JP3299010B2 - Image generation method and image generation device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[Table of Contents] The present invention will be described in the following order. Industrial application Conventional technology (FIGS. 12 to 16) Problems to be solved by the invention (FIGS. 12 to 16) Means for solving the problems (FIGS. 1 and 5) Action (FIGS. 1 to 5) Example (FIGS. 1 to 11) (1) Configuration of Image Generation Device (FIG. 1) (2) Configuration of Main Storage Device (FIGS. 2 to 4) (3) Internal Configuration of Central Processing Unit (FIG. 5) (4) Radiity calculation unit based on division based on BF product (FIGS. 6 and 7) (5) Radiity calculation unit based on division based on unevenness of the incident surface (FIGS. 8 and 9) (6) Based on moment (FIGS. 10 and 11) (7) Effects of Embodiment Effects of the Invention

[0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image generating method and an image generating apparatus, and more particularly, to an image generating method and an image generating apparatus using an adaptive division method.

[0003]

2. Description of the Related Art Conventionally, the radiocity method is one of methods for obtaining surface energy in computer graphics (CG). In order to create an image by the radiocity method, a scene is created. A scene is a collection of objects to be displayed, and an object is represented by a group of surfaces called patches.

[0004] The radiocity algorithm can be considered as a finite element method for the energy of light. First, the surface of the object constituting the scene is decomposed into sufficiently small portions (patches). Next, the transfer of light energy between these patches is represented by a coefficient called a form factor.

[0005] The form factor is defined as the proportion of the part of the energy emitted by a patch that reaches the patch of interest. The form factor from patch i to patch j is denoted Fi-j. The energy balance per unit area of patch i is calculated using the form factor

(Equation 1) Can be written as FIG. 12 shows the transfer of energy between patches. Here, Bi is the energy per unit area emitted from the patch i, that is, the radioactivity to be obtained. Ei is the energy per unit area at which the patch i emits light, and ρi is the diffuse reflection coefficient of the patch, which is given in advance as the intensity of the light source and the color of the surface. Ai and Aj are patch areas, which can also be calculated in advance from coordinate values. Equation (1) further defines Fi-j Ai = F
It is simplified using the relationship ji Aj. When a similar expression is made for n patches, an n-order system of equations relating to B is obtained, and a radicality can be obtained by solving this.

[0006] Here, as a conventional radiometric method by adaptive division, if there are n patches, calculation of the number of combinations, ie, O (n ² ) form factors, is originally necessary. O (n ² ) in conventional radio
Was the biggest challenge (Cohen86), but in recent years the computational complexity has been reduced to O (n) by the application of techniques for solving N-body problems in scientific calculations (Hanrahan91), (Smits92). These improved algorithms make the decomposition into patches hierarchical and adaptive (adaptiv
In step e), the number of form factors to be calculated is reduced by transmitting and receiving energy between groups as far as the error permits, with respect to a remote patch having little effect.

FIG. 13 shows the configuration of processing on the central processing unit in the conventional radiometric method by adaptive division. This central processing unit includes a radioactivity calculation unit 2 based on division based on a BF product which is a product of a radioactivity and a form factor, and a radiocity calculation unit 3 based on division based on non-uniformity of an incident surface.

[0008] Both of the radiometric operation units 2 and 3 start with the transmission and reception of light between the initial patches (interactions) constituting the scene in the initial state, and then adaptively divide the patches when the error is large. ,
Find the radiocity in the equilibrium state. If it is decided to perform the division, the patch is divided and the original interaction is replaced with the interaction with each of the divided patches. Patch splitting is continued recursively until the radiocity error falls within a predetermined tolerance.

When the errors of all the interactions fall within a predetermined range, the interaction is traced at each node while descending the patch tree structure from the root.
Multiply the radiocity from the source. This time, the radio node of the intermediate node is determined while moving up the tree. The above-described steps of the error evaluation and the collection of the radioactivity are repeated until the values converge, and a solution of the radiocity is obtained.

The two types of radio frequency computing units 2 and 3 differ from each other in the method of determining whether a patch can be divided. In the radioactivity calculation unit 2 based on the division based on the BF product, an interaction having a large transfer of the BF product is regarded as having a large error, and is divided until the BF product becomes smaller than a certain size. In addition, the radioactivity calculation unit 3 based on the division based on the non-uniformity of the incident surface calculates the difference between the average of the radiocity obtained at a plurality of points on the incident surface and the radioactivity obtained at a single point on the incident surface. And whether or not the division is possible is determined. From the above, the non-uniformity of the radioactivity on the radiation surface side is not taken into account in the conventional determination of the possibility of division.

FIG. 14 shows a processing configuration on the central processing unit 1 in the conventional radiometric method based on equal division. The radiometric operation unit 4 based on equal division only performs division until a certain degree of fineness is obtained in places where the nonuniformity is large and small in the radiocity operation. FIG. 15 and FIG. 16 show the processing of the radioactivity calculation unit 4 by the equal division. This calculation unit performs division based on the area. From the above, the radiometric operation unit based on equal division can perform calculations by a simple procedure, and by making the division fine, it is possible to accurately calculate the influence due to the unevenness of the radiation surface.

[0012]

By the way, radiocity is an important display technique in computer graphics, and is particularly suitable for a realistic representation of a space illuminated by indirect light, such as a room. However, in the conventional method, an error of the radiocity caused by the nonuniformity of the radioactivity of the radiation surface is not sufficiently evaluated, and as a result, the incident surface may not be appropriately divided.
For this reason, for example, when the carpet or the wallpaper is not plain, or when the brightness of the radiating surface is large due to shadows or spotlights, the error is extremely large in the conventional method.

When the equal division is used, if the division is made fine in order to suppress the error, the calculation amount increases in proportion to the square of the number of patches obtained as the division result.
For this reason, application to practical scenes is difficult. From the above, there has been a demand for an adaptive division determination method capable of performing accurate radiocity calculation with a small number of divisions even in a scene where unevenness of the radiation surface is reflected on other surfaces.

SUMMARY OF THE INVENTION The present invention has been made in view of the above points, and has an image generating method and an image generating method capable of performing accurate radioactivity calculation with a small number of divisions even in a scene where unevenness of a radiation surface is reflected on other surfaces. An image generation device is proposed.

[0015]

According to the present invention, an input image is divided into a plurality of patches to represent the transfer of light energy between an arbitrary radiation surface patch and an arbitrary incident surface patch. When the form factor is calculated, the coefficient of the first order or higher of the radial moment of the radiation patch obtained as the product of the moment of the texture of the radiation surface and the incident radiation and the form factor obtained discretely are interpolated. Using the product sum of the form factor and the first or higher order interpolation coefficient, the radiometric prediction error, which is the predicted value of the error between the average radiocity within the radiation surface patch and the radioactivity of each small area portion within the radiation surface patch, is calculated. The predicted and calculated radioactivity prediction error is compared with a predetermined error threshold, and the It is greater than, and so further to divide the radiation surface Patsuchi.

[0016]

[0017]

The radiation surface is divided based on the error caused by the non-uniformity of the radiation on the radiation surface on the radiation on the incident surface, and a portion for calculating the radiation is added. In the radioactivity calculation unit, the function of the radioactivity distribution of the radiation surface patch and the form factor distribution between the radiation surface patch and the incident surface patch were used in the radioactivity calculation unit to judge the propriety of dividing the radiation surface. Accordingly, accurate radioactivity calculation can be performed even in a scene in which unevenness of the radiation surface is reflected on other surfaces.

[0018]

BRIEF DESCRIPTION OF THE DRAWINGS FIG.

(1) Configuration of Image Generation Apparatus In this embodiment, the radioactivity is obtained by using the moment of the radiocity, and an image in which the reflection is appropriately expressed is created. FIG. 1 shows an image generating apparatus 10 for generating an image from shape, coordinate, color, and light source data using the radiocity method in this method. The image generation device 10 includes an input device 11, a central processing unit 12, a main storage device 13, and an output device 14. The input device 11 inputs shape data of a figure, coordinate data, color, light intensity of a light source, and a texture image necessary for creating an image by the radiocity method. The input device 11 includes, for example, a combination of the following device groups.

That is, the CAD device 11A is for inputting a figure, and the coordinate input device 11B is for picking up coordinates. In addition, disk device 11C
Is an external storage device for inputting the stored coordinates and texture image data to the central processing unit 12, and the scanner device 11D is for reading texture image data and the like from the paper.

The main storage device 13 is provided with a central processing unit 12.
Is stored. The central processing unit 12
Temporarily save the results of any of the above processing units. further,
The stored result is supplied to an arbitrary processing unit on the central processing unit 12 as needed. The central processing unit 12 follows the processing procedure extracted from the main storage device 13 and
Performs processing on the shape and coordinate data input from 1
Creates rendering result image data, and outputs
Output to

The output device 14 displays or stores the processing result of the rendering unit in the central processing unit 12, and is composed of, for example, a combination of the following device groups. That is, C
The RT device 14A is a graphic display device for displaying an image of a rendering result on a screen. The printer device 14B is for storing the image of the rendering result on paper. Furthermore, the disk device 14
C is for storing an image as a rendering result.

(2) Structure of Main Storage Device As shown in FIG. 2, the interior of the main storage device 13 includes a patch tree array 13A, texture image data 13B, texture average data 13C, texture moment data 13D, and rendering image data. 13E. The patch tree sequence 13A is a sequence of patches representing the shape of the scene, and is a texture image data 13B.
Is texture image data before initialization, and the texture average data 13C is texture data corresponding to the shape of the scene. In addition, texture moment data 1
3D is texture moment data corresponding to the shape of the scene, and rendering image data 13E is image data for storing a rendering result of the scene.

First, the patch tree row 13A is:
As shown in FIG. 3, for example, patch trees P00 and P00 corresponding to initial patches P0 and P1 forming the shape of the entire scene.
01, P02, P03, P000, P001, P002, and P003 are all arranged. Patches that transfer energy on different patch trees are related by an interaction list.

A single patch tree is a data structure corresponding to one of the initial patches that make up the scene. Before the calculation of the radiocity, there is only a single initial patch in a single patch tree, but as the calculation of the radiocity proceeds, it is divided into hierarchical patch groups. Assuming that the patch before division is a parent patch and the patch after division is a child patch, the parent patch and the child patch are related by link data. After the radiocity calculation, the entire patch group having a hierarchical parent-child relationship linked by the link data becomes a single patch tree.

Here, the patch is a link data L as shown in FIG. 4 as patch data P holding its shape, coordinates, color, and the radioactivity obtained as a calculation result.
K, coordinate data ZH, and radiocity data RD. That is, in this embodiment, a case where the patch P is divided into four rectangles will be described as the link data LK. The division of an arbitrary polygon into an arbitrary number can be dealt with by increasing the coordinate data and the address of the child patch.

The address of the child patch divides one patch into four for each division. Therefore, one patch has four child patch addresses. At the time the patch is created, the addresses of the child patches are not stored. When the incident plane is divided, four small patches are created, and the addresses of the created small patches are stored. Also, since there is only one parent for a child patch, it has one parent patch address. The interaction list is a list that holds a relationship between a radiation surface and an incident surface. When the patch reaches the entrance surface, all other patches incident on that patch are retained. When the radiation surface is divided, each of the divided radiation surfaces is added to the interaction list, and the radiation surface before the division is deleted.

The coordinate data ZH is data related to the shape of the figure in the patch data, and is composed of modeling coordinates, texture coordinates, and patch coordinates. The modeling coordinates store patch coordinates in a modeling coordinate system that determines the arrangement of patches in the entire figure. The texture coordinates store the coordinates of patches in a texture coordinate system that determines the correspondence between patches and textures.
The patch coordinates store the coordinates of the patch in the patch coordinate system for determining the location of the child patch among the parent patches.

The radiocity data RD is radioactivity moment data obtained as a result of the radiocity calculation. Since the radiocity moments are calculated in a hierarchical manner, the descendants, ancestors, oneself, Classified as total. Among them, the radioactivity moment of the patch of the descendant and the radiation surface interacting with the patch group of the descendant are stored. The radioactivity moment of the ancestral patch and the radiation surface interacting with the patch of the ancestor are stored in the radioactivity moment of the ancestor patch.

In the BF total, the total BF product of the own patch and the interacting radiation surface is integrated. The radioactivity moment of the own patch and the radiation surface interacting with the own patch are stored in the radioactivity moment of the own patch. The total of the radioactivity moments stores the ancestors from the initial patch to the self, and the total of the radioactivity moments of the self and the descendant patches. In the radio vertex of the patch vertex, the radio vertex of the patch vertex obtained by performing the interpolation is stored in the sum of the radio moments.

Second, the texture image data 13B
Consists of a two-dimensional array that stores pixel values of image data. In this embodiment, moments for u, v, and uv are necessary as texture average data 13C in order to obtain coefficients for error prediction.

In this embodiment, since the maximum number of divisions is set to six, 4096 are calculated from the texture average data initial patch.
It is a two-dimensional array corresponding to the number of divisions up to the division, and the initialized texture in the initial patch,
Initialized texture in 4-split patch, initialized texture in 16-split patch, initialized texture in 64-split patch, initialized texture in 256-split patch, initialized texture in 1024-split patch, 4096 split patch It is made up of textured textures.

Fourth, as the texture moment data 13D, in this embodiment, the first moment about the texture u, the first moment about the texture v, and the bi-first moment about the texture uv are used. I have. The moments for u, v, and uv are needed to determine coefficients for error prediction.

In this embodiment, since the maximum number of divisions is determined to be six, the textures in the initial patch correspond to the respective number of divisions from the above-mentioned initial moment patches of the three types of moments to those of the 4096 divisions. Moment, texture moment in 4-split patch, texture moment in 16-split patch, texture moment in 64-split patch, texture moment in 256-split patch, texture moment in 1024-split patch, texture moment in 4096-split patch It is composed of a corresponding two-dimensional array.

Fifth, rendering image data 13E
Since the rendering by the radiocity method of this embodiment is performed by the Z-buffer method, rendering image data obtained as a result of rendering is two-dimensional of R, G, B, and Z values representing pixels of an output image. Consists of an array.

(3) Internal Configuration of Central Processing Unit FIG. 5 shows the configuration of processing performed in the central processing unit 12. In the case of this embodiment, the processing inside the central processing unit 12 includes an initialization unit 12A, an incidence plane scan unit 12B, a radioactivity calculation unit 12C based on a radiation plane / incidence plane division based on a BF product, an end condition determination unit 12D,
It is composed of a radioactivity computing unit 12E based on the incidence plane division based on the non-uniformity of the entrance plane, a radioactivity computation unit 12F based on the radiation plane division based on the moment, a radiocity interpolation unit 12G, a rendering unit 12H, and an output unit 12I. And calculates a radiocity based on input data such as coordinate data to generate a rendered image.

First, the initialization unit 12A generates a patch group from the coordinate data obtained from the input unit. Next, an interaction is set between all patches and all other patches. Next, preprocessing of the texture image is performed. The moment of the texture in a state where the texture image is divided into each division stage is obtained and stored in the moment of each divided patch. Similarly, the average of each division stage is obtained and stored in the moment of the 0th-order texture. As a result of the initialization processing, a patch tree arrangement in which initial patches are set is output. It also outputs the pixels of the texture image. Further, the moment of the hierarchically divided texture image and the average of the hierarchically divided texture image are output.

In the incident surface scanning section 12B, the incident surfaces are scanned one by one. Next, the total collected BF of the scanned incident surface is initialized to 0, and each incident surface data is output. Further, the radioactivity computing unit 12C based on the radiation surface / incident surface division based on the BF product obtains the radiocity gathering on a specific incident surface by the radiation surface / incident surface division based on the BF product.

The end condition judging unit 12D waits for the end of the processing on all the incident surfaces. Next, it is checked whether or not the radiocity has sufficiently converged, and whether or not the division is sufficient is detected. As the convergence condition, for example, a condition that there is no newly divided patch in the previously executed radiometric operation can be used.

The radioactivity calculator 12E based on the splitting of the incident surface based on the non-uniformity of the incident surface finds the radioactivity that converges on a specific incident surface by splitting the incident surface based on the non-uniformity of the incident surface. In addition, the radioactivity computing unit 12F based on the radiation surface division based on the moment obtains the radiocity that converges on a specific incident surface by the radiation surface division based on the moment. Further, the radiocity interpolation unit 12G uses the radioactivity of each patch obtained by the processing of each radioactivity operation, and obtains the radiocity of the vertex of each patch by the interpolation operation.

The rendering unit 12H determines the shape of the patch to be drawn by applying coordinate transformation to the coordinates of each patch, and draws the patch using the colors of the vertices obtained by the illuminance interpolation unit. When drawing, the hidden surface is erased by, for example, the Z buffer method. The color on the patch is determined by the product of the average of the radiocity on the patch and the pixel value of the texture image data corresponding to the texture coordinates on the patch. As a result of drawing, rendering image data is obtained and output from the output unit 12I.

(4) Radiity calculation unit based on division based on BF product Here, processing related to radio calculation based on division based on the BF product in the radiation calculation unit 12C based on splitting of a radiation surface and an incident surface based on division based on a BF product 6 and 7 are shown in FIGS. A patch data group and a specific incident surface of interest are given as inputs to the arithmetic unit. The output is obtained by dividing the input patch data group on the basis of the BF product, and obtaining the radioactivity for the divided patch of an arbitrary level.

In this process, the determination unit for a new radiation surface checks whether there is any radiation surface that has not yet been taken out of the radiation surface having the input incidence surface and the interaction. In the take-out part of the new radiation surface, the radiation surface with interaction is sequentially picked up for the input incident surface.

As the calculation of the BF product, the BF product representing the transmission of the radioactivity from the radiation surface to the entrance surface is obtained by the product of the form factor by the microsurface-microsurface approximation and the radioactivity of the radiation surface. Form factor by micro-surface approximation is

(Equation 2) Required by Here, θi is the angle formed by the line segment connecting the center of the radiation surface and the center of the incident surface and the normal line of patch i, and θj is the line segment connecting the center of the radiation surface and the center of the incident surface and the normal line of patch j. The angle formed, d, is the distance between the centers of patch i and patch j.

In order to determine that BF is larger than the threshold value, the threshold value in this embodiment is set to 10% of the radioactivity of the patch having the largest self-emission energy.
And It is determined whether the area of the surface to be divided is larger than the threshold value, and only the surface having the area equal to or larger than the threshold value is divided. By this determination, it is possible to prevent the processing from becoming an infinite loop due to the infinite division. In this embodiment, it is used as the threshold value of 2 ^-12 of the area of initial Patsuchi. The comparing section of the area of the radiation surface and the area of the incidence surface determines whether the area of the radiation surface is larger than the area of the incidence surface. With this determination, the larger surface is divided.

In the process of dividing the incident surface, it is checked whether a child patch of the incident surface obtained as a result of the division has already been created. If not, a data area necessary to store four patches is secured. The address of the parent patch is stored in the address of the parent patch on the four child patches. Next, the address of the data area of the child patch group secured at the address of the child patch of the parent patch (incident surface) is stored. The address of the radiation surface is deleted from the interaction list of the parent patch on the parent patch, and the address of the radiation surface is set in each of the interaction lists of the child patch group.

In the radiation surface division, it is checked whether or not a child patch of the radiation surface obtained as a result of the division has already been created. If not, a data area necessary to store four patches is secured. The address of the parent patch (radiation surface) is stored in the address of the parent patch on the four child patches. Next, the address of the data area of the child patch group secured at the address of the child patch of the parent patch is stored. The address of the parent patch (radiation surface) is deleted from the interaction list on the incident surface, and the addresses of the child patch groups are stored.

In the update unit for the total BF of the incident surface, if it is determined that the division is unnecessary for both the incident surface and the radiation surface, the obtained BF product is added to the total BF of the incident surface. In determining whether or not the incident surface is divided, it is determined whether or not the incident surface is already divided. Further, in the inheritance of the radiocity from the parent patch to the child patch, the sum of the radiocity of the ancestor patch and the radiocity of the own patch on the parent patch is stored in the ancestor radiocity on the child patch.

Further, the radiometric operation section having the child patches 1 to 4 as an incident surface recursively executes the radiometric operation by division based on the BF product using each of the child patches 1 to 4 as an incident surface. Further, as the inheritance of the radio patchy from the child patch to the parent patch, the child patch radiocity operation unit obtains the radio patch of the parent patch from the recursively obtained radio patch group. The incidence surface radioactivity calculator multiplies the calculated total BF by the average texture and adds self-emission to determine the incidence surface radioactivity.

(5) Radiity Calculation Unit Based on Division Based on Unevenness of Incident Surface FIG. 8 and FIG. 9 show processing of the radio-calculation unit 12E based on division based on non-uniformity of the incident surface. A patch data group and a specific incident surface of interest are given as inputs to the arithmetic unit 12E. The output is obtained by dividing the input patch data group based on the non-uniformity of the incident surface, and obtaining the radioactivity for the divided patch of an arbitrary level.

In this process, in the calculation of the BF product and the error prediction due to the non-uniformity of the incident surface, the sum of the BF products when each of the radiation surface and the incident surface divided into four is approximated by a minute surface is obtained. Next, the BF product when both the radiation surface and the incident surface are approximated by a single minute surface is determined, and the difference between the two is used as a prediction error. The prediction error is compared with the threshold value to determine the magnitude. In this embodiment, the prediction error 1 [%] is used as the threshold value.
Further, the radiometric operation unit using the child patches 1 to 4 as an incident surface recursively executes a radiometric operation by division based on the non-uniformity of the incident surface, using the child patches 1 to 4 as an incident surface.

(6) Radiity Calculation Unit Based on Moment-Based Division Processing FIGS. 10 and 11 show the processing of the radio-activity calculation unit 12F based on moment-based division. In the radioactivity calculation unit 12F, a patch data group and a specific incident surface of interest are given as inputs. The output is obtained by dividing the input patch data group based on the moment, and obtaining the radioactivity for the divided patch at an arbitrary level.

For calculating the BF product and estimating the error based on the moment, the following equation is used for the equation (1) of the radiocity.

(Equation 3) Strictly speaking, the radiocity that reaches a point on patch i from patch j is

(Equation 4) Given by Here, d is the distance between dA _i and dA _j . Conventionally, it was assumed that the radiocity was a constant value B _Aj in the plane, and this integral was decomposed into the product of the radiocity and the form factor. That is,

(Equation 5) Further, the _{form factor} F _dAi-Aj is approximately obtained, and using this, the following equation is obtained.

(Equation 6) Equation (1) was derived.

The transformation from equation (4) to equation (6) is correct when the radiocity is constant inside patch j.
In fact, in most cases the radiocity is uneven. Of course, a correct value can be obtained by performing numerical integration faithfully according to the equation (4), but it takes too much time. Therefore, (4)
Transform the formula into a more easily calculated form. First, the form factor term in equation (4) is expressed by the following equation using coordinate systems u and v on patch j.

(Equation 7) Expand like By rewriting equation (4) using this, the following equation can be obtained.

(Equation 8) If the coefficient is taken out and regrouped,

(Equation 9) Is obtained. In the equation (9), the first term corresponds to the average radiocity in the patch, and the moments of the radioactivity after the second term indicate the prediction error.

From the above, patch coordinate systems u and v are determined on the radiation surface. Assuming that the radiation surface is divided into four parts, approximate coefficients F, Fu, Fv, and Fuv are obtained when the form factor group obtained at the center of the four divisions is approximated by a polynomial of u and v. Also,

(Equation 10)

[Equation 11]

(Equation 12)

(Equation 13) Is calculated from the product of the radiation surface moment and the texture moment stored in the texture image data.

The determination section as to whether or not the prediction error based on the moment is larger than the threshold value determines whether the prediction error based on the moment represented by the second and subsequent terms of equation (9) is larger than the threshold value. Do. In this embodiment, the prediction error 1
[%] Is used as the threshold value. In addition, the update unit of the total BF moment of the incident surface adds the calculated BF moment to the total BF moment of the incident surface when it is determined that the division is unnecessary for both the incident surface and the radiation surface.

In the inheritance part of the radioactivity moment from the parent patch to the child patch, the sum of the radioactivity moment of the ancestor patch on the parent patch and the radioactivity moment of the own patch is calculated on the child patch. Store it in the ancestor's radiocity moment. Further, the radio moment calculation unit using the child patches 1 to 4 as an incident surface recursively executes a radio calculation by division based on the moment using the child patches 1 to 4 as an incident surface. In the inheritance unit of the radio patch moment from the child patch to the parent patch, the radio patch calculation unit of the child patch obtains the radio patch moment of the parent patch from the recursive radio moment of the child patch group.

(7) Effects of the Embodiment According to the above configuration, in the radiocity method by adaptive division, the incidence plane division is appropriately performed by predicting the radiocity error caused by the nonuniformity of the radiation plane of the radiation plane. By doing so, accurate reflections can be calculated, for example, when the carpet or wallpaper is not plain, or when the brightness of the emitting surface is large due to shadows or spotlights. Was.

Also, an accurate result can be obtained with a much smaller number of interactions as compared with the case of using even division, which is another method for calculating an accurate reflection.
Therefore, application to practical scenes is possible.

[0060]

As described above, according to the present invention, the radiation surface is divided on the basis of the error caused by the non-uniformity of the radioactivity on the radiation surface on the radioactivity on the incidence surface, and a portion for calculating the radioactivity is added. In this way, the non-uniformity of the radiating surface is reflected on other surfaces, and the radioactivity calculation unit determines the appropriateness of dividing the radiating surface by determining the distribution of the radiating surface patch and the radiating surface patch and the incident surface patch. By using the function of the form factor distribution between the two, it is possible to realize an image generation method and an image generation apparatus capable of performing accurate radioactivity calculation even in a scene in which unevenness of a radiation surface is reflected on another surface.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an overall configuration of an image generating apparatus to which an image generating method according to the present invention is applied.

FIG. 2 is a schematic diagram illustrating an internal configuration of a main storage device of the image generating apparatus.

FIG. 3 is a schematic diagram illustrating a configuration of a patch tree arrangement in a main storage device.

FIG. 4 is a schematic diagram illustrating a configuration of patch data in a main storage device.

FIG. 5 is a block diagram illustrating a configuration of a process performed by a central processing unit of the image generating apparatus.

FIG. 6 is a flowchart showing processing of a radio-activity calculating unit based on division based on a BF product.

FIG. 7 is a flowchart showing processing of a radio-activity calculating unit by division based on a BF product.

FIG. 8 is a flowchart showing the processing of the radioactivity calculation unit based on division based on non-uniformity of the radiation surface.

FIG. 9 is a flowchart showing the processing of the radioactivity calculation unit based on division based on non-uniformity of the radiation surface.

FIG. 10 is a flowchart showing the processing of a radiocity operation unit by division based on a moment.

FIG. 11 is a flowchart showing a process of a radiography operation unit by division based on a moment.

FIG. 12 is a schematic diagram for explaining the transfer of energy between patches.

FIG. 13 is a block diagram showing a configuration of processing of a conventional radiometric method by adaptive division.

FIG. 14 is a block diagram showing a configuration of a conventional processing of a radiometric method by equal division.

FIG. 15 is a flowchart showing the processing of the radiography operation unit by equal division.

FIG. 16 is a flowchart showing the processing of a radiography operation unit by equal division.

[Explanation of symbols]

10 image generating device, 11 input device, 11A
CAD device, 11B coordinate input device, 11C disk device, 11D scanner device, 12 central processing device, 12A initialization unit, 12B incident surface scan unit, 12C BF product Radiation surface calculation unit based on radiation plane / incidence plane division, 12D... Termination condition determination unit, 12E Radiation surface calculation unit based on incidence plane division based on non-uniformity of entrance plane, 12F... Radiation plane based on moment Radiometric operation unit by division, 12
G ... radiocity interpolation unit, 12H ... rendering unit, 12I ... output unit, 13 ... main storage device, 13A ...
... Patch tree, 13B ... Texture image data, 1
3C: Texture average data, 13D: Texture moment data, 13E: Rendered image data, 14: Output device, 14A: CRT device, 14B
... Printer device, 14C disk device.

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-4-107689 (JP, A) JP-A-2-287776 (JP, A) Hajime Hosaka, Takushi Totsuka, radiocity considering texture on surface, Sony Research Forum 1993 Transactions, Japan, Sony Corporation General Technology Group, February 25, 1994, 71-76, Application for this application (issued after filing the application) Gen Kumamoto et al., Form factor calculation in the radiosity method Proceedings of the 1993 IEICE Spring Conference, Proc. Of the IEICE, IEICE, March 15, 1993, no. 7,370 (58) Field surveyed (Int.Cl. ⁷ , DB name) G06T 15/50 240 JICST file (JOIS)

Claims (2)

(57) [Claims] (1)ImageRadiocity method with adaptive partitioning
In the image generation method byPatch splitting step to split the input image into multiple patches
When, The light energy between any emitting surface patch and any incident surface patch
A fore that calculates the form factor that represents the movement of the lugie
Muh actor calculation step; Radiation surface texture moment and incident radioactivity.
Of the radiation surface patch as the product of
Coefficient of first order or higher and discrete form
When the factor is polynomial interpolated, the primary of the form factor
Using the sum of products with the above interpolation coefficients,
Average radioactivity and small size within the radiation surface patch
The estimated value of the error with the radiocity of each area
A radiocity that predicts a certain radiocity prediction error
Error prediction steps, The calculated radiocity prediction error and a predetermined error threshold
, And the above-mentioned radio-city prediction error is
If the value is larger than the value, the radiation surface patch is further divided
Patch subdivision steps Equipped with An image generation method characterized by the following. (2)A radiocity method for adaptive segmentation of images.
In the image generation device according to Patch dividing means for dividing the input image into a plurality of patches; The light energy between any emitting surface patch and any incident surface patch
A fore that calculates the form factor that represents the movement of the lugie
Muff actor calculating means; Radiation surface texture moment and incident radioactivity.
Of the radiation surface patch as the product of
Coefficient of first order or higher and discrete form
When the factor is polynomial interpolated, the primary of the form factor
Using the sum of products with the above interpolation coefficients,
Average radioactivity and small size within the radiation surface patch
The estimated value of the error with the radiocity of each area
A radiocity that predicts a certain radiocity prediction error
Error prediction means; The calculated radiocity prediction error and a predetermined error threshold
, And the above-mentioned radio-city prediction error is
When the value is larger than the value, the radiation surface patch is further increased. To divide
Patch subdivision means An image generating apparatus comprising: