JP2000306118A - Tree texture generating device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a tree texture generating device which automatically constructs a three-dimensional tree model holding the reality of an actually photographed picture of a tree based on the picture. SOLUTION: This tree texture generating device picks up the image of an object tree from 1 plural view points by a camera 2 and generates volume data by a computer 3 for picture generation and generates a tree model consisting of segments of branches and the trunk based on the volume data and extracts bark reference pixels of the generated tree model to update color information of a texture template of the bark read out from a disk 5 and extracts leaf surface reference pixels from pixels, where bark reference pixels are removed, and updates color information of a texture template of leaves read out from the disk 5 based on the color information of these leaf surface reference pixels and generates a tree model with updated bark pixels and leaf pixels.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a tree texture generating apparatus, and more particularly to a tree texture generating apparatus capable of automatically constructing a realistic three-dimensional tree model based on a photographed image.

[0002]

2. Description of the Related Art In recent years, research on a communication environment via virtual scenes of persons at a distance from each other has been advanced. In order to realize such an environment, it is necessary to establish a virtual scene generation method.

As a general method for generating such a virtual scene, three-dimensional computer graphics (hereinafter referred to as C
G). In this method, objects in the scene are modeled as geometric objects, and the scene viewed from the virtual camera is rendered. in this way,
In a modeled object, deformation, movement, and the like can be easily expressed by setting attributes close to reality. Further, since it has a complete three-dimensional structure, it is easy to generate an image viewed from an arbitrary viewpoint.

[0004]

As a technique of three-dimensional CG, the expression of an artificial object has been relatively studied, and the texture has been improved. On the other hand, it is not easy to model natural objects and natural phenomena such as trees. In particular, when modeling a tree or the like that already exists, it is currently performed by the CG artist's sensitivity and manual work, which requires a great deal of labor and time.

SUMMARY OF THE INVENTION Accordingly, a main object of the present invention is to provide a tree texture generating apparatus capable of automatically constructing a three-dimensional tree model retaining its reality based on a photographed image of the tree. It is.

[0006]

The invention according to claim 1 is
A tree texture generation device that generates bark and leaf textures based on an image of a target tree, and volume data creation means that creates volume data that wraps the tree based on images from multiple viewpoints, A tree model creating means for creating a tree model composed of branches and trunk segments based on the created volume data, and specifying the position of the trunk of the created tree model, defining surrounding pixels as bark reference pixels, Bark pixel extracting means for limiting the color range of the bark reference pixel and extracting pixels in the color range, and a pixel indicating the surface of the leaf as a pixel remaining in the image obtained by removing the bark reference pixel by the bark pixel extracting means. Leaf pixel extracting means for extracting as surface reference pixels, and leaf and bark textures each containing color information are stored in advance as templates A bark updating unit that updates the color information of the bark texture template read from the storage unit based on the extracted bark reference pixel color information; and a bark updating unit that extracts the leaf surface reference pixel color information. Updating means for updating the color information of the leaf texture template read from the storage means, and rendering means for creating and rendering a tree model with the bark pixels and the leaf pixels updated by the bark updating means and the leaf updating means, respectively. It is comprised including.

In the invention according to claim 2, the bark pixel extracting means according to claim 1 calculates an average and a variance of hue (H), saturation (S), and brightness (V) of a bark reference pixel, and obtains an HSV color space. The color range within.

According to the third aspect of the present invention, the leaf pixel extracting means of the first aspect includes a hue (H), a saturation (S),
The average and variance of lightness (V) are obtained, and the color range in the HSV space is limited.

[0009] The invention according to claim 4 further includes photographing means for photographing a target tree from a plurality of viewpoints, and removes a background image from an image photographed before creating volume data. The transformation is performed to normalize the tree size and stem position.

In the invention according to claim 5, the normalized volume data is shaped into a tree shape.

[0011]

FIG. 1 is a block diagram of one embodiment of the present invention. In FIG. 1, a plurality of cameras 2 for photographing the target tree 1 from a plurality of viewpoints are installed around the target tree 1. The camera 2 may be a camera using a CCD element or an optical camera. Image data captured by the camera 2 is provided to an image generation computer 3.

The image generating computer 3 mainly has a texture creating function, a volume creating function, a tree model creating function, and a rendering creating function in order to create a tree model from the image data input from the camera 2. ing.

The texture creation function creates a tree texture according to a program based on a flowchart shown in FIG. The volume data creation function creates volume data surrounding a tree based on images taken from a plurality of viewpoints captured by the camera 2. The tree model creation function creates a tree model composed of branch and trunk segments based on the created volume data. The rendering function adds a leaf model to the created tree model and renders the tree model.
Create an image of a tree with the shape of The tree image generated by the image generation computer 3 is displayed on the display 4.

FIG. 2 is a flowchart for explaining a specific operation of the embodiment of the present invention. FIG. 3 is a diagram showing an example of a tree of an original image. FIG. 4 is a diagram for removing a background from the original image. FIG. 7 is a diagram showing an example of a processed image.

Next, a specific operation of the embodiment of the present invention will be described. Trees of the original image as shown in FIG. 3 are photographed from the surroundings by the plurality of cameras 2 shown in FIG. Subsequently, as shown in FIG. 4, the background is removed from the photographed image, color conversion is performed, and the tree size and the position of the trunk are normalized. Background image removal is performed manually, and color conversion is performed by gray scale conversion. After conversion, dynamic range normalization is performed to obtain a rectangle including the target tree 1 and normalize its size. Thus, the size of the target tree 1 is obtained. Next, volume data is created by the image generating computer 3.

FIGS. 5 and 6 are diagrams for explaining a method of creating volume data. In order to create volume data from live-action video, the concept of silhouette back projection is introduced. When the silhouette obtained from the image is antiparallel-projected into a three-dimensional space, an infinite-length column having the silhouette at this time as a cross section is obtained as shown in FIG. When the same processing is repeated with images from a plurality of viewpoints, an area where the subject overlaps, that is, an area where the target shape exists as a logical expression is limited. The volume data obtained in this way is not the shape itself of the target tree 1, but can be defined as a shape including the shape.

In this embodiment, unlike the simple silhouette back projection method, the volume data has 256 gradation values as the volume data.

Next, the volume data is shaped. When a large number of images of the object are prepared over the entire circumference, the shape approximation ability of the required volume data also increases. However, when the number of provided images is small, an unintended shape is generated. For example, if an image captured by two cameras as shown in FIG. 5 is read and the two cameras are arranged at right angles, the volume data to be obtained is a rectangular parallelepiped as shown in FIG. The shape is close to, and is obviously different from the original tree shape.

In order to avoid this, volume data obtained by back projection is shaped. Generally, a tree has a shape close to a circle when viewed from above vertically. Therefore, the outermost point when the trunk is centered for each scan line of the captured image after processing is determined, and the distance from the trunk is defined as the outermost point of the volume data. Shaping can be performed by deleting the voxels existing outside the outermost point.

The input pixels of the tree shaped in this way are stored in a step S1 shown in FIG.
Is converted to the HSV color system. Here, the HSV color system is a color system represented by hue (H), saturation (S), and lightness (V). In step S2, a root region of the trunk is detected.

FIG. 7 shows the root of the trunk. As shown in FIG. 7, the root regions of the trunk are all represented as segments having the same length and the same thickness. Subsequently, in step S3, a pixel group included in the root region of the trunk is defined as a bark reference pixel. In step S4, the average and variance of the hue, saturation, and lightness of the bark reference pixel are obtained, and the color in the HSV color space is obtained. The range is limited, and pixels in the color range are extracted from the entire image.

On the other hand, from disk 5 to 4 as shown in FIG.
The leaf and bark texture templates are read out and converted to the HSV color system in step S5. 8A shows a leaf texture template, and FIG. 8B shows a bark texture template. In step S6, the average t _avr and the variance t _var of each value of the leaf and bark texture templates are calculated. Then, in step S7, the color information of the bark texture template is updated according to the bark color information in the color range extracted in step S4.

The updated color information tex (u, v) is calculated from the average ravr and variance rvar of the reference pixel, the average tavr and variance tvar of the template, and the pixel value tmp (u, v) of the template according to the following equation. You.

[0024]

(Equation 1)

Although the expression (1) shows the hue value H, the saturation value S and the lightness value V are similarly calculated.

Here, equation (1) will be described. H
_tex (u, v) is the texture of the finally output target image. u and v indicate that they are u and v planes. H _ravr is the average value of the reference image, H _rvar is the variance value of the reference image, H _tmp is the pixel value of the template, H _tavr is the average value of the template, and H _tvar is the variance value of the template. is there.

In equation (1), H _tmp -H _tavr indicates how far the template pixels are from the average. When this is divided by the variance value Ht _var of the template, it indicates the ratio of pixels to the variance. By multiplying this by the variance value H _rvar of the reference image, the pixel value is updated according to the variance of the given image. By adding this ratio to the average value H _ravr of the reference image, the texture for the target image that is finally output is output. As a result, the color of the input image can be approximated while maintaining the variance ratio.

On the other hand, in step S8, pixels included in the color range of the bark defined in step S4 are removed from the input image, and in step S9, the remaining pixel groups are used as leaf reference pixels. In step S10, the average ravr and the variance rvar of the leaf reference pixel group are calculated.

In step S11, the average tavr and variance tvar of each value, which is the color information of the leaf texture template calculated in step S6, are calculated using the average ravr and variance rva of the leaf reference pixel group calculated in step S10.
In accordance with r, it is updated using the above-mentioned equation (1). Thereby, the pixel value is updated at a rate corresponding to the variance of the given image. Then, in step S12,
A texture as shown in FIG. 9 is generated based on the bark pixels and leaf pixels updated in steps S7 and S11, respectively, and displayed on the display unit 4. An example of a tree texture displayed on the display unit 4 is shown in FIG.

The embodiments disclosed this time are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

[0031]

As described above, according to the present invention, the position of the trunk of the tree model obtained from a plurality of viewpoints is specified, and the range in which peripheral pixels are defined as bark reference pixels is specified. The pixels remaining after removing the bark reference pixels in the range are extracted as leaf surface reference pixels, the color information of the bark texture template is updated based on the color information of the extracted bark reference pixels, and the extracted leaf surface By updating the color information of the leaf texture template based on the color information of the reference pixels and creating a tree model using the updated bark pixels and leaf pixels, a model having a tree-like branch structure can be constructed.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram of an embodiment of the present invention.

FIG. 2 is a flowchart for explaining a specific operation of the embodiment of the present invention;

FIG. 3 is a diagram illustrating an example of a tree of an original image.

FIG. 4 is a diagram illustrating an example of an image obtained by removing a background from an original image.

FIG. 5 is a diagram showing a method of creating volume data.

FIG. 6 is a diagram illustrating a method of shaping volume data.

FIG. 7 is a diagram showing a root of a tree trunk.

FIG. 8 is a diagram showing an example of a leaf and bark texture template.

FIG. 9 is a diagram illustrating a tree of a generated three-dimensional model.

FIG. 10 is a diagram showing an example of a tree image rendered and displayed.

[Explanation of symbols]

 1 target tree 2 camera 3 computer for image generation 4 display unit 5 disk

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Atsushi Otani 5 Sanraya, Seiya-cho, Soraku-cho, Kyoto, Japan F-Term in Intelligent Telecommunications Imaging and Communications Laboratory, Inc. 5B050 BA11 CA07 DA04 EA06 EA28 FA02 5B057 BA02 DA06 DB03 DB06 DC30 DC31

Claims (5)

[Claims] 1. A tree texture generating apparatus for generating bark and leaf textures based on an image of a target tree, wherein the apparatus generates volume data enclosing the tree based on images from a plurality of viewpoints. Volume data creating means, a tree model creating means for creating a tree model composed of branches and trunk segments based on the volume data created by the volume data creating means, a trunk of the tree model created by the tree model creating means Bark pixel extraction means for defining the position of the bark reference pixel, defining the surrounding pixel as a bark reference pixel, limiting the color range of the bark reference pixel, and extracting pixels in the color range, the bark pixel extraction means A leaf pixel for extracting a pixel remaining in the image from which the pixel has been removed as a leaf surface reference pixel indicating the leaf surface Output means, storage means for storing in advance the leaf and bark textures each including color information as a template, bark read from the storage means based on the color information of the bark reference pixels extracted by the bark pixel extraction means A bark updating unit for updating the color information of the texture template, and updating the color information of the texture template of the leaf read from the storage unit based on the color information of the leaf surface reference pixels extracted by the leaf pixel extracting unit. A tree texture generating apparatus, comprising: a leaf updating unit; and a rendering unit configured to create and render a tree model using the bark pixel and the leaf pixel updated by the bark updating unit and the leaf updating unit, respectively. 2. The bark pixel extracting means calculates an average and a variance of hue (H), saturation (S), and brightness (V) of the bark reference pixel, and limits a color range in an HSV color space. The tree texture generating device according to claim 1, wherein: 3. The leaf pixel extracting means calculates an average and a variance of hue (H), saturation (S), and lightness (V) of the leaf reference pixel, and limits a color range in an HSV space. The tree texture generating device according to claim 1, characterized in that: 4. A photographing means for photographing the target tree from a plurality of viewpoints, wherein a background image is removed from the photographed image and color conversion is performed before creating the volume data. 2. The tree texture generating apparatus according to claim 1, wherein the tree size and the position of the trunk are normalized by performing the following. 5. The method according to claim 1, wherein the normalized volume data is shaped into a tree shape.
A tree texture generating apparatus according to claim 4.