JPH0973561A - Stereoscopic image generating device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate a correct stereoscopic image even when an object having a complicated shape like a tree is deformed and an intricate part is exposed. SOLUTION: This device is provided with plural simplification models 281 having precision corresponding to a distance to the object from an observer's viewpoint based on a fractal model, a simplification model selection part 241 for previously generating plural partial textures for each simplification model by projecting a part where the fractal model corresponds to each element constituting each simplification model 281 and selecting corresponding one from among the simplificaiton models 281 in accordance with the sight line data, a partial texture selection part 242 for selecting the plural partial textures 301 corresponding to the selected simplification model in accordance with the viewpoint data, and a rendering processing part 243 for mapping the selected plural partial textures to the selected simplification model.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a stereoscopic image generating apparatus, and more particularly to a stereoscopic image generating apparatus for generating a stereoscopic image according to a viewer's viewpoint by using fractal geometry.

[0002]

2. Description of the Related Art In the field of computer graphics, there is a motion vision display system that defines an object with a three-dimensional model, measures the viewpoint position of a user, and generates and displays an image viewed through a screen from that viewpoint. Being developed. Also, a system for simulating a natural landscape has been developed. Furthermore, using wide-field, high-definition, stereoscopic, and kinetic vision, immersive communication has been developed that can provide the feeling that people in remote areas are in the same space.

In computer graphics, when an object to be displayed is defined by a three-dimensional model and a high-quality image is generated, the object is finely sampled, light source calculation is performed, and perspective display is performed. Realistic images can be obtained. Furthermore, if stereoscopic display, motion visual display, and the like are performed, the stereoscopic effect of the displayed image becomes higher. When generating such a high-quality image, the number of required display model vertices becomes enormous, which increases the amount of calculation such as rendering processing and projection conversion processing, which tends to increase the processing time. .

The above-described presence-oriented communication conference system is required to generate and display a high quality image in real time. Therefore, various researches such as improving the processing speed of the computer, developing high-speed algorithms for various processing such as light source calculation, creating a model with vertex removed using visual characteristics, and displaying using that model have been promoted. There is.

By the way, in order to display a shape with a huge amount of data such as a tree, it is necessary to delete the number of vertices of the display model by utilizing visual characteristics. However, it has been impossible to reduce the amount of data to the extent that an image can be displayed in real time with the conventionally proposed method. In addition, a method has been proposed in which a texture obtained by projecting a three-dimensional model in parallel is registered in advance and the texture is mapped in real time on a polygon composed of a plurality of intersecting planes. The image is distorted. Furthermore, with this method, it was not possible to generate in real time a deformation that complies with the physical laws such as fluctuation due to wind.

On the other hand, the Institute of Electronics, Information and Communication Engineers, D-II, Vol. J78-T-I
I, No. 7th pages 1091 to 1104, "a hierarchical tree using fractals that displays a two-lens stereoscopic image without distortion with respect to viewpoint movement and can be deformed in real time according to a physical law. A high-speed method for generating a three-dimensional tree image using shape representation "has been proposed. This method models an object with a complicated shape such as a tree using fractal geometry, obtains sufficient roughness for display according to the distance from the observer's viewpoint to the object, and determines the roughness. Accordingly, a simplified model is generated, and the texture of the tree image from an appropriate viewpoint is further mapped to the simplified model.

[0007]

However, in this method, since the texture information on the intricate portion of the tree is not prepared, a complete stereoscopic image cannot be displayed when the intricate portion is exposed due to deformation. was there. Further, there is a problem that about 18 tree images are required to display a distortion-free twin-lens stereoscopic image with respect to the viewpoint movement over the entire circumference of the tree of 360 °. It should be noted that one tree image at this time can handle a viewpoint movement of about 20 °.

Therefore, it is an object of the present invention to provide a stereoscopic image generating apparatus capable of generating an accurate stereoscopic image even when an object having a complicated shape is deformed and its intricate portion is exposed.

[0009]

According to the present invention, a stereoscopic image generating apparatus for generating a stereoscopic image in accordance with an observer's viewpoint models an object to be generated as a stereoscopic image using fractal geometry. A fractal model generating means for generating a fractal model, and a simplified model generating means for generating, based on the fractal model, a plurality of simplified models having a definition according to the distance from the observer's viewpoint to the object. , A partial texture generation means for generating a plurality of partial textures for each simplified model by projecting the corresponding portion of the fractal model onto each of the elements constituting each simplified model, and an observation given from the outside A simplified model selecting means for selecting a corresponding one of the simplified models according to the viewpoint data relating to the viewpoint of the user; Comprises a portion texture selecting means for selecting a plurality of portions texture corresponding to the selected simplified models according to the data, and a mapping means for mapping the plurality of partial textures selected by the selected simplified models.

Therefore, according to the present invention, even if the intricate portion of the object is exposed by the deformation according to the physical law, a plurality of partial textures are individually provided for all the elements constituting each simplified model. Since it is generated, the complete twin-lens stereoscopic image can be always displayed without the exposed intricate parts being lost, and the entire circumference of the tree can be displayed as compared with the conventional method described above. Therefore, the amount of data required can be significantly reduced.

[0011]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a block diagram showing the overall configuration of a motion vision display system including an image generation unit according to the present invention. Referring to FIG. 1, the motion vision display system includes a screen 12 for displaying an object 14.
And a magnetic sensor 20 for detecting the viewpoint of the observer 18.
A magnetic field generator 16 that generates a predetermined magnetic field, an image generator 24 that generates an image of the object 14 to be displayed on the screen 12, a magnetic field generator 16 and an image generator 2.
4 controls the observer 1 in response to the output of the magnetic sensor 20.
The control unit 22 that supplies the viewpoint data regarding the eight viewpoints to the image generation unit 24.

The three-dimensional world coordinate system in this motion visual display system is defined by the xyz coordinate system 10. First, the screen 12 and the object 14 in this three-dimensional world coordinate system
And the three-dimensional position of the magnetic field generator 16 is obtained in advance.
A magnetic sensor 20 is attached to the head of the observer 18. The magnetic sensor 20 detects the magnetic field generated by the magnetic field generation unit 16, and thereby the distance from the magnetic field generation unit 16 to the viewpoint of the observer 18 is obtained in real time.
In response to the output of the magnetic sensor 20, the viewpoint data regarding the viewpoint of the observer 18 is supplied from the controller 22 to the image generator 24.

In the embodiment shown here, the magnetic sensor 20 is used to detect the viewpoint of the observer 18, but the method for detecting the position of the viewpoint is not limited by this, and the viewpoint is not limited. Any method of detecting the position can be adopted.

When the viewpoint of the observer 18, the positional relationship between the screen 12 and the object 14 is thus determined, the perspective projection image of the object 14 observed from the viewpoint through the screen 12 is generated by the image generator 24. Further, it is displayed on the screen 12. For example, observer 18
When is moved to the right, the right side of the object 14 that can be observed from the new right side viewpoint is displayed on the screen 12. On the other hand, when the observer 18 moves to the left, the left side of the object 14 that can be observed from the new left side viewpoint is displayed on the screen 12.

As described above, in order to display a stereoscopic image in real time according to the viewpoint of the observer, a model serving as a skeleton of the object 14 to be displayed, a texture to be attached to the model, and the like are specified. It is necessary to prepare in advance. FIG. 2 is a block diagram showing an apparatus for generating a model and a texture necessary for displaying an object having a complicated shape such as a tree. With reference to FIG. 2, this apparatus includes a fractal model generation unit 261 that generates a fractal model by modeling an object having a complicated shape such as a tree using fractal geometry, and the generated fractal model. Hierarchical simplified model generation unit 262 that generates a plurality of simplified models 281 based on
And a hierarchical partial texture generation unit that generates a plurality of partial textures 301 for each simplified model by projecting corresponding portions of the generated fractal model onto each of the elements that form each simplified model. 263
And a texture management table generation unit 264 that generates a texture management table 302 for managing which element in which simplified model the generated partial texture corresponds to, and the generated simplified model. A database 28 that stores 281 and the generated partial texture 301 and management table 30.
2 and a database 30 for storing 2.

In the fractal model generation unit 261, the tree to be displayed is modeled using fractal geometry, thereby generating a fractal model composed of shape data with the highest definition. By using the data of this fractal model, a tree 261a having a three-dimensional shape with the highest definition as shown in FIG. 3 can be displayed.

Next, the hierarchical simplified model generator 262 generates three simplified models 281 based on the generated fractal model. Here, as illustrated in FIG. 4, three simplified models 281a to 281c having different degrees of definition are generated according to the distance from the viewpoint of the observer 18 to the tree that is the object 14. In general, there is a visual characteristic that a tree located closer to the viewpoint can be observed in more detail, while a tree located farther from the viewpoint can be observed more roughly. Therefore, a tree located closer to the viewpoint needs shape data with higher definition, but a tree located far from the viewpoint does not require shape data having higher definition.

Therefore, here, as shown in FIG. 4A, a simplified model 281 having a high definition as data for displaying a tree located at a position close to the viewpoint.
a is generated based on the fractal model. Further, as shown in FIG. 4B, a simplified model 28 having a medium definition is used as data for displaying a tree at a medium position which is not so close or far from the viewpoint.
1b is generated based on the fractal model. Further, as shown in FIG. 4C, a simplified model 281c having a low definition is generated based on the fractal model as data for displaying a tree located far from the viewpoint. The three simplified models 281 (281a to 281c) thus generated are stored in the database 2
8 is stored.

FIG. 5 shows the data structure of the simplified model 281 stored in the database. As shown in FIG.
The data of the simplified model 281 includes a branch data structure 40 and a leaf data structure 50. The branch data structure 40 is
It has a pointer 41 to a child branch, a range 42 of descendant branches, simplified shape data 43, a texture identifier 44 for simplified shape data, and branch shape data 45. The leaf data structure 50 has leaf shape data 51. The shape roughness due to perspective from the viewpoint and the range 42 of descendant branches are sequentially compared from the original trunk 46 to the child branch 47. Range 4 of descendant branches
When 2 becomes smaller than the above-mentioned shape roughness, the simplified shape data 43 at that time is used instead of drawing the subsequent shape data.

On the other hand, the hierarchical partial texture generator 263.
Then, the partial texture 301 is generated for the element 32 forming each of the simplified models 281a to 281c. As shown in FIG. 3, one element 32 forming the simplified model 281a is selected, and the corresponding portion 34 of the tree 261a having the highest definition using the fractal model is selected. A corresponding portion 34 of the fractal model is projected onto the selected element 32,
Thereby, one partial texture 301 is generated.
Similarly, a plurality of partial textures 301 corresponding to all the elements 32 of the simplified model 281a are generated.

Each element 32 of the simplified model 281 is composed of polygons assembled in a cross shape. Therefore,
The texture 301 is obtained by projecting the corresponding model of the fractal model 261a parallel to each polygon. Further, as the data for texture, four types corresponding to the front, back, left and right of the simplified model 281 are used.

The vertical and horizontal size Stex of the texture is determined according to the following equation (1). Text = Tree × A ^B (1) Here, “Tree” is the vertical and horizontal size of the tree assumed when the computer graphics are displayed. Further, A is represented by the following equation (2). A = 1 / √C (2) Also, B is the generation of the branch for simplification, and C is IFS.
Is the number of reduced maps of. By doing so, the amount of texture data required for the C ^B simplified models can be expressed by the following expression (3). Stex ² × C ^B × 4 = Tree ² × 4 (3) Here, the relationship between the vertical and horizontal size Tree of the tree and the simplification level B, which are assumed when the computer graphics are displayed, is expressed by the following formula (4). To do. 0 ≦ B ≦ E−1,
The larger B is, the finer the image is. Tree = Smax × D ^EB (4) Here, Smax is the maximum vertical and horizontal size of the tree assumed at the time of displaying computer graphics, D is the average reduction ratio of each reduction map, and E is the maximum. It is the generation of branches that gives definition. According to this expression (4), the size of the simplified shape on the display surface is constant regardless of the size. The amount of texture data required at this time is expressed by the following equation (5).

[0023]

[Equation 1]

If, for example, D = 0.5 and E = 8 are given as typical values in this equation (5), the required data amount for texture becomes Smax ² × 5.074.

In this way, a plurality of partial textures 301 are individually generated for each simplified model 281, but since a large number of partial textures are generated, the simplified models 281a ... 281
It is necessary to manage which element of the simplified model corresponds to c. Creating a management table to manage such partial textures
This is the texture management table generation unit 264 shown in FIG.

As shown in FIG. 6, the generated partial textures are grouped for each layer and referred to by the texture identifier 44 for simplified shape data uniquely defined in each group. The simplified shape data texture data management table 60 stores the simplified shape data texture identifiers 44 in the grouped image data 62 with an offset 63 and vertical and horizontal sizes 64 and 65.
Is associated with.

The partial texture 31 and the management table 302 thus generated are stored in the database 30. As a result, all the partial textures 301 necessary for the simplified model 281 stored in the database 28 are prepared, and all the partial textures 301 are prepared.
The data about which element of which simplified model corresponds to is prepared.

FIG. 7 shows the details of the image generation unit 24 for generating a stereoscopic image in real time according to the viewpoint of the observer 18 based on the simplified model 281, the partial texture 301 and the management table 302 which are stored in the database. It is a block diagram showing. Referring to FIG. 7, this image generation unit 24
Is selected by the simplified model selection unit 241 that selects a corresponding one of the simplified models 281 according to the viewpoint data from the control unit 22 and the simplified model selection unit 241 according to the viewpoint data from the control unit 22. A partial texture selecting unit 242 that selects a plurality of partial textures corresponding to the simplified model selected, and a plurality of partial textures selected by the partial texture selecting unit 242 to the simplified model selected by the simplified model selecting unit 241. A rendering processing unit 243 that performs mapping according to the management table 302.

Therefore, the simplified model selection unit 241 selects one of the three simplified models 281a to 281c stored in the database 28. If the distance from the viewpoint of the observer 18 to the tree that is the object 14 is short, the simplified model 281a with high definition shown in FIG. 4A is selected. If the distance from the observer 18 to the tree, which is the object 14, is medium, FIG.
The simplified model 281b having the medium definition shown in (b) is selected. If Observer 18 to Object 1
If the distance to the tree of 4 is long, the simplified model 281c with low definition shown in FIG. 4C is selected.

On the other hand, in the partial texture selection unit 242,
From the partial textures 301 stored in the database 30, a plurality of partial textures 301 corresponding to the simplified model selected by the simplified model selection unit 241 are selected. If the simplified model selection unit 241 selects the high-definition simplified model 281a shown in FIG. 4A, all the partial textures 301 corresponding to the selected simplified model 281a are selected. It

In the rendering processing unit 243, the corresponding one of the selected partial textures is mapped to each element 32 of the simplified model 281 selected according to the management table 302 stored in the database 30. As a result, the data necessary for the three-dimensional display of the tree is supplied to the screen 12, and the three-dimensional image of the tree is displayed on the screen 12.

As described above, according to this embodiment, since the tree shape data corresponding to the distance from the viewpoint is used for drawing, it is necessary to draw a large amount of tree data for landscape simulation. Even in this case, the display image can be changed in real time with respect to the movement of the viewpoint. Therefore, it is possible to display a twin-lens stereoscopic image that is not distorted with respect to the viewpoint movement. Moreover, since the texture is generated by projecting the corresponding part of the fractal model to each element of the simplified model, not the entire simplified model, the deformation of the tree according to the physical law such as fluctuation due to wind Even if the intricate portion is exposed by, the texture of the newly exposed portion due to the deformation can naturally display it.

Further, in the above-mentioned conventional method, about 18 tree images are required to display a twin-lens stereoscopic image without distortion with respect to the viewpoint movement over the entire circumference of the tree of 360 °. On the other hand, in this embodiment, since the entire circumference of the tree can be displayed with the data amount equivalent to the above-mentioned five tree images even at the simplification level having a relatively high definition, the required data amount can be significantly reduced. .

Therefore, in the motion visual display using computer graphics, by generating the shape data with the roughness according to the viewpoint position, even if a large number of trees such as a landscape simulation have to be displayed, the real time display is possible. You can draw with. Even if the shape of the tree image is simplified by texture-mapping the tree image to the simplified shape, the original tree color will remain fine and the three-dimensional tree image will remain. Can be displayed.

[Brief description of drawings]

FIG. 1 is a block diagram showing an overall configuration of a motion vision display system including a stereoscopic image generation device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a part of a stereoscopic image generation device according to an embodiment of the present invention, which is a part for preparing data necessary for the image generation unit in FIG.

FIG. 3 is an explanatory diagram showing a method of generating a partial texture for each element of each simplified model in the hierarchical partial texture generation unit in FIG.

FIG. 4 is a diagram showing a simplified model generated according to a distance from a viewpoint in a hierarchical simplified model generation unit in FIG.

FIG. 5 is an explanatory diagram showing a hierarchical model of a simplified model generated by a hierarchical simplified model generation unit in FIG.

FIG. 6 is an explanatory diagram showing a method of generating a partial texture and a management table in a hierarchical partial texture generation unit and a texture management table generation unit in FIG.

7 is a block diagram showing a detailed configuration of an image generation unit in FIG. 1 and its periphery.

[Explanation of symbols]

 24 image generation unit 28,30 database 241 simplified model selection unit 242 partial texture selection unit 243 rendering processing unit 261 fractal model generation unit 262 hierarchical simplified model generation unit 263 hierarchical partial texture generation unit 264 texture management table generation unit 281 , 281a to 281c Simplified model 301 Partial texture 32 elements

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Katsuyuki Omura, Inventor Katsuyuki Omura, Seika-cho, Soraku-gun, Kyoto Prefecture, 5 Mihiradani, Osamu Osamu, Kyoto, Japan (72) Inventor Shinichi Shiwa, Kyoto Prefecture Soraku Gunma Seika-cho, Osamu Osamu, Osamu Osamu, 5 Hiratani, ATR Co., Ltd.

Claims (1)

[Claims] 1. A stereoscopic image generation device for generating a stereoscopic image according to a viewpoint of an observer, wherein a fractal model is generated by modeling an object to be generated as the stereoscopic image using fractal geometry. Fractal model generation means for generating, a simplified model generation means for generating a plurality of simplified models having a definition according to the distance from the observer's viewpoint to the object, based on the fractal model; Partial texture generation means for generating a plurality of partial textures for each simplified model by projecting the corresponding part of the fractal model onto each of the elements constituting each simplified model, and the observation given from the outside. Model selection means for selecting a corresponding one of the simplified models according to the viewpoint data regarding the viewpoint of the user A partial texture selecting means for selecting a plurality of partial textures corresponding to the selected simplified model according to the viewpoint data, and mapping the selected plurality of partial textures to the selected simplified model. A stereoscopic image generation device comprising a mapping means.