JP2001022939A - Method and device for extracting three-dimensional information and recording medium with three- dimensional information extraction program recorded therein - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce an area taken by a storing means for holding a voxel space, suppress voting errors and extract color information. SOLUTION: A voxel space generating part 11 selects a plurality of images obtained, by photographing the entire shape of a subject as a contour image among photographed images, generates voxel space information by using the contour information of a subject area on the contour images and stores the voxel space information in an information storage part 15. A voxel generating part 12 defines the prescribed number of images as a reference image among a plurality of images and their adjacent images as reference images, generates a voxel which is a unit three-dimensional space going through a characteristic point in a voxel space about the characteristic point on the reference images and stores the voxel and color information C of the original characteristic point in the part 15. A voxel information comparing part 13 extends a projection line from the generated voxel toward each reference image and compares the color information of an intersection with the reference image with the information C. A three-dimensional information extracting part 14 extracts the three- dimensional coordinates and color information of the original characteristic point, when the difference between the color information and the information C is below a threshold.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for extracting three-dimensional information for extracting a three-dimensional shape of a photographing subject.

[0002]

2. Description of the Related Art As a technique for restoring the three-dimensional coordinates of a subject, as shown in FIG. 10, an image based on a large number of two-dimensional images of the same subject taken from different viewpoints using a turntable or the like is conventionally known. By extending a straight line from the viewpoint to a feature point representing an edge or a boundary area of the object appearing above, and passing the object through the three-dimensional space where the object exists,
There is known a method of voting (voting) to three-dimensional coordinates where a feature point of a subject may exist (FIG. 11).
reference). Note that, based on a plurality of two-dimensional images obtained by imaging the same subject from different viewpoints, a straight line extending from a feature point on the subject in the three-dimensional space toward the viewpoint is referred to as a “projection line”. "Projection" means passing a projection line
That. Conversely, a straight line extended from the viewpoint toward the feature point of the subject appearing on the image is called a “back-projection line”,
Passing a backprojection line onto a three-dimensional space in which a subject exists is called “backprojection”. Also, as shown in FIG. 11, in the voting method, a “voxel space” that is a voting target space is defined as a set of “voxels” that are unit three-dimensional spaces. Is passed, the voting operation is performed on the voxel. Here, the "voting operation" is to increase the function value of the voxel through which the back projection line passes by 1, and this function often represents the number of back projection lines passing. Also,
In order to execute a voting operation, camera parameters (photographing position and direction) at the time of photographing need to be known.

[0003] As a conventional method in the three-dimensional coordinate restoration technique, for example, reference 1 by Hamano et al., “A method of extracting three-dimensional environment information by voting into space” (Research in IEICE
I), J75-D-II, pp. 342-350,199
In 2-02), a storage means for storing the number of votes is previously provided for all voxel positions in the three-dimensional voxel space, and back projection lines are extended from all feature points on all captured images, and The number of voxel back-projected lines that pass is measured as the number of votes. Then, a voxel position where the number of votes is the maximum is searched for in the three-dimensional voxel space, and the voxel position where the number of votes is the maximum is the three-dimensional feature point. Extracted as position. However, this method has a problem that voxels through which backprojection lines from feature points that are not originally corresponding points pass have a local maximum value, and in that case, an incorrect voting result (erroneous voting point) is obtained. is there.

On the other hand, Kawato's reference 2 "Extraction of three-dimensional information by two-stage voting into space" (IEICE (D-II), J
77-D-II, pp. 2334-2341, 1994
12) proposes a method of preventing erroneous voting points by performing a voting operation twice. In this method, the number of passages for all voxel positions through which all back projection lines pass is measured in the same manner as in the method of the above-mentioned document 1, and then the fact that only one feature point exists on the back projection line is used. The extraction of erroneous voting is suppressed by the second passage number measurement. That is, in the method of Reference 2, two storage means for storing the number of votes are prepared for all the voxel positions in the three-dimensional voxel space, and the voting operation for all back projection lines is performed in the same manner as in the method of Reference 1. And the resulting vote count is k ₁ , the vote count k ₁ is stored for each voxel, the backprojection line from the feature point on each image is drawn again, and each voxel on the backprojection line is drawn. locate the position of the voxel votes k ₁ corresponding to the the maximum, to vote operation on the corresponding voxels of the second time of the number of votes storing means only for voxels votes k ₁ is the largest. The maximum value search and voting operation and result of for all backprojection line, the number of votes stored in each voxel positions and k _2. Finally, k ₂ in each voxel is obtained, those whose values are equal to or larger than a certain threshold value are extracted, and the positions are output as the three-dimensional coordinates of the feature points. In this way, the maximum value of the number of votes corresponding to each voxel on one back projection line is extracted, and therefore, the method of Document 2 suppresses the extraction of erroneous voting points. In addition, the present method employs an octree-type data structure to represent a voxel space. In the method, first, a sufficiently large voxel including the object region is prepared, the above-mentioned voting operation is performed on the voxel twice, and only the voxel for which the voting result is obtained is divided into eight equal parts (see FIG. 12). ), A method of repeatedly executing the voting operation twice and repeating the voxel division to a required resolution while repeatedly executing the voting operation twice. By adopting this structure, it is possible to reduce the storage means and the useless voxel search required for configuring the space. Voxel division is performed until one back projection line passes through one voxel.

On the other hand, Hosoya et al., Reference 3, "Method and apparatus for extracting three-dimensional information of feature points" (Japanese Patent Application No. 7-32299).
4) proposes a method of treating a voxel space as a set of two-dimensional voxel planes. In this method, the captured image is classified into two types, a reference image and a reference image, and the back projection line L ₂ of the feature point on the reference image is set for each voxel through which the back projection line L ₁ of the feature point on the reference image passes. Is performed, the number of votes is increased by one. The voting is performed while sequentially following the voxel space from a voxel plane close to the reference image to a plane far from the voxel plane. At this time, not all voxels in the voxel space are stored as information,
The number of feature points on the image, that is, the number of backprojection lines, is set, and the voting operation from the reference image is performed along the trajectory where the backprojection lines pass through the voxel space, thereby reducing unnecessary hardware area. . Further, at the time of voting, the position of the voxel having the largest number of votes on the back-projected line from the reference image is always stored, thereby avoiding repetition of the search in the voxel space and achieving high speed. However, this method still requires that the voxel space be defined in advance.

[0006]

As described above, in the conventional method, the position and the size of the voxel space are designated in advance according to the position and the size in the three-dimensional space including the subject. Need to be kept. In Reference 1,
This is to store the number of votes for all voxel positions in the three-dimensional voxel space, and it is necessary to prepare storage means for storing the number of votes for all voxel positions. There is a problem that means is required. In addition, false voting points are inevitable. on the other hand,
In the method of Reference 2, the voxel space is represented using an octree structure to prevent useless storage areas from being used. However, it is necessary to specify the position and size of the voxel space in advance. In addition, in order to equally divide the voxel for which the voting result is obtained, two voting operations are repeated each time. Therefore, Document 1
Although this method requires less than the conventional method, there is a problem that a storage means area for holding voxel information over the entire subject area is required as a result. In the method of Literature 3, it is expected that a good voting result can be obtained in the vicinity of the reference image by a method in which a voting point having the largest number of votes is always provided on a back projection line from the reference image. But,
A voting result cannot be obtained for a portion that cannot be seen on the reference image. In addition, it is necessary that the reference image and the voxel plane are almost opposed to each other. If the three-dimensional information on the entire circumference of the subject is to be extracted while changing the reference image, the data structure is complicated. Is very difficult.

Further, in the conventional methods, since the coordinates of the feature points of each image are input and the back projection line is extended from each feature point toward the three-dimensional space, the feature points extracted for each image are used. Implicit understanding is necessary, especially when voting using images with similar viewpoints, the error due to quantization greatly changes the voting voxel,
There is a drawback that voting is not performed from an image whose shooting position is close. In addition, information relating to the color of a feature point is not handled, and the only purpose is to restore coordinates.
Furthermore, in the voting methods of Literatures 1 and 2, a voting operation is performed from a part that is not clearly visible, such as when viewed from a direction in which the subject is rotated 180 degrees, in order to perform a voting operation from all captured images. In addition, it interacts with the above-mentioned drawbacks, and further derives erroneous voting points.

An object of the present invention is to reduce the area taken by the storage means for holding the voxel space, and to improve the accuracy of voting to suppress false voting and to extract color information. An object of the present invention is to provide a method and an apparatus for extracting dimensional information and a recording medium on which a program for extracting 3D information is recorded.

[0009]

According to the three-dimensional information extraction method of the present invention, a plurality of images taken from a direction which can restore the entire shape of a subject in a photographed image are selected as contour images, A voxel space generation step of generating voxel space information using contour line information of the subject region, a voxel space storage step of storing the generated voxel space information, and a reference to a predetermined number of images of the plurality of images. An image, a voxel generation unit that generates a voxel that is a unit three-dimensional space passing through the feature point in the voxel space as a reference image using an image in the vicinity thereof as a reference image, and used to generate the generated voxel and the voxel. A step of storing voxel / color information for storing color information C of feature points on the reference image, and extending a boundary line from the generated voxel toward each reference image, A voxel information comparison step of comparing the color information at the intersection with the image and the color information C; and a feature on the reference image used to generate the voxel when the difference between the two color information is equal to or less than a threshold value. There is a three-dimensional information extraction step for extracting three-dimensional coordinates and color information of the point.

First, the step of generating a voxel space will be described. From the captured images, a plurality of images are selected as contour images so that the entire shape of the subject can be seen. Here, a contour line representing a subject area is extracted using chroma key processing.

Now, one of the selected contour images is defined as a reference contour image and the others are defined as reference contour images. Back projection line from viewpoint I ₁ onto point f ₁ on the contour line on the defined reference contour image on three-dimensional space

[0012]

[Outside 1] (FIG. 7). Next, one of the reference contour images is selected, and the back projection line is selected.

[0013]

[Outside 2] Viewpoint I ₂ of the reference outline image to calculate the epipolar plane makes with. Next, the line of intersection between this epipolar plane and the reference contour image

[0014]

[Outside 3] The intersection between the (epipolar line) and the contour on the reference contour image is determined. Intersection is an epipolar line

[0015]

[Outside 4] Are viewed in order from both ends to the inside, and are defined as two points of pixels that first come into contact with the contour line. Two straight lines and back projection lines that pass through these two points and viewpoint I ₂

[0016]

[Outside 5] 2 intersections with

[0017]

[Outside 6] Is the line segment

[0018]

[Outside 7] Can be regarded as indicating a range in which a point on the subject corresponding to the point f ₁ on the contour line may exist. Apply the same operation to all the other reference contour images, and move

[0019]

[Outside 8] And closest

[0020]

[Outside 9] Is selected, the point on the subject corresponding to f ₁ becomes 3
This is a range that can exist in the dimensional space (FIG. 9). If the above operation is performed on points on the contour line on the entire contour image, a three-dimensional convex hull region formed by intersecting back projection lines on all contour lines can be finally created. This includes a three-dimensional space region where the subject always exists, as well as a convex hull region surrounding the subject, and can be used as an outline of the voxel space at the time of voting operation.

Next, the voting operation will be described. Here, one of the captured images is selected as a reference image, and a plurality of images are selected as reference images.

A voxel is generated on a back projection line drawn from the feature point on the reference image into the outline of the generated voxel space. Here, the color of the feature point is C. Draw a projection line from the center of each generated voxel to each reference image,
The color at the intersection with the reference image is compared with C. If the difference is equal to or less than the threshold, a voting operation is performed on the voxel. Here, since the back projection line from the reference image exists only in the voxel space region described above, as a result, the voting is performed by searching on the epipolar line in the subject region on the reference image. As a result, a highly accurate voting operation can be performed without depending on the position of the feature point on the reference image and without performing useless search. After the voting for one back projection line is completed, the coordinates of the original feature point in the three-dimensional space are determined by performing a threshold operation on the voxel. Here, the threshold operation is to select a voxel that satisfies the conditions by giving some conditions such as the required minimum number of votes.

According to the present invention, in a device for restoring three-dimensional coordinates of feature points using the voting principle, first, contour information of a voxel space is generated from a contour of a subject on a contour image. Next, a voting operation is performed for each feature point on the reference image using the generated voxel space information. In this case, since only the voxels passing on the back projection line of the feature point are generated, it is not necessary to have all the voxel information, and the storage area for holding the voxel information can be greatly reduced. Finally, the three-dimensional space coordinates of the original feature point are determined by a threshold operation from the voxels for which voting has been performed.

In terms of quality, since the contour of the voxel space generated first is the convex hull area of the subject, it is possible to prevent a search for a useless area outside the subject area during the voting operation. In addition, erroneous voting outside the convex hull region can be suppressed. Also, in the conventional method,
Although it was difficult to add the number of votes from the neighboring images, in the present invention, by extending the projection line from the voxel on the back projection line toward the reference image and comparing the color information, the voting operation of It is possible.

For the above reasons, according to the present invention, it is possible to obtain high-quality and high-precision three-dimensional information extraction results while reducing storage means required for processing.

[0026]

Next, embodiments of the present invention will be described with reference to the drawings.

Referring to FIG. 1, a three-dimensional information extracting apparatus according to an embodiment of the present invention inputs a photographed image, viewpoint information, feature point information, and contour line information, and the entire shape of a subject in the photographed image is determined. A plurality of images captured from each direction that can be restored are selected as contour images, voxel spatial information is generated using contour information of a subject area on the contour image, and the voxel spatial information is stored in the information storage unit 15. The voxel space generating unit 11 performs a predetermined three-dimensional image among the plurality of images as a reference image, and sets a nearby image as a reference image, and sets a feature point on the reference image in a unit three-dimensional space passing through the feature point. A voxel generation unit that generates a voxel in a voxel space and stores color information C of the voxel and original feature points (feature points on a reference image used to generate voxels) in an information storage unit 15. 2, a voxel information comparison unit 13 that extends a projection line from the generated voxel toward each reference image, compares the color information of the intersection with the reference image with C, and stores the result in the information storage unit 15. When the difference between the color information and the color information C is equal to or smaller than the threshold, the three-dimensional information extraction unit 14 extracts the three-dimensional coordinates and color information of the original feature point and stores the extracted three-dimensional coordinates and information in the information storage unit 15.

Here, the terms will be explained.

A photographed image An image of a photographed object.

Viewpoint information The camera position and the direction toward the subject when the subject is photographed. Here, it is used in the same meaning as the camera parameter.

Characteristic point information Information on the edge of a subject or a point in a boundary area on a photographed image.

Contour information Information indicating a boundary between a subject area and a background area on a photographed image. It is represented as a set of points on a line.

Camera parameters The camera position and orientation with respect to the subject when the subject is photographed. However, in this specification, it is assumed that there is no lens distortion of the camera at the time of shooting, and if not, it is necessary to include it in the camera parameters.

In the present embodiment, comparison of color information is considered using RGB notation. In this notation, the color of a voxel of one pixel is represented using three primary colors of R (red) value, G (green) value, and B (blue) value. When comparing the color information of a pixel and a voxel, a difference value is individually obtained for each of the R value, the G value, and the B value, and when all three difference values are smaller than a given threshold, voting for the voxel is performed. I do.

First Embodiment In this embodiment, one set of a reference image and a reference image is selected.
The three-dimensional information viewed from the reference image is extracted.

Next, the flow of the entire processing of this embodiment will be described with reference to the flowchart of FIG.

First, voxel space outline information is generated from the reference outline image and the reference outline image (step 21). FIG.
9 is a flowchart showing a detailed process of generating voxel space outline information. One of the contour images is selected as a reference contour image, and the other images are defined as reference contour images. When the following processing is completed once for all the contour images as the reference contour image, all the processing is completed (steps 40 and 41).

Next, back projection is performed on the pixel f ₁ on the contour line of the subject on the defined reference contour image.

[0039]

[Outside 10] The formulas representing, determined as a straight line passing through the viewpoint I ₁ and f ₁ (step 42-44). If there is an unprocessed reference contour image, select the reference contour image

[0040]

[Outside 11] Viewpoint I ₂ of the reference outline image to calculate the epipolar plane makes with (Step 45-47). Next, the epipolar line

[0041]

[Outside 12] Is obtained as an intersection line between the epipolar plane and the reference contour image (step 48). Next, two intersections

[0042]

[Outside 13] The epipolar line

[0043]

[Outside 14] Are sequentially examined from both ends toward the inside, and are obtained as two intersections of a point where the outline first meets the contour (step 49).
Next, these two points, two straight lines passing through the viewpoint I ₂ and a back projection line

[0044]

[Outside 15] 2 intersections with

[0045]

[Outside 16] (Step 50). For other reference contour images, use the same method

[0046]

[Outside 17] The gradually determined, adding a closest point P ₂ and the farthest point P ₁ when viewed from the reference contour image as part of the outline of the voxel space, will change (step 51).

Next, returning to FIG. 2, if there is an unprocessed feature point on the reference image, a back projection line is drawn from the feature point on the reference image into the voxel space, and all voxels passing therethrough are extracted. Generate the color of the original feature point in each voxel with R value, G value,
The B value is given as three values and stored in the information storage unit 15 (steps 24 and 25).

Next, if there is an unprocessed reference image, the reference image is selected, a projection line is drawn from the center of the generated voxel to each reference image, and the color (R.
G. FIG. If the difference between the (B value) and that of the voxel is equal to or smaller than the threshold value, the voting value 1 is added to the voxel (steps 28 and 29). If there is no unprocessed reference image, the following conditions are satisfied: • One voxel with the maximum number of votes is selected. • The number of votes is greater than the threshold and the voxel that satisfies the maximum is selected. (Steps 26 and 2)
7). If there is no voxel that satisfies the condition, the three-dimensional space coordinates of the feature point are regarded as unrecoverable, and are not added to the voting result. Thereafter, the processing from step 23 is repeated . Second Embodiment In this embodiment, the size of the voxel space is given in advance, and the three-dimensional shape of the entire subject is extracted. The flowchart of FIG. In the present embodiment, the three-dimensional shape of the entire subject is extracted by selecting a plurality of combinations of the reference image and the reference image (steps 31, 32, and 28). Here, as for the reference image, it is necessary to make the entire shape of the subject visible as in the case of selecting the contour image. At this time, for example, as shown in FIG. 10, if an image photographed at a fixed angle using a turntable is used, an image having a known photographing position and orientation can be easily obtained.

Also, since the size of the voxel space is given in advance, in the present embodiment, the spatial information is changed using the principle of the present invention. FIG. 5 shows the flowchart. In this case, back projection line

[0050]

[Outside 18] , And the epipolar line of each reference contour image as in the first embodiment.

[0051]

[Outside 19] Does not need to be explicitly determined from the given region of voxel space.

[0052]

[Outside 20] (Step 5)
2) In that range

[0053]

[Outside 21] Intersection by searching directly above

[0054]

[Outside 22] Can be directly obtained (step 53).

According to the present embodiment, the voting operation can be performed after the voxel space, which had to be specified with a sufficient size in the conventional method, is reduced to the necessary minimum size.

In the present embodiment, a case is described in which a rotary table is used as a method of photographing a subject. However, the present invention is not limited to this, and any photographed image from which camera parameters can be obtained may be used. .

Third Embodiment In the above embodiments, all applicable ranges are not limited to these. For example, a method of generating a voxel space more efficiently by combining the outline information generation of the voxel space of the first embodiment and the flowchart of the voxel space change in the second embodiment may be considered.

In the present embodiment, a voxel space is generated without initial information by applying the voxel space outline information generation processing of the first embodiment to a part of a contour, and the second embodiment is performed on the result. The detailed shape of the outline of the voxel space is determined by applying the processing of changing the outline information of the voxel space. After the outline of the voxel space is generated, three-dimensional information is extracted using the flowchart of FIG. 2 in the first embodiment.

The algorithm for generating a voxel space in this embodiment is as follows.

Determination of the outline shape of the voxel space Four points (the lowermost and right and left four points) from which pixels on the contour line on the reference contour image can form a rectangle surrounding the subject area.
Point), and the intersection of the back projection line for the four points with the back projection line from the reference contour image

[0061]

[Outside 23] (FIG. 8A). As a result, a total of eight intersections are obtained for one reference contour image. The routine for finding this intersection is executed for all the reference contour images, and a rectangular parallelepiped having a plane parallel to the xy, yz, and xz planes including all the obtained intersections is created (FIG. 8B). This is used as the outline shape of the outline of the voxel space.

Using the voxel space determined by the voxel space detailed shape determination as the initial voxel space, the processing of changing the voxel space outline information in the second embodiment is applied as it is to determine the detailed shape of the voxel space.

Referring to FIG. 6, a three-dimensional information extracting apparatus according to another embodiment of the present invention comprises an input device 61, a storage device 62,
63, an output device 64, a recording medium 65, and a data processing device 6
6.

The input device 61 is for inputting photographing information, viewpoint information, feature point information, contour information and the like. The storage device 62 corresponds to the information storage unit 15 in FIG. The storage device 63 is a hard disk. The output device 64 displays or prints the extracted three-dimensional information. The recording medium 65 is a recording medium such as an FD (floppy disk), a CD-ROM, or an MD (magneto-optical disk) on which a three-dimensional information extraction program including the processing described in each of the above embodiments is recorded. The data processing device 66 is a CPU that reads the three-dimensional information extraction program from the recording medium 65 and executes the program.

[0065]

As described above, according to the present invention, there are provided: 1. A convex hull region in which a subject is photographed can be automatically generated as a voxel space without previously specifying a voxel space to be a voting operation target, or the voxel space can be reduced to a convex hull region. High quality and high accuracy 3D information extraction results can be obtained.
The effect is obtained.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a three-dimensional information extraction device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing the overall processing flow of the three-dimensional information extraction device of FIG. 1;

FIG. 3 is a flowchart showing a flow of voxel space generation in the embodiment of FIG.

FIG. 4 is a flowchart illustrating a flow of an entire process according to a second embodiment of the present invention.

FIG. 5 is a flowchart illustrating a flow of voxel space generation according to the second embodiment.

FIG. 6 is a configuration diagram of a three-dimensional information extraction device according to another embodiment of the present invention.

FIG. 7 is an explanatory diagram of the principle of the present invention.

FIG. 8 is an explanatory diagram of the principle of the present invention.

FIG. 9 is an explanatory diagram of the principle of the present invention.

FIG. 10 is a diagram illustrating a state in which a subject is photographed using a turntable.

FIG. 11 is a diagram showing how a voxel range is determined when a voxel space is generated.

FIG. 12 is a conceptual diagram of octree generation.

[Explanation of symbols]

 Reference Signs List 11 voxel space generation unit 12 voxel generation unit 13 voxel information comparison unit 14 three-dimensional information extraction unit 15 information storage unit 21 to 29, 30 to 32, 40 to 53 step 61 input device 62 storage device 63 storage device 64 output device 65 recording Medium 66 Data processing device

Continued on the front page (72) Inventor Akihiko Miyazaki 3-19-2 Nishi-Shinjuku, Shinjuku-ku, Tokyo F-term within Nippon Telegraph and Telephone Corporation (reference) 2F065 AA04 DD00 DD07 FF04 JJ03 JJ26 KK03 MM04 QQ24 QQ25 UU05 5B057 BA02 CE16 CH01 CH11 DB03 DB06 DB09 DC05 DC09 DC16 DC25 5L096 AA09 CA05 FA06 FA15 GA40 LA05

Claims (12)

[Claims] 1. A three-dimensional information extraction method for inputting a plurality of images of the same subject photographed from a plurality of viewpoint directions and extracting information on three-dimensional coordinates and colors of the subject. A voxel space generation step of selecting a plurality of images taken from each direction such that the entire shape of the subject can be restored as a contour image, and generating voxel space information using contour information of a subject area on the contour image, A voxel space storage step of storing the generated voxel space information; and a unit 3 passing a feature point of a predetermined number of images among the plurality of images as a reference image and images near the reference image as reference images.
A voxel generation step of generating a voxel that is a dimensional space in the voxel space, and storing voxels and color information for storing the generated voxels and color information C of feature points on the reference image used for generating the voxels. A voxel information comparison step of extending a projection line from the generated voxel toward each reference image and comparing color information of an intersection with the reference image with the color information C of the feature point. A step of extracting three-dimensional information and extracting three-dimensional coordinates and color information of a feature point on the reference coordinates used to generate the voxel when the information difference is equal to or smaller than the threshold value;
Dimension information extraction method. 2. The step of generating a voxel space includes the steps of selecting a point on the contour of the reference contour image, extending a back projection line from the viewpoint to the point on a three-dimensional space, and selecting all reference contour images. The epipolar plane formed by the back projection line and the viewpoint of the reference contour image is calculated, and the epipolar line of the intersection of the epipolar surface and the reference contour image and the contour line on the reference contour image are calculated. By looking inward from both ends and finding the two intersections, which are the two pixels that first hit the contour, create a three-dimensional convex hull region created by intersecting backprojection lines on all contours The method of claim 1, comprising the step of: 3. The voxel space generation step generates voxel space information representing a schematic shape of a subject from initial voxel space information in addition to the contour line information. In the voxel generation step, the reference image and the reference image are generated. The method according to claim 1, wherein a plurality of sets are used. 4. The voxel space generating step includes selecting four points that can surround the subject area from pixels on the contour line on the reference contour image, and back-projecting the four points from the reference contour image. The method according to claim 1, wherein obtaining an intersection with the back projection line is performed for all the reference contour images, a rectangular parallelepiped including all the obtained intersections is created, and the rectangular parallelepiped is defined as a schematic shape of an outline of the voxel space. Method. 5. A three-dimensional information extracting apparatus for inputting a plurality of images of the same subject photographed from a number of viewpoints and extracting information on three-dimensional coordinates and colors of the subject. A voxel space generating means for selecting a plurality of images photographed from each direction such that the full shape of the object can be restored as a contour image, and generating voxel space information using contour line information of a subject area on the contour image; First information storage means for recording the obtained voxel spatial information, a predetermined number of images of the plurality of images as a reference image, the image in the vicinity thereof as a reference image, the feature points on the reference image, A voxel generating means for generating a voxel, which is a unit three-dimensional space passing through a feature point, in the voxel space, and the generated voxel and a voxel space used for generating the voxel space A second information storage unit for storing color information C of a feature point on the reference image, extending a projection line from the generated voxel toward each reference image, and obtaining color information of an intersection with the reference image and the feature point. And voxel information storage means for storing the result in a second information storage means; and a reference used for generating a voxel when a difference between the two color information is equal to or less than a threshold value. A three-dimensional information extraction device, comprising: means for extracting three-dimensional coordinates and color information of a feature point on coordinates and storing the information in the first information storage means. 6. The voxel space generating means includes means for selecting a point on the contour line of the reference contour image, extending a back projection line from the viewpoint to the point on the three-dimensional space, and selecting all reference contour images. , The epipolar plane formed by the back projection line and the viewpoint of the reference contour image is calculated, and the epipolar line of the intersection of the epipolar surface and the reference contour image and the contour line on the reference contour image are calculated. By looking at the line inward from both ends and finding the two intersections, which are the two pixels that first hit the contour, the three-dimensional convex hull region created by intersecting the backprojection lines on all the contours The apparatus of claim 5, comprising means for creating. 7. The voxel space generating means generates voxel space information representing a schematic shape of a subject from initial voxel space information in addition to the contour line information. 6. The apparatus according to claim 5, wherein a plurality of sets are used. 8. The voxel space generation means selects four points that can surround the subject area from pixels on the contour line on the reference contour image, and performs back projection lines for the four points from the reference contour image. The method according to claim 5, wherein obtaining an intersection with the back projection line is performed for all the reference contour images, a rectangular parallelepiped including all the obtained intersections is created, and the rectangular parallelepiped is defined as the outline shape of the outline of the voxel space. apparatus. 9. A three-dimensional information extraction program for inputting a plurality of images obtained by photographing the same subject from a number of viewpoints and extracting information on three-dimensional coordinates and colors of the subject, the program comprising: A voxel space generation process of selecting a plurality of images taken from each direction such that the full shape of the object can be restored as a contour image, and generating voxel space information using contour information of a subject area on the contour image; A first information storage process of recording the obtained voxel spatial information, a predetermined number of images of the plurality of images as a reference image, and images in the vicinity thereof as reference images, and for feature points on the reference image, A voxel generation process for generating a voxel, which is a unit three-dimensional space passing through a feature point, in a voting target space, and is used for generating the generated voxel and the voxel. A second information storage process for storing color information C of a feature point on a reference image, extending a projection line from the generated voxel toward each reference image, and obtaining color information of an intersection with a reference image and the feature Voxel information storage processing for comparing the color information C of the point with the result and storing the result in the second information storage means; and when the difference between the two pieces of color information is equal to or less than a threshold, the voxel information is used to generate a voxel. A recording medium storing a three-dimensional information extraction program for extracting a three-dimensional coordinate and color information of a feature point on reference coordinates and causing a computer to execute a process of storing the information in the first information storage means. 10. The voxel space generation process includes a process of selecting a point on the contour line of the reference contour image, a process of extending a back projection line from the viewpoint to the point on a three-dimensional space,
For all the reference contour images, an epipolar plane formed by the back projection line and the viewpoint of the reference contour image is calculated, and an epipolar line which is an intersection of the epipolar surface and the reference contour image, and a contour line on the reference contour image are calculated. Looking inward from both ends of the epipolar line, and finding the two intersections that are the two pixels that first come into contact with the contour, a three-dimensional projection created by intersecting the backprojected lines on all the contours The recording medium according to claim 9, further comprising a process of creating a wrapping area. 11. In the voxel space generation processing, voxel space information representing a schematic shape of a subject is generated from initial voxel space information in addition to the contour line information, and the voxel space information is generated. The recording medium according to claim 9, wherein a plurality of sets of the reference image and the reference image are used. 12. The voxel space generation processing selects four points that can surround a subject area from pixels on the contour line on the reference contour image, and performs back projection lines for the four points from the reference contour image. The method according to claim 9, wherein obtaining an intersection with the back projection line is performed for all reference contour images, a rectangular parallelepiped including all the obtained intersections is created, and the rectangular parallelepiped is defined as a schematic shape of an outline of the voxel space. recoding media.