This coordinate transformation is performed by multiplying both the coordinates of the position of the center of the sphere and an arbitrary point on the pope by the following matrix. In the matrix, A is a rotation parameter, B is a projection parameter, C is a translation parameter, and D is a scaling parameter.

In addition, clipping processing is performed in the same manner as in the prior art. This clipping process is performed by multiplying both the coordinates of the center position of the sphere whose coordinates have been transformed and the coordinates of an arbitrary point on the pope by the following matrix 1. A perspective transformation (according to the conventional technique described above) is then performed as necessary.

After these conversions, span data is generated for the sphere (object data) that is sequentially processed in the span decomposition task. Generation of this span data (first
(See 312 in Fig. 3) is a section plane when the method is cut by the plane subjected to the clipping process in the viewpoint coordinate system, and the section planes are sequentially cut at the resolution of the generated image as shown in Fig. 9. The intersection line (chord) formed when cutting is the span referred to in the present invention, and the span data is determined from the following data. Span start and end positions (PL
, PR), the radius of the sphere, the coordinates of its center position, and the color information of the starting position (display color of the method)2, and its PLXPR is expressed by the following equation.

PL-(x, Y-Yl, Z) PR= (x, Y+Y1.Z) However, Yl in the above formula is Yl-Katana Goka V Near Door 7. This span data is transferred via the network 10 to the filling task of each PE, which is uniquely determined as described below.

Each of the span data obtained by decomposing each two-dimensional figure onto the projection plane into spans as described above is sent from the PE in charge of the decomposition processing to the screen as a PE in charge of the same span data string via the network 10. It is transferred to the filling task of the PE, which is uniquely determined from the generation. This is illustrated in FIGS. 10 and 11. These figures show the development of the image memory contents corresponding to the display device screen, and the image data displayed on the display device screen is divided into vertical lines as shown in FIG. Decide in advance. 3. Assuming that the reading direction of the image memory is perpendicular to the screen task, assign vertical lines to each PE (as shown in the figure, the first vertical line from the left is the PEO, the second vertical line is the PEO, etc.) is like assigning a PEI). When the reading direction of the image memory and the screen task match, instead of the vertical line shown in FIG. 10, a horizontal line perpendicular thereto is assigned to each PE. For lines in either direction, the PE in charge of that line is uniquely determined from the above if screen generation is a task method, and once determined, there is a fixed line between each line and the PE. Become a relationship. Therefore, due to this relationship, the span data generated by each PE in charge of disassembly processing and located on the same line as described above is transferred only to the PE in charge of the same line via the Hennessoto work 10. (See 313 in Figure 7).

In this way, each of the span data corresponding to the filling processing line is transferred to the filling task of each P'E (see 321 in Fig. 7), and each filling 4 screen is divided into pixels according to the shading model for the span data. Interpolation of brightness values and filling of the image memory are performed (see 322 in FIG. 7). The first example shows this situation.
This is Figure 2.

FIG. 12 shows an example of filling in a sphere in a molecular model, etc., as described above. In this figure, for convenience of explanation, one sphere is shown on the viewpoint coordinate system, which is illustrated by rotating the screen corresponding to development shown in FIG. 12 by 90 degrees. As mentioned above, the span data SD for this ball is the p E i of the line that includes the span data SD.
(The meaning of J is the same as the explanation in the [Operation] section.) is transferred via the network 10 from the span decomposition task described above. The span decomposition task is performed using other P
The entire span decomposition task of H may include that of the self PE, or the entire span decomposition task of H may not include the span decomposition task of the self PE.

The span data SD is the span end position PL (=Y-Y
1), PR(-Y+Y1), sphere radius r,
5 The center position ○ coordinates (x, y, z) are determined from the display color of the sphere, and from the span decomposition task of the PE in charge of the sphere shown in FIG. It will be forwarded to PEJ who is in charge of filling it out. Each point P ( Find the position of x, y, z). Each of these points P is on the same coordinates for X, y
is the value of (x, y, z) when the value of the Y coordinate is changed by a variation determined by the resolution of the image. The second direct is PL - (x, Y-Yl, Z),
Since PR= (xy+Y1.Z), it becomes. The normal vector N at each of these points P is given as N=(x-X, y-Y, z-Z) since its surface is a spherical surface.

Using this normal vector N and each data set in advance at the start of processing regarding the position and direction of the light source that irradiates light to the three-dimensional object (described above) projected on the projection plane, a phono model is created. Sphere filling process according to (
shading processing) (see 322 in FIG. 7). The brightness ■ at each point P for this filling process is
Using the normalized normal vector N' and the normalized ray direction vector L for each point, I=k
・It is obtained as (N' ・L) +a. However, in this second case, a is a constant for display.

If the sphere cut by the line segment represented by the span data SD is filled in according to the luminance I at each point P of the nuclear sphere while performing hidden surface processing, the span data SD in the filling process is performed. The processing for the display line segment defined by is completed.

Each display line segment for the ball existing on both sides of this display line segment is displayed by the PE in charge of the line where each display line segment is located.
, the same fill-in process as that performed in the PEJ described above is carried out concurrently.

7 For example, a Z-buffer algorithm is used for the hidden surface processing described above. The Z buffer algorithm performs hidden surface processing using a memory (Z buffer) that stores the distance from the viewpoint corresponding to each pixel of image data. Clear the Z buffer (i.e., set it to the value at infinity) when you start drawing.

When writing an input pixel of a figure, the distance Z value of the pixel is compared with the distance Z value of the pixel corresponding to the input pixel in the Z buffer, and the input pixel is determined to be within the Z buffer corresponding to the input pixel. If the input pixel is in front of the pixel, it is determined that the input pixel is visible from the viewpoint, and the input pixel is written to the pixel storage location in the Z buffer that is the comparison target. Hidden surface processing is performed by repeating processing for each input pixel in a figure without performing update processing.

As is clear from the above, the processing of each task (applied task, span decomposition task, and fill-in task) is progressed asynchronously within each PE and between PEs, and each PE has 8 The processing priority of the tasks is set to decrease in the order of filling tasks, span decomposition tasks, and application tasks, so even if there is a bias in the amount of object data generated in the application tasks of each PE, Since each PE is asynchronously distributed to all span decomposition tasks in the system including its own span decomposition task, the processing of the entire system is completed in relation to the processing priority (see 33 in Figure 6).
By then, the distribution of the object data to each PE has been equalized.

While such object data is being distributed, the filling process for span data is also sequentially performed by the P in charge of each line.
This is done in parallel with E.

The pixel data subjected to the filling process is sequentially written as described above to the storage locations corresponding to the display screen pixels in the image memory of the above-mentioned line PE to which the pixel data belongs.

Assuming that the readout direction of each image memory is diagrammatically shown as a straight chain line SL in FIG. 11, the readout of each image memory is exactly -straight chain, %9! When reaching SL, the read address for the image memory of each vertical line PE is output from the address counter for the image memory, and is as follows. Said-
Dianzhen #i! The image memory read address corresponding to the intersection of the - dotted chain line SL and the first vertical line is outputted from the corresponding address counter to the image memory of the PE in charge of the first vertical line on the left side on SL, and the second PE is read out at the next read clock. The image memory read address corresponding to the intersection of the -dot chain line SL and the second vertical line is output to the image memory of the PE in charge of the th vertical line from the address count of the PE in charge of the line. The same applies to sequential read clocks.

In this way, by the time the reading corresponding to the -dotted chain line SL is completed from the image memory of the PH in charge of the leftmost vertical line to the image memory of the PE in charge of the rightmost vertical line, the -dotted chain line S
Each pixel data of the pixel data string for the L-compatible display device task is sequentially output to each pixel data output from the image memory of the PE in charge of the leftmost vertical line to the image memory of the PE in charge of the rightmost vertical line. . Each of the pixel data sequentially outputted is supplied from the multiplexer 12 to the display device 4 via the video controller 26 in response to a selection signal sequentially supplied from the video controller 26 to the multiplexer 12, and is displayed on the screen of the display device 4. Display the tasks corresponding to the chain line SL. Such task-type reading is also performed sequentially in reading corresponding to other tasks. The above reading is then repeated a predetermined number of times, for example 30 or 60 times per second, depending on whether the display device is of an interlaced scanning type or not.

In the task-type readout for each image memory that has been filled in as described above, the readout speed of the image memory is reduced to 1/the number of image memories of the pixel data clock in the display device due to the multiple readout described above.

In the above embodiments, an example in which the three-dimensional object is formed by a three-dimensional combination of spheres has been described, but the three-dimensional object may have another single shape.

Further, the image memory only needs to have a capacity for at least three lines handled by the PE, and does not necessarily have a capacity for one screen. In this case, the readout control will naturally be different from that described above. In any case, the image memory only needs to be compatible with the filling task of the PE, and does not necessarily have to be provided in the PE.

The present invention may be practiced even if the system is configured such that initial data input is provided via a communication line.

〔Effect of the invention〕

As is clear from the above description, according to the present invention,
Multiple processors with the same configuration are used to process the data necessary for displaying the three-dimensional object that you want to display, and each processor is responsible for one processor that processes the initial data about the three-dimensional object that you want to display. , or generates object data representing two or more processing unit three-dimensional parts, and sends those object data to all two processors, including or excluding the processor that generated the object data, to generate span data. While distributing it asynchronously, each processor performs the filling process using the span data assigned to it, thereby achieving asynchronous distributed processing of the processing of the processing unit 3D part and asynchronous concentration of the filling process. Since the data processing necessary for displaying the three-dimensional object to be displayed is achieved through processing, there is no need for dedicated hardware to obtain image data about the three-dimensional object to be displayed, and the three-dimensional object to be displayed can be displayed. It is possible to enjoy high-speed drawing of a three-dimensional object (immediate operational responsiveness) by avoiding slowdown of processing due to division into a large number of small polygons and by making it easy to equalize the load on each processor.

Parallel processing of data necessary for displaying a three-dimensional object to be displayed by each of the processors improves the utilization rate of the processors. Furthermore, it is advantageous for high-resolution display devices to be able to multiplexly read image data about a three-dimensional object to be displayed that is being written to the image memory of each processor or has been processed, from each image memory. 3. It can give the system a great deal of adaptability.

[Brief explanation of drawings]

Figure 1 is a block diagram of the principle of the present invention, Figure 2 is a diagram showing an embodiment of the present invention, Figure 3 is a configuration diagram of a processor, Figure 4 is a configuration diagram of a network, and Figure 5 is a diagram of each PE. Figure 6 is a diagram showing the image output and multiplexer control circuit system, Figure 6 is a diagram showing the system processing flow, Figure 7 is a diagram showing each P
Figure 8 shows an example of the distribution of object data to PEs, Figure 9 is an explanatory diagram of span data generation, and Figure 10 shows the allocation to each PE to the filling processing line. Figure 11 is a diagram showing only the filling process line assigned to PEO in the distribution example of Figure 8, Figure 12 is an explanatory diagram of the filling process for the example of the filling process line, and Figure 13 is a conventional three-dimensional diagram. FIG. 4 is a diagram showing the image generation system. FIG. 14 is a diagram showing the processing flow in the system shown in FIG. 1 to 6, 2 is an input unit (host computer 2), 4 is a display device, 6i is a processing unit (processor 248), 8L is an image data storage unit (image memo 1J40i), 10
is a network, 12 is a multiplexer, and 14 is an output control circuit (video controller 26). Abrig-Siri/Task JP-A-3 85688 (13) Span #pI Kusatastan t Suyoru 2 L7 Tas 7 So PE's Numarifu D- Figure 9 Labor Yi 4- To Keyu's PE ○ Zuke 13ro Pig q fig.

Claims (1)

[Claims] (1) Input section for inputting initial data of a three-dimensional object (2)
, a display device (4), and an image area that is connected to the input section (2) and is common for processing and display control of the processing unit stereoscopic part of the three-dimensional object to be displayed on the screen of the display device (4). A plurality of processing units (6_i) (i=1, 2,
..., n), an image data storage unit (8_i) corresponding to the plurality of processing units (6_i), and a network (10) provided for asynchronous data transfer between the plurality of processing units (6_i). , multiplexer (
12), the image outputs of the plurality of image data storage units (8_i) are sent to the display device (4) via the multiplexer (12) in a predetermined display time series in synchronization with the display of the display device (4); and an output control circuit (14) for outputting the output, and the processing for each processing unit three-dimensional part of the three-dimensional object to be displayed is distributed to each processing unit (6_i) via the network (10). The processing of the image area portion that is common in terms of display control between the processed processing unit three-dimensional portions is transferred, or not transferred, to a processing unit previously allocated to the common image area portion, via the network (10). The common image area partial image data obtained by the above processing is written into the corresponding image data storage section, and is sent to the display device (4) via the multiplexer (12) under the control of the output control circuit (14). A stereoscopic image display system characterized by displaying images on a screen.