JPH1091786A - Three-dimensional image information extraction method and image forming method using the extraction method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To extract the three-dimensional image information from plural two-dimensional images by producing an edge map where the dots gather, erasing the noises, producing a vectorized line image, and specifying a three-dimensional coordinate in regard to a vectorized line to extract the three-dimensional image information. SOLUTION: An edge map where the dots gather is produced, the noises are erased, and a line image ia produced at a border edge between the luminance and the hue. Then a three-dimensional coordinate is decided by the vector correspondence processing in regard to a vectorized line. When the correspondence of segments are secured among plural coordinates, the three- dimensional coordinate is decided from these segments in a triangulation principle. Furthermore, an angle division distribution pattern is produced against a line image. i.e., a two-dimensional vector. Then the depth data are extracted from the distributed angles in a vector and the three-dimensional coordinates of all lines are decided.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for extracting three-dimensional image information from a plurality of two-dimensional images having parallax and a technique effective when applied to a method for creating a three-dimensional image using the same. is there.

[0002]

2. Description of the Related Art Conventionally, a three-dimensional image obtained by a three-dimensional measurement system has properties that are not found in ordinary light and dark images, and is an effective approach for robot vision such as recognition of the position and shape of an object. is there. However, even if it is a three-dimensional image, the structure of the obtained image data is different if the measuring means is different. An image processing algorithm appropriate for the input type must be applied. For example, a slit image obtained by the slit light projection method and a distance image obtained by the optical radar method are heterogeneous data,
From an image processing standpoint, it requires a completely different approach. Therefore, in a specific application, attention must be paid to the selection of a measurement method that meets the purpose and the application of an image processing algorithm suitable for each.

[0003] The technology relating to the three-dimensional image processing, in particular,
Some algorithms disclosed so far are described in, for example, Shokodo Inc., May 10, 1996.
It is described in Seiji Iguchi and Kosuke Sato, published in the 5th edition of the first edition, "3D Image Measurement", pp. 121-141.

[0004]

The present inventor has found the following problems as a result of studying the above prior art.

In the prior art, since the outline of an image of an object is recognized only by luminance, an object having a simple shape can be recognized.
Recognition of contours of homogeneous objects, objects of the same color, etc. was obtained with high accuracy.However, objects of complex shapes, objects of heterogeneous objects, objects of multicolored colors, various objects in nature, etc. However, there has been a problem that sufficient contour recognition accuracy cannot be obtained.

An object of the present invention is to provide a method for extracting three-dimensional image information from a plurality of two-dimensional images.

Another object of the present invention is to provide a technique capable of obtaining a highly accurate three-dimensional image from a plurality of two-dimensional images.

Another object of the present invention is to provide a technique capable of obtaining sufficient contour recognition accuracy even for various objects in the natural world.

The above and other objects and novel features of the present invention will become apparent from the description of the present specification and the accompanying drawings.

[0010]

SUMMARY OF THE INVENTION Among the inventions disclosed in the present application, the outline of a representative one will be briefly described.
It is as follows.

(1) A first step of loading a plurality of two-dimensional images having parallax into a memory, a second step of correcting the plurality of two-dimensional images, and a line drawing of each corrected two-dimensional image. A third process of creating an edge map in which dots are collected, a fifth process of creating an edge map in which dots are collected, a fifth process of creating a vectorized line image by removing noise, and a vector. A three-dimensional image information extracting method including a sixth step of extracting three-dimensional image information by specifying three-dimensional coordinates with respect to a transformed line.

(2) A first step of loading a plurality of two-dimensional images having parallax into a memory, a second step of correcting the plurality of two-dimensional images, and a line drawing of each corrected two-dimensional image. A third process of creating an edge map in which dots are collected, a fifth process of creating an edge map in which dots are collected, a fifth process of creating a vectorized line image by removing noise, and a vector. A sixth process of extracting three-dimensional image information by specifying three-dimensional coordinates with respect to a transformed line, and terming a region of a minimum unit surrounded by a continuous line as an “region”, and decomposing the image into a plurality of regions Clarify the seventh process and the relationship between the line and the surface, and rank them.
The process of assuming undetermined boundaries and surfaces in the hidden parts,
A three-dimensional image creating method includes an eighth step of specifying three-dimensional coordinates inside a region surrounded by a line and a ninth step of determining a part whose three-dimensional coordinates have not yet been determined by interpolation. .

(3) A first step of loading a plurality of two-dimensional images having parallax into a memory, a second step of correcting the plurality of two-dimensional images, and a line drawing of each corrected two-dimensional image. A third process of creating an edge map in which dots are collected, a fifth process of creating an edge map in which dots are collected, a fifth process of creating a vectorized line image by removing noise, and a vector. A sixth process of extracting three-dimensional image information by specifying three-dimensional coordinates with respect to a transformed line, and terming a region of a minimum unit surrounded by a continuous line as an “region”, and decomposing the image into a plurality of regions Clarify the seventh process and the relationship between the line and the surface, and rank them.
The process of assuming undetermined boundaries and surfaces in the hidden parts,
Eighth process for specifying three-dimensional coordinates inside a region surrounded by a line, ninth process for determining a part for which three-dimensional coordinates have not yet been determined by interpolation, and the earth as a region or a set of regions. A tenth process of separating from the others as (group), an eleventh process of separating the background from the others as a region or a set of regions (group), a set of regions of the background, the earth, and the object and no object This is a three-dimensional image creation method including a twelfth step of forming a three-dimensional image from space.

(4) A first step of loading a plurality of two-dimensional images having parallax into a memory, a second step of correcting the plurality of two-dimensional images, and a line drawing of each corrected two-dimensional image. A third process of creating an edge map in which dots are collected, a fifth process of creating an edge map in which dots are collected, a fifth process of creating a vectorized line image by removing noise, and a vector. A sixth process of extracting three-dimensional image information by specifying three-dimensional coordinates with respect to a transformed line, and terming a region of a minimum unit surrounded by a continuous line as an “region”, and decomposing the image into a plurality of regions Clarify the seventh process and the relationship between the line and the surface, and rank them.
The process of assuming undetermined boundaries and surfaces in the hidden parts,
Eighth process for specifying three-dimensional coordinates inside a region surrounded by a line, ninth process for determining a part for which three-dimensional coordinates have not yet been determined by interpolation, and the earth as a region or a set of regions. A tenth process of separating from the others as (group), an eleventh process of separating the background from the others as a region or a set of regions (group), a set of regions of the background, the earth, and the object and no object A twelfth process of forming a three-dimensional image from space and a thirteenth process of creating a work plane suitable for image processing
And a fourteenth step of performing predetermined image processing as necessary.

According to the above-described means, three-dimensional image information can be extracted from a plurality of two-dimensional images by creating a line image of a two-dimensional image by luminance edge processing and hue edge processing. As a result, sufficient contour recognition accuracy can be obtained even for an object having a complicated shape, an object made of a foreign material, an object made of multicolored colors, and various objects in the natural world.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail along with embodiments (examples).

The three-dimensional image information extraction method and the three-dimensional image creation method using the same according to the embodiment (example) of the present invention are described below.
It consists of a first stage, a second stage, and a third stage.

(First Stage) Processing Step (1) Two or more two-dimensional images having parallax by a plurality of video cameras are loaded into a memory. -The optimal number of images (number of cameras) is determined by experiment. A known distance between cameras for obtaining two or more two-dimensional images having the parallax is used. • Test three cameras as a set to get the final image with smooth edges. The image of the camera at the center is used as the main image, and the cameras on both sides are used as auxiliary images for obtaining a parallax image and correcting the boundary of the cutout surface. For example, in the case of a vehicle, two cameras are sufficient. At that time, the main image is switched according to the moving direction of the camera. -The selection of the type of camera is performed by AD conversion from a digital still camera, a digital TV camera, or a film camera. Digital still cameras are advantageous in image quality,
In moving image processing, the storage medium becomes enormous and becomes impractical. Therefore, a digital TV camera that can use a video tape as a storage medium is used.

Processing Step (2) The plurality of two-dimensional images are corrected. That is, first, optical correction such as correction of spherical aberration and the like is performed. Correct the perspective so that it is correctly orthogonal (square) processed.

Next, the vanishing point (the focus of the perspective) is corrected as required. Create a correction function or correction table.
There is no breakthrough because it uses known technology.

(Second Stage) Processing Step (3) A line drawing of each of the two-dimensional images is created (at this stage, it is still a two-dimensional image).

An image consisting only of a luminance signal is created by edge processing of luminance (BRIGHTNESS). For example, a spatial differentiating circuit is formed by a digital filter, and an edge signal is extracted to create an edge map. A known digital two-dimensional filter for edge detection is used.

Next, hue (HUE) edge processing is performed. In this method, saturation (CHROMA) is saturated to form an image consisting only of hue signals, and hue edges are detected (a digital filter is also used here). The hue signal is characterized in that it depends only on the light wavelength reflection characteristics of the object and the color temperature of the light source, and does not depend on the intensity of the light source or the illuminance of the object.
Thus, the hue signal does not depend on the light intensity, but only on the color temperature of the light source and thus the optical properties of the surface of the object.

Due to the nature of the object, regardless of the luminance or illuminance (whether it is in the sun or in the shade), it can be separated for each object quality by the hue signal. Good agreement.

That is, a distribution characteristic diagram of hue signals in an image is created, a hue map is created, and regions having the same hue and the region boundaries are determined. For example, color signals X, Y,
A hue signal is directly generated from the Z signal, or the phase of a subcarrier is detected, and a phase distribution diagram (hue map) is generated.

A fragmentary (discrete) phase boundary is set as a new edge. When the quality of the object changes, the luminance and the hue change in a fragmentary manner (discretely). Therefore, the change point of the luminance and the hue includes the boundary of the object, and the boundary of the object very often changes the luminance and the hue. .

However, the hue edge is not always the boundary of the object. However, the object boundary always includes a hue edge and a luminance edge. A 99% object boundary can be detected by the luminance edge and the hue edge. Edges that cannot be detected by the above-described method will be points that cannot be detected visually.

Processing Step (4) Next, an edge map in which dots are gathered is created, noise is eliminated, vectorized, and a line image is created. That is, a line drawing is formed by edges at the boundary between luminance and hue. Edge maps are only needed to determine the extent of processing,
It may be slightly larger than the target object. The part of the image that is cut off by the edge and escapes from the image is captured from the adjacent image, so there is no worry that the area created by the edge is too large. -It is divided into a closed line segment that forms a surface (region boundary edge) and a single line segment that does not form a surface (independent edge).・ A single picture element dot that does not create inconsistent data or edges is regarded as noise. Establish noise standards. -When three-dimensional coordinates are overlapped, such as a distant view reflected on the water surface, an image reflected on a mirror, or an object reflected on a metallic glossy surface, special care must be taken as a mirror effect for processing. Care should be taken not to judge the mirror effect as contradiction or noise. An edge is formed from an edge map created from edge processing of luminance (BRIGHTNESS) and edge processing of hue (HUE), and a group of dots is vectorized and converted into a line segment. Combine vectorized line segment maps created by luminance and hue to complete line drawings of a plurality of two-dimensional images. As a result, a part where the quality of the surface of the object changes, a region of the same quality, and a plurality of two-dimensional images divided into the region boundary are formed.

Processing Step (5) Next, three-dimensional coordinates are specified for the vectorized line. There are the following three methods for determining three-dimensional coordinates of a line segment from a plurality of images formed by the line segment converted into a two-dimensional vector. The accuracy can be improved by using two of these methods in combination.

In the method (1), three-dimensional coordinates are specified by "vector correspondence processing". If a correspondence between each line segment is provided between a plurality of images, three-dimensional coordinates can be determined from the line segment by the principle of triangulation (trigonometric function).

First, finding a corresponding point of a line segment between a plurality of images and calculating three-dimensional coordinates is called "vector correspondence processing". In order to make each line segment correspond to a plurality of images, it is easier to determine the area after determining the area. It is quite possible to classify line segments and predict corresponding point locations and vectors.

A corresponding vector is determined by selecting from candidate vectors (at most several). By narrowing down the range where the corresponding points exist, the corresponding points are finally narrowed down to one.

In the method (2), the three-dimensional coordinates of the line segment are specified by “division angle distribution processing”. The original three-dimensional coordinates of the line part (two-dimensional vector) of the line image of the main image are extracted from the line image of the other image by the angle division processing to extract the three-dimensional vector.

The feature of the angle-of-segment distribution processing is that three-dimensional coordinates can be obtained even if the correspondence between vectors is not established. Split-angle distribution processing is a three-dimensional coordinate determination method that creates a "square-angle distribution pattern" for a line drawing (two-dimensional vector), extracts depth data as a vector from the distributed angle, and determines the three-dimensional coordinates of all lines. (At this stage, it becomes a three-dimensional line drawing for the first time).

Line drawings (two-dimensional vectors) are "0" and "1".
Therefore, both the vector correspondence processing and the angle division distribution processing are easy. Lines in a line drawing are classified into discontinuous independent line segments and region boundary line segments that are continuously closed to form a surface.

The method (3) is performed by direct processing using two-dimensional correlation. Correlate a plurality of images to determine three-dimensional coordinates. This method can be applied to a part of the screen.

Processing Step (6) Next, the surface of the minimum unit surrounded by the continuous lines is named "region", and the image is decomposed into a plurality of regions. From the three-dimensional coordinates of the obtained line drawing, a surface boundary line segment is found, and the continuous line is traced to specify a “surface” (curved surface at this stage) of a region surrounded by the line. It is possible to divide the specified plane into a plurality of planes and to vectorize the plane by its normal direction and area. Here, the image has been aggregated into a single basic image composed of regions, that is, a line vector and a plane vector.

It should be noted that it is convenient to think primarily of an image taken by a single camera, assuming that the apparent two-dimensional image has three-dimensional information including a shadowed part.

Processing Step (7) Next, the relationship between the line and the surface is clarified, the order is determined, and an undetermined boundary and surface are assumed in the hidden part.

The identified and named areas;
The boundary of another area hidden behind the area is corrected by parallax from an image of an adjacent camera, and information is added. Although the additional signal in that case is still two-dimensional, the three-dimensional coordinate is predicted and added as a smooth extension from the known three-dimensional coordinate of the same area.

Still, it is impossible to obtain all the information of the hidden part, so the boundary of the hidden part remains unknown. The prediction of the three-dimensional coordinates of the hidden part is performed until the hidden edge appears. Don't expect beyond the edge. By adding information as three-dimensional coordinates, a smooth image boundary is realized.

Further, if data of a hidden part is required, the number of cameras is increased. If you increase it vertically,
It is possible to obtain an overhead view.

Therefore, the edge of the surface as far as possible is positioned as a temporary boundary. Temporary boundaries are treated differently from other boundaries that have already been determined, giving them the flexibility to add and change information at any time. The part cut by the frame of the image is also a temporary boundary. From the three-dimensional coordinates, the relationship between the line and the surface surrounding the line is clarified, and the arrangement order from the viewpoint direction is assigned (the arrangement order of the regions).

Processing Step (8) Next, the three-dimensional coordinates inside the area surrounded by the line are specified (up to the stage of the three-dimensional drawing). Returning to the original image to determine the three-dimensional coordinates in the area surrounded by the already determined line drawing, it is not the edge of the already determined “0” and “1”.
A vector correspondence process and an angle distribution process are performed on the remaining analog, continuously changing luminance gradient, saturation gradient, and hue gradient. The edge is the maximum frequency F of the system
(Max) (processed).

After the edge is processed as the contour of the object, the frequency component F (h) that has not been determined to be an edge is sequentially subjected to a vector distribution process (and a vector correspondence process) to extract each three-dimensional vector. And fill the surface.

The planes of all the regions are taken out one by one, enlarged if necessary for each region, and three-dimensional coordinates are determined in detail. If your computer has the power,
The processing of the step (4) of “creating an edge map in which dots are gathered, eliminating noise and vectorizing to form a line drawing” and the processing step (8) of “3D coordinates inside the area surrounded by the line” The process of “specify” can be performed as one process, or the entire area can be processed at once. As a result, further detailed three-dimensional coordinates inside the boundary of the region whose three-dimensional coordinates have been determined by the edge processing are determined.

Since the edge of the surface in the area always coincides with the boundary line formed by the edge (the property of a real object), as a result, the coordinates of the line of the line drawing are changed with increased accuracy.
In this processing, although a two-dimensional image is apparent, many points in the screen have three-dimensional coordinates.

(1) Identification of Three-Dimensional Coordinates on Plane (Completion of Plan in Area) The line-processed image is a binary image of “0” and “1”, and the three-dimensional coordinates are determined. Although it will be easy, the in-plane of the area is not a line segment of “0” and “1” but an analog quantity of a waveform that changes continuously, and requires intermediate treatment between edge processing and correlation processing. Perform vector correspondence processing and angle division distribution processing according to the luminance waveform and the hue waveform. -Whether it is a dot or a line or a waveform, it can be said that it is a two-dimensional specifying process. By developing it, one region is collectively specified as three-dimensional coordinates as a "plane". If a corresponding plane in each image of the plane vectorized by the normal direction and the area is obtained, the vector correspondence processing can be performed. Areas where the correspondence of the planes is not required are subjected to angle distribution processing.

(2) The three-dimensional coordinates of the detailed plane are determined by the orthogonal function conversion processing. Find the optimal orthogonal function in relation to the coordinates of the ground and the shooting axis. By the modified Fourier transform, the traveling direction of the photographer is X
A power spectrum is created with the axis as the axis, the height as the Y axis, and the depth as the Z axis, and the movement amount and the Z axis are detected by focusing on the phase velocity of the frequency component. The movement of an object due to parallax can be expected to depend on the phase of the spectrum created by the object. -In addition to this processing, it is also possible in principle to perform orthogonal function conversion directly from a moving image (a continuously changing image) itself to determine three-dimensional coordinates (see the principle of X-ray CT).

Processing Step (9) The part for which the three-dimensional coordinates have not yet been determined is determined by interpolation. Although the three-dimensional coordinates of the surface created by the region have already been obtained, the process here is a process for further determining internal details.

Although the three-dimensional coordinates of a considerable portion are determined by F (h), the portion where the luminance and hue do not change (DC component)
In principle, three-dimensional coordinates cannot be determined by either the vector correspondence processing or the angle distribution processing. The DC portion is approximately obtained from surrounding coordinates by interpolation (not an unreasonable approximation). By this processing, the three-dimensional coordinates of all parts in the image have been specified.

By the above processing, all parts in the image are constituted by the three-dimensional vectorized line and the three-dimensional vectorized surface.

Finally, the edge is corrected with a supplementary image from an adjacent image in order to eliminate the glare of the edge.
The detected edges do not appear directly in the image.

(Third Step) Processing Step (10) The earth is separated from the others as areas or groups (sets of areas). At this stage, the earth is composed of multiple areas. Due to the nature of objects on the ground, most objects are in contact with the earth due to gravity, so the area is in contact with the earth, but for the sake of image processing, the area is separated from the earth and separated into one area .

Processing Step (11) The background is separated from the others as regions or groups (sets of regions). A distant view that is more than a certain degree is treated as a series of curved surfaces as a background. At this stage, the background consists of a group of sets of a plurality of regions.

Processing Step (12) Completion of a three-dimensional image composed of a background, the ground, a group of objects, and a space without objects.

An area in which detailed three-dimensional coordinates are determined generally exists as a set of a plurality of areas. (A) A group is formed by collecting a plurality of three-dimensionally adjacent regions. (B) A plurality of regions that are close to each other are collected to form a group. (C) The group of the background, the ground, and the object always has a surface of a part that cannot be seen from the main image. (D) Since the image includes a part that cannot be seen from the main image, the edge is finely processed (to eliminate glare and unnaturalness such as chroma key).

Thus, a three-dimensional image as a set of groups is completed. The image is completed by attaching the luminance, hue, and luminance (chroma) from the plurality of images to the three-dimensional coordinates.

It should be noted that a group of the pasted images forms one buffer memory plane. In order to reconstruct the region and the group, a buffer memory during this processing is prepared.

Processing Step (13) A work plane suitable for image processing is created. Road landscape, landscape with sea, mountain landscape, city landscape, landscape in architecture,
Night scenery, moving scenery, winter scenery and other scenery are judged from the color composition, parameters suitable for each are prepared, the meaning of all objects in the scenery is judged,
Classify.

The meaning of an object is selected, classified, judged, and named by a "three-dimensional digital filter" and a "three-dimensional mask" having parametric variables.
Rather than integrating only physically close coordinates, each group is separated or combined and reconstructed for each meaningful part so as to correspond to each object. Or,
Return to the domain and separate and rejoin to form a meaningful set.

A set of areas collected for each meaningful part is defined as a work group. Features of the work group: When an object moves, only the work group relating to the object moves while changing its shape, and the other work groups do not change.

This group of operations is often re-changed from the buffer memory according to the purpose and other needs.

After the scenery is selected, one background, one earth,
Each work group of a plurality of objects having various meanings is configured.

At the stage of classifying and reconstructing into groups, a part that does not fall into any region is a space where no object exists, and this space is a space that may be used in a future image processing stage.

Therefore, it is desirable that this space is positioned as one work group and has three-dimensional coordinates. Each task group has its own memory plane, is independent and can work independently. Thus, the basic three-dimensional image is completed.

Processing Step (14) Here, for example, the following image processing is performed. -Coordinate transformation of viewpoint change, prediction of an image of a blind spot, separation or combination with an adjacent work group. -The point that the actual captured image captured by the camera is converted into three-dimensional images and replaced with CG to construct a three-dimensional whole image or a partial image is an important step for realizing artificial vision. It is the main feature.・ Three-dimensional normalization ......... Three-dimensional images with three-dimensional coordinates are
Since the size on the coordinates has already been determined, it is possible to directly standardize as it is by a process of justifying the directions. The standardized work group (area is also possible) can be easily compared with other work groups, measurement, determination for each purpose, mask processing, filter processing, and the like can be easily performed.・ Average averaging of objects: Since the size of a two-dimensional image differs depending on the position of the viewpoint, the averaging process cannot be performed as it is. Since the averaging can be easily performed even if the apparent sizes are different, it is expected that the S / N of the image and the measurement accuracy of the object are greatly improved. -Transmission by each work group (or area) makes it possible to realize an optimum transmission frequency for each work group, so that extremely large bandwidth compression can be realized. The images are transmitted separately for each work group, and the receiving side reconstructs a three-dimensional image. Bandwidth compression by transmission for each work group is a main feature of the present invention. -By performing the processing for each work group in this way, extremely efficient processing can be realized.

Processing step (15) Free reconstruction of a three-dimensional image is performed. A two-dimensional image from any viewpoint can be obtained from the square-transformed three-dimensional coordinates at any time.・ If you change the viewpoint continuously, you can create a video that moves in any direction at any time. -If a vertical image is captured, a bird's-eye view is also possible. The image is reconstructed without contradiction by overwriting from the far side, that is, in the reverse order of the arrangement order of the regions or the arrangement order of the groups.

Processing Step (16) The processing steps (3) to (9) are implemented as an LSI chip. By making the processing steps (3) to (9) into chips, handling of a three-dimensional image becomes easy, and the application range is rapidly expanded. -Various applications can be opened by forming chips in the processing steps (10) to (13).・ Error in general triangulation is 1 for 100 meters
Although it has an accuracy of millimeters or less, that is, an accuracy of 0.001% or more, the accuracy of triangulation with an image to be realized now is:
An accuracy of about 5 cm at 100 meters, that is, about 0.5% is sufficient.・ Understand the optical properties of the camera and correctly calibrate the square conversion to the image on the film surface.

The present invention has been described with reference to the above embodiment (example).
However, the present invention is not limited to the above-described embodiment, and it is needless to say that various changes can be made without departing from the scope of the present invention.

[0071]

The effects obtained by typical aspects of the invention disclosed in the present application will be briefly described as follows.

By creating a line image of a two-dimensional image by luminance edge processing and hue edge processing, three-dimensional image information can be extracted from a plurality of two-dimensional images. As a result, objects with complex shapes, objects made of foreign materials,
Sufficient three-dimensional recognition accuracy of contours and surfaces can be obtained even for objects composed of multiple colors, various objects in the natural world, and the like.

Claims (4)

[Claims] A first step of loading a plurality of two-dimensional images having parallax into a memory; a second step of correcting the plurality of two-dimensional images; A third process of creating an edge map in which dots are collected, a fourth process of creating an edge map in which dots are collected, a fifth process of creating a vectorized line image by eliminating noise, and a vectorization process Sixth extracting three-dimensional image information by specifying three-dimensional coordinates with respect to the extracted line
A three-dimensional image information extracting method, comprising the steps of: 2. A first step of loading a plurality of two-dimensional images having parallax into a memory; a second step of correcting the plurality of two-dimensional images; A third process of creating an edge map in which dots are collected, a fourth process of creating an edge map in which dots are collected, a fifth process of creating a vectorized line image by eliminating noise, and a vectorization process Sixth extracting three-dimensional image information by specifying three-dimensional coordinates with respect to the extracted line
, The surface of the smallest unit surrounded by a continuous line is named “region”, the seventh process of decomposing the image into a plurality of regions, and the order of the line and the surface is clarified and ranked. A process of assuming an undetermined boundary and surface in a hidden portion, an eighth process of specifying three-dimensional coordinates inside a region surrounded by a line, and an interpolation method of a portion in which three-dimensional coordinates have not yet been determined A three-dimensional image creating method, comprising: a ninth step of determining a three-dimensional image. 3. A first step of loading a plurality of two-dimensional images having parallax into a memory, a second step of correcting the plurality of two-dimensional images, and a step of converting a line drawing of each of the corrected two-dimensional images into a luminance. A third process of creating an edge map in which dots are collected, a fourth process of creating an edge map in which dots are collected, a fifth process of creating a vectorized line image by eliminating noise, and a vectorization process Sixth extracting three-dimensional image information by specifying three-dimensional coordinates with respect to the extracted line
, The surface of the smallest unit surrounded by a continuous line is named “region”, the seventh process of decomposing the image into a plurality of regions, and the order of the line and the surface is clarified and ranked. A process of assuming an undetermined boundary and surface in a hidden portion, an eighth process of specifying three-dimensional coordinates inside a region surrounded by a line, and an interpolation method of a portion in which three-dimensional coordinates have not yet been determined A ninth step of determining the ground, a tenth step of separating the ground from the others as a region or a set of regions, and an eleventh process of separating the background from the others as a region or a set of regions A three-dimensional image creating method, comprising: a twelfth step of constructing a three-dimensional image from a process, a set of regions of the background, the earth, and the object and a space without the object. 4. A first step of loading a plurality of two-dimensional images having parallax into a memory, a second step of correcting the plurality of two-dimensional images, and a step of converting a line drawing of each of the corrected two-dimensional images into a luminance. A third process of creating an edge map in which dots are collected, a fourth process of creating an edge map in which dots are collected, a fifth process of creating a vectorized line image by eliminating noise, and a vectorization process Sixth extracting three-dimensional image information by specifying three-dimensional coordinates with respect to the extracted line
, The surface of the smallest unit surrounded by a continuous line is named “region”, the seventh process of decomposing the image into a plurality of regions, and the order of the line and the surface is clarified and ranked. A process of assuming an undetermined boundary and surface in a hidden portion, an eighth process of specifying three-dimensional coordinates inside a region surrounded by a line, and an interpolation method of a portion in which three-dimensional coordinates have not yet been determined A ninth step of determining the ground, a tenth step of separating the ground from the others as a region or a set of regions, and an eleventh process of separating the background from the others as a region or a set of regions A process, a twelfth process of forming a three-dimensional image from a set of regions of the background, the earth, and the object and a space without the object, and a thirteenth process of creating a work plane suitable for image processing. A fourteenth step of performing predetermined image processing in response to the request. Image creation method to.