JP2002230582A - Method for generating three-dimensional shape model from pattern shape model and image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate a three-dimensional shape model similar to an impression of an image from a three-dimensional shape model as a pattern and an image as a target. SOLUTION: A frame of a pattern shape model is deformed so that an outline of an image which the pattern shape model is projected, and an outline of a target image approach, and then a cross sectional shape of the pattern shape model is scaled so as to make the outline of the image which the pattern shape model is projected coincide with the outline of the target image.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a method for generating a three-dimensional shape model in a three-dimensional CAD system and a computer graphics system.

[0002]

2. Description of the Related Art Conventionally, as a method of generating a smooth three-dimensional shape model from an outline of an image, Japanese Patent Application No. Hei 8-17164.
Two have been proposed. In addition, as a method of deforming the three-dimensional shape model, a method in which the operator moves one point on the three-dimensional shape surface, deforms a shape within a certain distance from the point to be moved, associates the shape model with its skeleton, There is a method (for example, “Bone function” of three-dimensional computer graphics software LightWave6 sold by Day Storm Co., Ltd.) that only deforms a skeleton element and deforms each part of the corresponding shape model. . Furthermore, a method of generating a realistic shape model similar to an image impression by deforming a three-dimensional shape model serving as a model to the contour of a target image for a human head (Iwanami A science library 62 "Invitation to face amount") has also been proposed.

[0003]

SUMMARY OF THE INVENTION Conventional Japanese Patent Application No. Hei 8-17
In the method of 1642, when the three-dimensional shape model is generated from the contour line of the target image, the cross-sectional shape perpendicular to the center axis of the shape model is always circular, and a very realistic shape model cannot be created. . For example, if a three-dimensional shape model is generated from a contour line when a human body is viewed from the front, the cross-sectional shape of the torso portion is circular, but the cross-section of the actual torso is almost elliptical.

[0004] Further, when trying to deform a model shape model in accordance with a contour line of a target image using a conventional method of deforming a three-dimensional shape model, an operator manually deforms each part of the shape model. Had to be done in and was very time consuming.

Further, when using a conventional method for deforming a three-dimensional shape model, the model shape model is deformed in accordance with the contour of the target image, and the operator only has to place the model shape model horizontally on the target image. Since it did not deform, the characteristics of the original shape model were lost after the deformation. For example, when deforming a shape with many branches, such as a human body or an animal, if the shape is deformed according to the thickness of the arms and legs,
The cross-sectional shape of the arm or leg becomes a flat ellipse, and it is impossible to create a realistic shape model.

According to the present invention, in a three-dimensional shape having many branches, a three-dimensional model model and a target image are used.
An object of the present invention is to easily generate a realistic three-dimensional shape model that retains the features of the model shape model and resembles the impression of the target image.

[0007]

In order to achieve the above-mentioned object, the present invention is carried out according to the procedure shown in FIG. First, in step 101, a target image is input. Next, step 10
In step 2, the operator selects a model model that matches the contents of the target image from a plurality of model models prepared in advance. Subsequently, in step 103, the operator deforms the framework of the model shape model according to the target image. Then, in step 104, each part of the model model is automatically deformed according to the deformed skeleton. Further, in step 105, the cross-sectional shape of the template shape model is automatically scaled so that the contour of the image on which the template shape model is projected matches the contour of the target image. Here, the finally deformed model shape model becomes the three-dimensional shape model obtained by the present invention, and the process ends in step 106.

[0008] Each procedure will be described in detail. First, Step 10
The input of one target image will be described. The target image is a photograph of an object, such as a person, an animal, or an aircraft, for which a three-dimensional shape model is created, or an illustration depicting the object. It is desirable that the target image be photographed or drawn from an angle at which the characteristics of the target object can be most easily captured. For example, the human body is better shot from the front than from the top or side,
For quadrupeds such as dogs and cats, images taken from the side, and for airplanes etc., images taken from the top are suitable. The input target image is displayed on a computer display.

In the next step 102, the operator selects a model model prepared in advance that suits the target of the target image. For example, when a photograph of a human body is a target image, a model of a human body is usually selected. In addition, the angle at which the target image is photographed (drawn) and the angle at which the model model is projected and displayed as a two-dimensional image are made to match. For example, when the target object of the target image is photographed from the front, the model model is also projected and displayed on the projection plane from the front. At this time, on the computer display,
Display over the target image.

The following procedure 103 will be described. Procedure 10
The skeleton conforming to the state of the branch of the shape is associated with the model shape model selected in 2 in advance. The framework is composed of a wire frame model composed of straight lines and vertices. For example, for a human model model as shown in FIG. 2A, the limbs correspond to the framework as shown in FIG. 2B. are doing. The operator moves each vertex of the skeleton horizontally to the projection plane and deforms it according to the shape of the target in the target image. FIG. 3A shows the state before the deformation of the skeleton.
(B) is a state after the frame deformation. Note that, in practice, the model shape model and the target image are superimposed on the projection plane, but FIG. 3 shows the model shape model and the target image side by side for explanation.

Subsequently, in step 104, as shown in FIG. 3C, each part of the model model is automatically deformed according to the deformed skeleton. That is, if each straight line of the skeleton has been translated after deformation, the portion of the model shape model corresponding to that straight line has also been translated, and if each straight line of the skeleton has been rotated, the model corresponding to that straight line has been rotated. The shape model portion is also rotated, and if the length of each straight line of the skeleton changes, the portion of the model shape model corresponding to the straight line is enlarged or reduced according to the length after deformation. At this time, the contour when the deformed model shape model is projected and displayed on the projection plane has a shape close to the contour of the target image, but does not completely match yet. That is, the deformed model model is a three-dimensional model that resembles the impression of the target image to some extent.
The thickness for the skeleton is significantly different from the target image.

In the next step 105, the cross-sectional shape of the model model is automatically adjusted so that the outline when the model model is projected on the projection plane and displayed matches the outline of the target image. Scaled. FIG. 4A is a diagram schematically showing a cross section when the model model is cut perpendicularly to a straight line of the skeleton. The cross-sectional shape 405 of the template shape model is scaled, further horizontally moved, and deformed to have the same width as the contour line 403 of the target image. That is, it is deformed as shown in FIG. By deforming the cross-sections at all positions in this way, the contour line when the model model is projected and displayed on the projection plane matches the contour line of the target image as shown in FIG. 5B. I will be. FIG. 5A shows a state before deformation of the model model, and FIG. 5B shows a state after deformation.

Finally, the process ends in step 106. Finally, the deformed model shape model becomes the three-dimensional shape model obtained by the present invention.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described. When the three-dimensional shape is a solid model of a boundary expression constituted by polygons, the deformation of the model model in each step is performed by moving the vertex coordinates of the polygon.

On the other hand, when the procedure 105 is performed, a cross section perpendicular to the skeleton cannot be obtained at the connection points of the skeleton, the branch points of the skeleton, and the ends of the skeleton.
The following special processing is performed for each.

At the connection point of the skeleton, first, FIG.
As shown in (a), the rotation center 604 is set near the connection point 603 of the skeleton. That is, when the skeleton of the model shape model is projected on the projection plane, a straight line that bisects the angle on the side where the angle of the skeleton at the connection point 603 is 180 degrees or less, and a contour line 602 of the target image The intersection of 1.2
The extended position of about 1.5 to 1.5 times is set as the rotation center 604. In the vicinity of the connection point of the skeleton, as shown in FIG. 6B, a cross-sectional shape obtained by cutting the model model at each cross-section 606 radially extended around the rotation center 604 and perpendicular to the projection plane is obtained. The scale is automatically scaled so that the cross-sectional shape matches the contour of the target image.

At the branch point of the skeleton, first, the average value S of the enlargement ratio of the cross section of the model model near the branch point is determined. Next, as shown in FIG. 7, the rotation centers 706 are respectively set on straight lines obtained by bisecting the angles of the adjacent skeletons on the projection plane, and the rotation centers 706 are centered, and the rotation centers 706 are perpendicular to the projection plane.
A cross-sectional shape obtained by cutting the model shape model at each cross-section extending radially is obtained. At this time, it is considered that the cross section extending from the rotation center to the skeleton extends at an angle to the next rotation center. That is, in the vicinity of the branch point, a cross section bent in a U-shape is obtained for the frame. Here, at the branch point of the skeleton, two methods are conceivable as a method of deforming the cross section of the cross shape. That is, a method of enlarging and reducing the cross-section of the V-shape as it is, and a method of dividing the cross-section of the V-shape into two parts with the skeleton as a boundary (however, enlargement in the direction perpendicular to the projection plane) The ratio is always an affine transformation fixed at S times). In the former method, a step occurs at the branch point, and in the latter method, the further away from the branch point, the less appropriate the shape of the depth of the projection plane. Therefore,
The former and the latter are deformed to have an intermediate shape of the cross-sectional shape deformed by both methods. That is, the cross section passing through the branch point is deformed only by the latter method, approaches the shape deformed by the former method as it moves away from the branch, and is deformed only by the former method at a place sufficiently distant from the branch.

At the end of the skeleton, by extending the skeleton in a straight line to the tip of the model model, a cross section of the model model is obtained, which is automatically scaled and deformed.

[0019]

Since the present invention is configured as described above, it has the following effects.

Since the operator only needs to deform the skeleton in accordance with the target image after selecting the model model, a three-dimensional model can be easily generated.

Since the template shape model is deformed so that the contour line when the template shape model is projected and displayed on the projection plane matches the contour of the target image, a three-dimensional image similar to the impression of the target image is obtained. A shape model can be generated.

Since the cross-sectional shape perpendicular to the skeleton is always deformed to a similar shape, a three-dimensional shape model can be generated while maintaining the features of the original model shape model. For example, in a human body, the cross-sectional shapes of arms and legs of a fat person and a thin person are almost similar, so that a realistic three-dimensional shape model can be generated from a single model model for both fat and thin persons.

Since the skeleton structure of the model shape model is utilized to the maximum extent, it is possible to efficiently generate a three-dimensional shape model having many branches such as a human body and an animal.

It is necessary to associate a skeleton with the template shape model in advance. However, even with a three-dimensional shape model generated on a conventional computer graphics system, the skeleton is already associated with a moving image. Often attached. These existing three-dimensional shape models can be immediately used as model shape models.

[Brief description of the drawings]

FIG. 1 is a flowchart showing a procedure of a process of the present invention.

FIG. 2 is a diagram showing a model shape model and its skeleton.

FIG. 3 is a diagram showing a transformation operation of a procedure 104.

FIG. 4 is a diagram showing details of a modification of the procedure 105.

FIG. 5 is a diagram showing a state of a modification of the procedure 105.

FIG. 6 is a diagram illustrating a deformation operation at a connection point of a skeleton.

FIG. 7 is a diagram showing a deformation operation at a branch point of the skeleton.

[Explanation of symbols]

 101 to 106 Procedures of the present invention 201, 301, 401 Model shape models 501, 601, 701 Model shape models 202, 302, 402, 605, 703 Frame 303, 502 Target image 403, 602, 702 Outline of target image Line 404 Projection plane 405 Cross-section of template model 603 Frame connection point 604, 706 Rotation center 606 Cross-section group 704 used for deformation 704 Cross-section group 705 Cross-section group near branch point 707 Branch point of skeleton

Claims (3)

[Claims] 1. A method of generating a three-dimensional shape model from a model shape model and a target image, wherein an operator can select the model shape model according to the contents of the target image. How to generate the model. 2. The three-dimensional shape model according to claim 1, wherein the model shape model is deformed by moving the framework of the model shape model so that the outline of the image projected from the model shape model and the outline of the target image are close to each other. Generation method. 3. The method according to claim 3, wherein the shape of the cross-section of the model is scaled so that the outline of the image projected from the model matches the outline of the target image.
How to generate a two-dimensional shape model.