JPH0786937B2 - Rendering system for 3D computer graphics - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a graphic processing and image processing technique by a computer, and more particularly to an image production support mechanism suitable for three-dimensional computer graphics.

[Conventional technology]

In the conventional 3D computer graphics system, when rendering information of modeled information such as shape and motion by a renderer to obtain video information, the parameters to be given to the renderer are Everything had to be decided and entered by the user. Since the number of combinations of these parameters is infinite, the user needs a large number of trials to obtain a desired result, resulting in an increase in man-hours for rendering.

So far, no attempt has been made to solve this problem in three-dimensional computer graphics, but there is an example in which the same kind of problem is dealt with as an application example of an expert system.

For example, Proceedings of AMC GC Symposium On-User Interface Software (1988), pages 117 to 128 (Pr
oceedings of the ACM SIGGRAPH Symposium on User In
terface Software (1988) pp.117-128), in order to design user interfaces such as texts, windows and icons, these colors such as texts are automatically displayed with the support of an expert system having artistic knowledge about color arrangement. The methods of making decisions are extended.

Also, in Ergonomics, Vol. 22, No. 1 (1986), pages 1 to 7, when an image of the interior requested by the customer is input as an adjective for designing the interior interior, it is input in advance. It describes how the system selects and displays the most suitable interior from the interior.

[Problems to be Solved by the Invention]

In the above-mentioned conventional technology, no consideration is given to the application of the three-dimensional computer graphics to the rendering process, and it is not possible to reflect the intention of the video effect on the shape or motion defined by the user in advance. There was a problem that I could not.

An object of the present invention is to provide a method for realizing a video effect support mechanism applicable to rendering processing of a three-dimensional computer graphics.

[Means for Solving the Problems]

To achieve the above object, the knowledge database 2 is used to provide an analyzer unit 3 and an interpreter unit 4 for converting a language into parameters that can be input to a renderer 5.

[Action]

The analyzer unit 3 extracts the noun data from the modeling database 1 and uses the noun rules of the knowledge database 2 to convert the noun data into reference parameters. Next, the interpreter unit 4 receives the reference parameter from the analyzer unit 3 and the adjective data from the man-machine unit 7 as input, and uses the adjective rule of the knowledge database 2 to output the correction parameter.

Therefore, the analyzer unit 3 can deal with the shape and motion defined in advance by the user, and the interpreter unit 4 can reflect the intention of the image production.

〔Example〕

An embodiment of the present invention will be described below with reference to FIG. First, an outline of the overall processing will be described, and then the details of the processing of each unit will be described.

As a flow of a normal rendering process, the model data of the shape and the motion pre-input to the modeling database 1 are subjected to the rendering process in the renderer unit 5 and output to the display unit 6 as image information to be displayed in a display. Is displayed. The present invention is intended to support the determination of the parameters to be input to the renderer unit 5 in this kind of processing.

In FIG. 1, first, the analyzer unit 3 extracts noun data from the modeling database 1 and applies the noun rule in the knowledge database 2 to convert the noun data into reference parameters. Next, the interpreter unit 4 receives the reference parameters from the analyzer unit 3 and the adjective data from the man-machine unit 7 as input, applies the adjective rules in the knowledge database 2, and uses the reference parameters and adjective data as correction parameters. Convert. Then, the renderer section 5 stores the model data from the modeling database 1,
With the correction parameters from the interpreter unit 4 as input,
Rendering processing is performed and the image information is output to the display unit 6 to finally obtain an image.

The processing will be described in detail below regarding the operation of each unit in FIG. The parameters to be input to the renderer unit 5 are color parameters, texture parameters, composition parameters,
Although there are illumination parameters and the like, a case of a method for supporting the determination of color parameters will be described below as an example.

Here, the color parameters will be described with reference to FIG. All colors are defined by three parameters of hue, lightness, and saturation, and the color solid defined by these has a double cone shape as shown in FIG. The central axis is lined with achromatic colors from white to black, and pure colors such as red, yellow, green, blue, and purple are lined up in the circumferential direction at the center of the brightness.
The hue represents the hue, the lightness represents the brightness, and the saturation represents the freshness. The renderer unit 5 can determine the color information of the model by designating these three parameters.

As the details of the processing of each unit, first, the processing of the analyzer unit 3 will be mainly described. The modeling database 1 referred to by the analyzer unit 3 is information that the user inputs in advance before the rendering work, and the internal structure thereof is shown in FIG. The inside is composed of information of the shape model 11 and the motion model 14, and each of them is further composed of noun data 12, 15 and model data 13, 16. This noun data 12,15
It is assumed that a noun expressing the semantic content of the corresponding model data 13 and 16 is input, and specifically, nouns such as “mountain” and “sea” correspond to this. The model data 13 and 16 are actually modeled data. For example, when the shape model is expressed as a set of triangle data, the vertex coordinate values of each triangle are stored here. .

Next, the structure of the analyzer unit 3 is shown in FIG. The analyzer unit 3 has a fuzzy reasoning mechanism 31, converts the noun data 12 and 15 from the modeling database 1 into reference parameters according to the noun rule from the knowledge database 2 and outputs them to the interpreter unit 4.

The fuzzy reasoning mechanism 31 will be briefly described with reference to FIG. First, the first rule is (if X ₁ is
Then, increase Y. ).
Then, in this rule, the conditional clause after "if" is the antecedent part, and the conclusion clause after "if" is the consequent part, and the member function is defined for each. The member ship function shows the probability distribution, and the position where the member ship value is 1 is most probable. In the actual calculation, from the membership value corresponding to the input value,
The area of the member function of the consequent part is calculated by min operation. The area to be obtained is a shaded area. For the second rule, (If X ₂ is small, Y is normal.) Similarly, find the area of the shaded area. The conclusion of these two rules is that the two shaded areas are combined by the max operation and the position of the center of gravity is calculated, and this is the output value.

The rules used for fuzzy reasoning in the analyzer unit 3 are obtained from the knowledge database 2. Here, the structure of the knowledge database 2 is shown in FIG. 6. The inside is divided into adjective rules 21 and noun rules 24, The analyzer part 3 uses the noun rule 24. Although the adjective rule 21 will be described later, a specific example of the noun rule 24 is shown in FIG. For example, (if noun = "mountain", the hue is green.) (If noun = "sea", the hue is blue.) Write a rule and define the member function. is there. Here, the membership function is always 1 in the antecedent part, and only the antecedent part is shown. Further, the shape of the member ship function is set as shown in the same figure so that the hue which is considered to be the reference has the peak of the member ship value. The analyzer unit 3 searches the noun rule 24 matching the noun data 12 by the fuzzy reasoning mechanism 31, and outputs the conclusion clause (consequent part) and the membership function to the interpreter unit 4 as reference parameters.

Hereinafter, the processing of the interpreter unit 4 will be mainly described.

The structure of the interpreter unit 4 is as shown in FIG. 8 and has a fuzzy reasoning mechanism 41 similar to that of the analyzer unit. The conclusion clause and the membership function output by the analyzer unit 3 are input as reference parameters, and the man-machine unit 7
The adjective rule 21 from the knowledge database 2 is applied to the adjective data from to determine the correction parameter, and this is output to the renderer unit 5.

Here, the content of the knowledge database in which the adjective rules are stored is as shown in FIG. 6, and the inside of the adjective rules is further divided into the art knowledge rule 22 and the subjective scale rule 23.
It is subdivided into two. Next, the content and construction method of these two rules will be shown.

The art knowledge rule 22 is a rule relating to the expertise used by artists and designers in the production of images, and is specifically a rule that can be extracted from a technical book or an interview with an expert. An example of this specialized knowledge is shown in FIG. This is a rule relating to color arrangement, and when a certain reference color P is defined, it indicates a range of hue values of colors in harmony with this. According to this knowledge, the harmony region and the inharmonicity region are present symmetrically around the reference color and alternate at a constant angle. The inharmonic regions are called the first and second indistinct regions, and the inharmonic regions indicated by the diagonal lines are called the same, similar, and contrastive harmony. The names of these harmony regions indicate that the reference color and the color of that part have the same, similar, and contrasting relationships.

Figure 10 shows the above art knowledge expressed in a rule format for fuzzy reasoning. The rule body is described as (if adjective = "contrast" and adjective = "harmonious", then hue is P '). Further, the membership function of the consequent part is determined such that the membership value becomes 1 in the section from P ″ to P and becomes 0 in the other section as shown in the figure. The above rule and the membership function are art knowledge rules. There are 22 contents.

Regarding the subjective measure rule 23, a construction method is first shown, and then an example of the rule is shown. The main function of the subjective scale rule is to express the user's intention of video calculation with adjectives and to convert these adjectives into parameters that the renderer can process.

To construct a subjective scale rule, it is necessary to present the sample to multiple subjects, evaluate the proportion of adjectives involved in it, and analyze the results. As a method for this analysis, SD (Semantic Different
ial) method will be described as an example. What is shown in FIG. 11 is an explanatory diagram of an example of a rating method using adjectives.
First, the adjectives that are considered to have a large contribution to the parameter to be converted are selected and combined with the opposite adjectives to predetermine the "light-dark" adjective pair. Next, the test subject is presented to the subject, and a mark is given to the part of the scale of each adjective pair where the relationship between the sample and the adjective seems to be most appropriate. As shown in the figure, the scale at this time is assumed to have seven levels of positive and negative from "very" to "neither", and points are scored from +3 to -3. Here, in order to investigate the relationship with the color parameters, for example, as shown in FIG. 12, assuming that the three parameters of hue, lightness, and saturation take three kinds of values, respectively, a total of 27 kinds of test pieces are tested. The item will be presented to the subject. Then, the evaluation with the pair of adjectives is performed for each sample.

A factor analysis is performed on the evaluation result of the adjective pairs to extract common factors. FIG. 13 shows an example of the extracted factors and factor loads. In this example, three factors, I, II, and III, can be extracted from 10 adjectives, and interpret the sense of activity, the sense of expansion, and the sense of complexity, respectively. In short, it is shown that the main measure for color is composed of, for example, these three factors. Here, the factor loading indicates the ratio of loading of the adjective pair on the factor, where a ₁ > a ₂
It is assumed that there is a relation of ＞ ・ ・ ・ ＞ a ₁₀ , b ₁ ＞ b ₂ ＞ ・ ・ ・ ＞ b ₁₀ , c ₁ ＞ c ₂ ＞ ・ ・ ・ ＞ c ₁₀ . In this example, the adjective pair number 5,4,, 2,10
Is interpreted as a factor of activity, 1,8,3 as a factor of expansion, and 9,6,7 as a factor of complexity. Further, the relationship between the color parameter and the adjective pair is calculated, and an example of the result is shown in FIG.
d, e, f are correlation coefficients of adjective pairs for hue, lightness, and saturation, respectively, and d ₁ ＞ d ₂ ＞ ・ ・ ・ ＞ d ₁₀ , e ₁ ＞ e ₂ ＞ ・ ・ ・ ＞ e ₁₀ ,
It is assumed that they have values of f ₁ > f ₂ >...> f ₁₀ . The hatched portion is a portion showing a high correlation equal to or higher than a certain threshold. For example, the adjective pair “sharp-dull” has a high correlation with hue. Next, regarding this adjective pair,
FIG. 15 shows an example of plotting factor scores when the hue is changed. According to this example, it is found that the hue should be close to α ₃ for expressing a sharp color, and should be close to α ₁ for a dull color.

The contents to be stored in the subjective scale rule of the knowledge database are the relational expressions of adjectives and parameters as shown in FIG. FIG. 16 shows the relational expression of FIG. 15 converted into a rule format that can be used in fuzzy reasoning. For this conversion, for example, on the axis where the factor score is 0,
Create a member function by dividing the graph of the relational expression in Fig. 15 into two. For example, the rule body is described as follows (if adjective = "sharp", hue is α ₃ ). (If adjective = "dull", hue is α ₁ ). The member ship function of the department is
The part with a factor score of 0 or more in Fig. 15 is assigned to the "sharp" rule, and the part with a factor score of 0 or less is assigned to the "sharp" rule. That is, the factor score is normalized from 0 to 1 and is used as the member ship value. Further, the member function of the antecedent part can be obtained from the factor loading shown in FIG. In the example of FIG. 13, the adjectives “strong”, “sharp”, “bold”, and “flashy” have a high load on the common factor of feeling of activity, and the factors a ₁ to a ₄ respectively. Have a load. Since the member function of the antecedent part needs to give the probability distribution of the input quantity,
The values of a ₁ to a ₄ are set as the member ship values. The membership function created in this way is shown in FIG.

As described above, the art knowledge rule and the subjective scale rule can be constructed. The interpreter unit 4 refers to this as an adjective rule, and searches for an applicable adjective data from the man-machine unit. Then, the final correction parameter is determined by adding the conclusion in the interpreter unit to the reference parameter which is the conclusion obtained from the analyzer unit 3. Specifically, as shown in FIG. 5, in this determination method, the member functions are combined by max operation, the center of gravity position is calculated, and this is used as a correction parameter.

The structure of the man-machine section 7 is as shown in FIG. 18, and there is an adjective input section 71 inside, which accepts adjectives input from a tablet or a keyboard.

The structure of the renderer unit 5 is as shown in FIG. 19, and is a projection conversion unit that converts the coordinates of a three-dimensional figure into a two-dimensional screen.
51, a hidden surface processing unit 52 that erases hidden surfaces, and a shadow processing unit 53 that shades the model surface. The model data 13 and 16 from the modeling database 1 and the correction parameters from the interpreter unit 4 are input, and the result of the rendering process is output to the display unit 6 as image information.

The structure of the display unit 6 is as shown in FIG.
The display control unit 61 converts the image information from the renderer unit into a video signal and displays it on the display.

In the above embodiment, the art knowledge rule 22 shown in FIG.
Can be subdivided into rendering rules 221 used by the renderer and non-rendering rules 226 not directly related to the renderer, as shown in FIG. 21, for example. Further, the rendering rail 221 can be further subdivided into rules such as color scheme 222, texture 223, composition 224, lighting 225, etc.
26 is also divided into makeup 227, clothing 228, renteria 229, etc.
When knowledge is stored by dividing rules into fields as described above, it is necessary to have a function of setting importance of rules for each field. This is a function of the content that attaches particular importance to only the rule of color arrangement. What realizes this function is the rule adjusting mechanism 230 shown in FIG.

The function of the rule adjusting mechanism 230 can be realized by changing the level of the member function as shown in FIG. That is, for a rule in an unimportant field, the level of the membership function of the antecedent and consequent parts of the rule is forcibly lowered. In the same figure, the dotted line part is the member function before adjustment. By lowering the level of the function, the rate of contribution of the rule is reduced in the calculation of the center of gravity for obtaining the output value, and the obtained output value changes to the one having the smaller degree of influence of the rule.

As described above, according to the present embodiment, since it is possible to adjust the strength of the importance of the rule for each field, there is an effect that the intention of the video effect can be more finely instructed.

In the above embodiment, as shown in FIG. 23, the man-machine section 7 can be further provided with an input section 72 directly. The direct input unit 72 does not input a physical parameter such as hue, but uses the data used to construct the subjective scale rule 23. An outline of the man-machine section 7 and the interpreter section 4 for this purpose is shown in FIG. As illustrated in FIG. 13, since there are three factors of the subjective scale for determining the color, the man-machine section 7 is provided with three input axes of the feeling of activity, the feeling of expansion and the feeling of complexity. For each input axis, it is assumed that an adjective pair with a large factor load is assigned.
Three types are used: “dull”, “bright-dark”, and “simple-complex”. Then, as shown in FIG. 14, the adjective pair “sharp-dull” has a high correlation with hue, “bright-dark” has a correlation with hue and saturation, and “simple-complex” has a correlation with lightness and saturation. Is high. Therefore, each input axis is connected to each parameter control section 42, 43, 44 of the interpreter section 4 only for those having a high correlation. Next, the interpreter unit 4 converts this direct input value into a parameter and outputs it to the renderer unit 5. This conversion uses the relationship between adjectives and parameters as shown in FIG.

As described above, according to the present embodiment, since the input value can be directly instructed, there is an effect that the intention of the video effect can be instructed more finely.

In the above embodiment, the construction of the subjective scale rule 23
As shown in Fig. 13 and Fig. 14, the member results are determined by factoring the evaluation results, removing common factors, and then determining the correspondence with the parameters. However, for the case shown in FIG. 12, the member ship function can be directly determined from the result of the evaluation shown in FIG. That is, instead of extracting the three factor axes as in the above embodiment, the ten adjective pairs used for the evaluation are directly used as the ten factor axes. For example, if the average value of the rating values is simply calculated, the same result as the relationship between the adjective and the parameter shown in FIG. 15 can be obtained.
Therefore, it is possible to construct the subjective scale rule 23 as shown in FIG. 16 also by this method.

As described above, according to this embodiment, there is an effect that the evaluation result can be directly reflected.

In the above-mentioned embodiment, the user can input an adjective as he / she thinks, without particularly limiting the adjectives that can be used. However, it is also possible to limit the use to only adjectives having strong parameter correlation as a result of performing factor analysis in advance. To this end, as shown in FIG. 25, the adjective pairs that can be used on the display are specified so that the user can select from them.

As described above, according to the present embodiment, since only adjectives having a strong correlation with the parameter can be selected, there is an effect that the accuracy in making the intention a parameter is improved.

〔The invention's effect〕

According to the present invention, since the intention of the video effect at the time of rendering processing can be instructed by the adjective, it becomes easier to reflect the intention as compared with the case of directly operating the physical parameter. Therefore, a desired video effect can be easily performed, so that the number of man-hours for rendering can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is an explanatory diagram of color parameters, FIG. 3 is an explanatory diagram of a modeling database, FIG. 4 is an explanatory diagram of an analyzer section, and FIG. 5 is fuzzy reasoning. FIG. 6 is an explanatory diagram of a mechanism, FIG. 6 is an explanatory diagram of a knowledge database, FIG. 7 is an explanatory diagram of an example of a noun rule, FIG. 8 is an explanatory diagram of an interpreter section, and FIG. 9 is an explanatory diagram of an example of art knowledge. Fig. 10 is an illustration of the art knowledge rule, Fig. 11 is an illustration of the adjective grading method, Fig. 12 is an illustration of the parameters at the time of grading, Fig. 13 is an illustration of an example of factor loading, and Fig. 14
Figure is an illustration of an example of the correlation coefficient of parameters, Figure 15 is an illustration of an example of the relationship between parameters and adjectives, Figure 16 is an illustration of an example of subjective scale rules, and Figure 17 is the antecedent of subjective scale rules. 18 is an explanatory view of the man-machine section, FIG. 19 is an explanatory view of the renderer section, FIG. 20 is an explanatory view of the display section, and FIG. 21 is a subdivided rule of FIG. FIG. 22 is an explanatory view when an adjusting mechanism is added, FIG. 22 is an explanatory view of processing of the rule adjusting mechanism, FIG. 23 is an explanatory view when a direct input unit is added to FIG. 18, and FIG. 24 is a direct input unit. Explanatory drawing of processing, No.
FIG. 25 is an explanatory diagram when the adjective is limited and used. 1 ... Modeling database, 2 ... Knowledge database, 3 ... Analyzer section, 4 ... Interpreter section, 5 ...
… Renderer section, 6 …… Display section, 7 …… Man-machine section, 21 …… Adjective rule, 22 …… Art knowledge rule, 23
…… Subjective scale rule, 24 …… Noun rule.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor ▲ Taka ▼ Hiroaki Tsukihiro 1-280, Higashi Koikeku, Kokubunji City, Tokyo Inside Design Research Laboratory, Hitachi, Ltd. (72) Inventor Ota ▲ Yoshi ▼ 4026 Kuji Town, Hitachi City, Ibaraki Prefecture Address: Hitachi Research Laboratory, Hitachi, Ltd. (72) Masao Yanaka 4026 Kuji Town, Hitachi City, Hitachi, Ibaraki Address: Hitachi Research Laboratory, Hitachi, Ltd.

Claims (9)

[Claims] 1. A three-dimensional computer graphics rendering system that obtains video information by performing a rendering process from modeled information of a shape, which represents model data defining the shape and meaning of the shape. A modeling database that holds modeling data configured by associating with noun data, an analyzer unit that extracts the noun data from the modeling data, and converts the noun data into a reference parameter, and a reference converted by the analyzer unit. Parameters and
It is characterized by comprising an interpreter unit for generating a parameter used for a rendering process from the adjective information inputted by the user, and a renderer unit for fetching the parameter generated by the interpreter unit and performing a render render process to output the video information. Rendering system for 3D computer graphics. 2. The analyzer unit and interpreter unit have a fuzzy inference mechanism, and by using each rule in a knowledge database holding a noun rule and an adjective rule, the analyzer unit and the interpreter unit respectively detect the reference parameter. The rendering system for a three-dimensional computer graphics according to claim 1, wherein the conversion or the generation of the parameter is performed by inference. 3. The adjective rule according to claim 3, wherein the renderer section provides an art knowledge rule based on an artist's or designer's expertise in image production and an adjective expressing a user's intention to produce an image. A rendering system for a three-dimensional computer graphics, comprising: a subjective scale rule for converting into a parameter for a rendering process to be executed. 4. The fuzzy inference mechanism in the interpreter unit makes inference by using a member ship function calculated from the relationship between the adjective that the user can input and the parameter. Computer graphics rendering system. 5. A rendering system for a three-dimensional computer graphics, wherein the parameter according to claim 1 is a color parameter, a texture parameter, a composition parameter, or an illumination parameter. 6. The subjective scale rule according to claim 3, wherein the subject is presented with a sample product in advance by a combination of a plurality of adjectives and a plurality of parameters, and the subject is evaluated. A rendering system for three-dimensional computer graphics, which is a relational expression between the adjectives and the parameters that have been defined. 7. A three-dimensional computer-graphics rendering system, wherein the level of the membership function according to claim 4 is varied to weight rules in a specific field and generate the parameters. . 8. The three-dimensional computer according to claim 1, wherein the adjective information input by the user is input by displaying a predetermined adjective pair in a menu in advance and selecting from the adjective pair displayed in the menu. Graphic system rendering system. 9. Image information is obtained by performing a rendering process from modeled information such as shape or motion.
A three-dimensional computer-graphics rendering system, characterized in that a parameter required for the rendering process is determined by giving an adjective for image production in the three-dimensional computer-graphics rendering system.