KR20100009452A - Method for image processing - Google Patents

Abstract

PURPOSE: An image processing method is provided to decide the ground line which passes the ground and sky or contacts the building and ground. CONSTITUTION: A first image is produced by perform the slope correction about an input color image(120). The local horizon of the first image pitches. One or more burning-off point of the first image is detected from the local horizon(130). A plurality of ground line candidates of the first image pitches based on burning-off points(140). The ground line of the input color images is detected based on the texture change of the texture change in the y-direction and the color variation in y-direction(150).

Description

[0001] The present invention relates to a method for image processing,

Some embodiments of the present invention relate to an image processing method, and more particularly, to a method for determining a ground line as a boundary between a building area and a ground area from an image, and a method for modeling a 3D image based on the ground line will be.

Recently, a method of modeling a 3-dimensional image from a color image has been studied. Generally, 3D information for 3D image modeling is composed of geometry information and color information. The color information is obtained by using a color image acquired by a video acquisition device such as a CCD (Charged Coupled Device) or a CMOS (Complementary Metal-Oxide-Semiconductor) camera.

On the other hand, the shape information can be obtained using a depth image. Examples of the method for acquiring the depth image include a method of directly acquiring the depth image through a hardware device such as a depth camera, There is a method of acquiring indirectly through a software image processing called a technique.

Also, a method of acquiring a depth image through the computer vision technique includes a method of using a plurality of input color images and a method of using only one input color image.

Among them, the study on the method of using one input color image is divided into two major directions. One is a method of dividing and processing a color input image into a ground, a vertical structure (a building or the like) (Vertical or Building), and a sky area. And the other is a method of training the input color image using the learning result by training the correspondence relationship between the color image and the corresponding depth image.

In the case of the method using the image segmentation, the preprocessing process for image segmentation is relatively complicated, and merely the geometry information of each divided region is not utilized. In the case of the learning method, the result of the image processing is greatly influenced by the accumulated training result (Training Set).

Therefore, it is possible to effectively and accurately detect the ground line, which is the boundary line between the vertical structure such as a building and the ground, or the boundary between the sky and the ground, without using the image segmentation or learning, You will be able to improve the quality.

Some embodiments of the present invention are directed to improving the efficiency of 3D image modeling and provide a method for determining a ground line where a vertical structure (building) and a ground meet from an input image or a sky and a ground meet.

Another embodiment of the present invention is to provide a method of 3D modeling an input 2D color image using the ground line.

According to an aspect of the present invention, there is provided a method of generating an image, the method comprising: generating a first image by performing skew correction on an input color image; determining a plurality of ground line candidates from the first image; Detecting a ground line of the input color image based on at least one of a color change in the y-axis direction of the ground line candidate and a texture change in the y-axis direction.

According to another aspect of the present invention, the step of determining the plurality of ground line candidates comprises the steps of determining a horizon of the first image, detecting at least one vanishing point of the first image from the horizon, And determining a plurality of ground line candidates of the first image based on the vanishing point.

According to another aspect of the present invention, the step of determining the horizon of the first image includes extracting at least one horizontal straight line from the first image, and extracting at least one horizontal straight line from the first image, And determining a horizon of the first image based on the coordinate value.

According to another aspect of the present invention, the vanishing point is extracted based on the x-axis coordinate value of the intersection of the at least one horizontal straight line.

According to another aspect of the present invention, the step of determining a plurality of ground line candidates of the first image based on the vanishing point comprises: Determining a boundary line of the group as a vertical boundary line of the first image, and determining a plurality of ground line candidates composed of a line segment having an edge at a vertical boundary line of the first image.

According to another aspect of the present invention, the step of determining a plurality of ground line candidates of the first image may include: determining a plurality of ground line candidates of the first image based on a color value change in the x- Determining a vertical boundary, and determining a plurality of ground line candidates comprising a line segment having an edge at a vertical boundary of the first image.

According to another aspect of the present invention, in the first image, a region of the ground line above the ground line is projected in a direction perpendicular to the ground, and the area of the ground line is projected in a direction parallel to the ground, And performing 3D modeling of the first image.

According to another aspect of the present invention, there is provided a method of generating an image, comprising the steps of: extracting at least one horizontal straight line from an input image; determining a horizon of the input image based on a y coordinate value of an intersection of the at least one horizontal straight line; And detecting at least one vanishing point from the horizon based on an x-coordinate value of an intersection of the at least one horizontal straight line.

According to some embodiments of the present invention, the efficiency of 3D image modeling is improved.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings and accompanying drawings, but the present invention is not limited to or limited by the embodiments. Like reference symbols in the drawings denote like elements.

1 is a flowchart showing an image processing method according to an embodiment of the present invention.

In step 110, an image is input. The image is a color image file composed of RGB data. The file format of the input image may be various forms such as a bitmap (bmp), jpeg or jpg, and a raw file. If the file format is data compression, the decompression may be performed prior to the image processing.

According to another embodiment of the present invention, the accuracy and / or efficiency of image processing is enhanced through various pre-processing of the input image. For example, the adjusted result image may be used as an input image of the image processing by adjusting brightness or value, saturation, etc. of the input image. In addition, various filter mask processing may be included in the pre-processing.

In the input image, there is a part which is a vertical structure as a building part. In the 3D image modeling process, the part which is a vertical structure will be erected in a direction perpendicular to the ground. Meanwhile, a ground portion exists in the input image, and the ground portion must be modeled in a direction parallel to the ground in the 3D image modeling process. On the other hand, the ground line is the boundary between the building portion and the ground portion of the input color image.

An input color image obtained at an arbitrary time point is an image that has not undergone tilt correction, and vertical straight lines may not be aligned with the y-axis direction and may be inclined or the horizon may be inconsistent with the x-axis direction have. Therefore, tilting correction is required for more accurate image processing.

In step 120, the input color image is skew corrected to generate a corrected first image.

The skew correction is performed to reduce the distortion in the three-dimensional modeling. This inclination correction can be performed by projecting the image to a virtual viewpoint by rotating the image with respect to x-axis coordinate, y-axis coordinate, and z-axis coordinate in space.

In step 130, the vanishing point of the first image is detected.

First, a horizontal straight line is extracted from the first image. There are various algorithms for extracting a straight line from an image. Some of the extracted straight lines in the horizontal direction pass through the same first vanishing point. Another straight line passes through the second vanishing point. In this way there may be several vanishing points other than the first vanishing point or the second vanishing point.

Then, an arbitrary two straight lines out of a plurality of extracted horizontal straight lines are selected, and an intersection point is obtained based on the directions of the two straight lines. In the same way, by finding another intersection point by another two straight lines, a plurality of intersection points can be obtained.

After the skew correction, the y coordinate values of the plurality of intersection points should be equal to the y coordinate of the horizon but may not be equal to each other. Therefore, the horizon of the first image can be determined by clustering the y coordinate values of the plurality of intersections, removing the outlier, and calculating the final y coordinate value.

And at least one final x coordinate value (s) is computed by removing the outlier so that the x coordinate values of the plurality of intersections are clustered, and wherein at least one of the points on the horizon having the final x coordinate value (S) is detected as the vanishing point of the first image.

In step 140, a plurality of ground line candidates of the first image are determined.

Straight lines in the horizontal direction of the first image are grouped. The vertical boundary of the first image is determined by checking the direction of the straight line (i.e., passing the same vanishing point) while checking the change of the color value in the x-axis direction while advancing the first image in the x-axis direction.

And a ground line candidate is determined at a portion below the horizon of the first image. Starting from any pixel at the left edge of the first image, it is possible to determine the ground line candidate as it proceeds in the x-axis direction.

In step 150, the ground line of the first image is determined.

A ground line candidate having the largest change is detected as a ground line 1010 of the first image based on at least one of a change in the color value in the y-axis direction and a change in the texture under the horizon of the first input image do.

In step 160, 3D modeling of the first image is performed.

The vertical portion of the 2D image, which is the upper portion with respect to the ground line, is projected in the direction perpendicular to the ground, and the ground portion in the lower portion is projected in the horizontal direction.

2 is a flowchart illustrating a method of detecting a vanishing point in an image processing method according to an embodiment of the present invention.

In step 210, a horizontal straight line is extracted from the first image.

There are various algorithms for extracting a straight line from an image. According to an exemplary method, a plurality of edges are extracted from an input image using a Sobel or Canny edge detector, and only a plurality of straight lines of the edge are detected. A non-horizontal component of the plurality of straight lines is filtered out to extract a straight line in the horizontal direction.

Some of the extracted straight lines in the horizontal direction pass through the same first vanishing point. Another straight line passes through the second vanishing point. In this way there may be several vanishing points other than the first vanishing point or the second vanishing point.

In step 220, the horizon of the first image is determined. An arbitrary straight line out of a plurality of horizontal straight lines extracted in step 210 is selected and an intersection point is obtained based on the directions of the two straight lines. In the same way, another number of intersections by two other straight lines can be obtained. If the error is ignored, the y coordinate values of the intersections are the same, and the straight line constituted by the pixels having the y coordinate value becomes the horizon line of the first image.

However, considering the error in actual image processing, the y coordinate values of a plurality of intersection points may not be equal to each other. Accordingly, according to an embodiment of the present invention, the horizon of the first image can be determined by clustering y coordinate values of the plurality of intersections, removing an outlier, and calculating a final y coordinate value .

At step 230, at least one vanishing point of the first image is detected.

If the error is ignored as described above, the intersection of any two straight lines in the plurality of horizontal directions can be determined as the vanishing point. However, in the actual image processing, it is necessary to consider the above-mentioned error.

Therefore, the x-coordinate values of the plurality of intersections are clustered to remove the outlier. At least one final x-coordinate value (s) is calculated, and the point (s) having the final x-coordinate value (s) of the points above the horizon is detected as the vanishing point of the first image.

3 is a flowchart illustrating a method of determining a plurality of ground line candidates in an image processing method according to an embodiment of the present invention.

In step 310, straight lines in the horizontal direction of the first image are grouped. The first image can be grouped into a plurality of groups while checking the direction of a straight line (i.e., passing through the same vanishing point) while proceeding in the x-axis direction. According to an embodiment of the present invention, the check is performed only on the horizon of the first image.

In step 320, the vertical boundary of the first image is determined. As a result of the check, there is a boundary line between the grouped groups in a part where the directions of the straight lines change greatly, and the boundary line is determined as a vertical boundary line dividing the first image into a plurality of groups.

Meanwhile, according to an embodiment of the present invention, a horizon line and an under-sky area are determined as in a building area in the input image, and the check is performed only within the area to determine a vertical boundary line of the first image.

Also, a change in the color value in the x-axis direction in the region above the horizon may be checked, for example, a portion where a difference in brightness or color difference greatly increases may be determined as a vertical boundary of the first image. Of course, it is also possible to determine the vertical boundary line by considering both the direction and the hue of the horizontal straight line. This vertical boundary line is a point where the ground line is bent, that is, a corner where the linear direction changes. These vertical boundaries generally occur at the edges of the building surface, at the boundary between the building and the building, or at the boundary of the building and the building's left or right sky.

In step 330, a plurality of ground line candidates of the first image are determined.

According to an embodiment of the present invention, a ground line candidate is determined in the lower part of the horizon of the first image. Starting from any pixel at the left edge of the first image, it is possible to determine the ground line candidate as it proceeds in the x-axis direction.

FIG. 4 is a diagram illustrating a color image input in step 110 of FIG. 1 according to an embodiment of the present invention.

Region 410 is a sky portion. In the image processing according to an embodiment of the present invention, the region 410 may be filtered out. And, the area 420 is a building part, and is a vertical structure. In the 3D image modeling process, this area 420 will be erected vertically. On the other hand, the region 430 is a ground portion, which must be a horizontal surface in the 3D image modeling process. On the other hand, the ground line 440 is a boundary between the vertical region 420 and the horizontal region 430 of the input color image.

The input color image of FIG. 4 is an image that has not undergone tilt correction, and the straight lines including the straight line 450 are inclined. Tilt correction is required for more accurate image processing.

FIG. 5 is a diagram illustrating a first image generated by performing skew correction on the image of FIG. 4 in step 120 of FIG. 1, according to an embodiment of the present invention.

The skew correction is performed to reduce the distortion in the three-dimensional modeling. The inclination correction may be performed by, for example, detecting the pitch and roll of the camera to correct the inclination of the input image. This inclination correction can be performed by rotating the image with respect to x-axis coordinate, y-axis coordinate, and z-axis coordinate in space and projecting the image to a virtual point of view. The input color image shown in FIG. 4 for illustrative purposes is an image to be skewed only in the x-axis coordinate direction, and the result is the first image shown in FIG.

It can be seen that all straight lines including the straight line 510 are orthogonal to the x-axis direction in the image and are skewed so that the horizon line is parallel to the x-axis direction. The image processing method according to an embodiment of the present invention detects a ground line 520 which is a boundary between a building and a ground. The upper portion may be vertically oriented with the ground line 520 as a reference, and the lower portion may be 3D modeled as a horizontal direction.

FIG. 6 shows a result of extracting a horizontal straight line from the first image of FIG. 5 in step 210 of FIG.

A number of horizontal straight lines including straight lines 611, 612, 613 and 614 and straight lines 621, 622, 623 and 624 were extracted from the first image. There are various algorithms for extracting a straight line from an image. According to an exemplary method, a plurality of edges are extracted from an input image using a Sobel or Canny edge detector, and only a plurality of straight lines of the edge are detected. A non-horizontal component of the plurality of straight lines is filtered out to extract a straight line in the horizontal direction.

The horizontal straight lines 611, 612, 613, and 614 pass the same first vanishing point (not shown), ignoring the error. The straight lines 621, 622, 623, and 624 also pass through the same second vanishing point (not shown) if they also ignore the error.

FIG. 7 is a diagram illustrating a result image obtained by determining the horizon of the first image in step 220 of FIG. 1 according to an embodiment of the present invention.

Two straight lines out of a plurality of horizontal straight lines extracted in FIG. 6 are selected, and an intersection point is obtained based on the directions of the two straight lines. If straight line 711 and straight line 712 are selected, they intersect at point 714. Also, if a straight line 721 and a straight line 722 are selected, they intersect at point 724.

In the same way, if straight line 711 and straight line 713 are selected, they intersect at point 714, neglecting the error. And intersects at point 724 if straight line 721 and straight line 723 are selected. In this case, since the tilt of the image is corrected, the y coordinate value of the point 714 is the same as the y coordinate value of the point 724. A straight line 730 including the point 714 and the point 724, It becomes a horizon line. That is, the horizon level is determined.

However, considering the error in the actual image processing, a plurality of intersection points will be obtained near the point 714, and a plurality of intersection points will be obtained near the point 724. The point 714, the point 724, and the y coordinate values of the plurality of intersections may not be equal to each other.

Thus, according to one embodiment of the present invention, the y coordinate values of a plurality of points of intersection near the point 714 and a plurality of points of intersection near the point 724 are clustering to remove the outlier, By calculating the final y coordinate value, a straight line 730, which is the horizon of the first image, can be determined.

According to an embodiment of the present invention, then, at least one vanishing point may be detected from a straight line 730 which is the horizon of the first image.

If the error is ignored as described above, the point 714 may be determined as the first vanishing point of the first image. Also, the point 724 may be determined as the second vanishing point of the first image. However, in the actual image processing, it is necessary to consider the above-mentioned error.

According to one embodiment of the present invention, first, the x-coordinate values of a plurality of intersections near the point 714 and a plurality of intersections near the point 724 are clustered to remove the outlier. So that at least one final x coordinate value (s) is calculated and the point (s) having the final x coordinate value (s) of the points on the straight line 730 are detected as the vanishing point of the first image . In FIG. 7, for ease of understanding, point 714 coincides with the first vanishing point of the first image, and point 724 corresponds to the second vanishing point of the first image.

8 is a diagram illustrating an image in which a vertical boundary line of the first image is determined according to an embodiment of the present invention.

The straight line 811 intersects the point 714 in Fig. 7, which is the first vanishing point of the first image. The straight lines 812 and 813 are also the same. However, the straight lines 821, 822, 823 intersect the point 724, which is the second vanishing point of the first image. As described in step 310 of FIG. 3, the first image can be grouped into a plurality of groups while checking the direction of a straight line (i.e., passing through the same vanishing point) while proceeding in the x-axis direction. According to one embodiment of the present invention, the check is performed only at the upper portion of the straight line 830, which is the horizon of the first image.

As a result of the check, straight lines 841, 842, 843, 844, and 845, which are portions in which the directions of the straight lines greatly change, were determined as vertical boundaries dividing the first image into six groups.

According to an embodiment of the present invention, steps 310 to 320 of FIG. 3 are performed only in the building area 420, except for the sky area 410 of FIG. 4, The straight lines 841, 842, 843, 844, and 845 as boundary lines are determined.

In addition, a change in the color value in the x-axis direction is checked in the building area 420. For example, a portion where the difference in the contrast and the difference in color is large is defined as the straight lines 841, 842, and 843 , 844, 845). Of course, it is also possible to determine the vertical boundary line by considering both the direction and the hue of the horizontal straight line.

As will be described later, the straight lines 841, 842, 843, 844, and 845, which are the vertical boundary lines, are the corners where the ground line is bent, that is, the straight line direction changes.

9 is a diagram illustrating an image in which a plurality of ground line candidates are determined according to an embodiment of the present invention.

In accordance with an embodiment of the present invention, the ground line candidates 931, 932, 933, 934, etc. are determined below the straight line 920, which is the horizon line. Although only four ground line candidates 931, 932, 933, and 934 are illustrated in FIG. 9 for the sake of convenience, it is possible to determine the ground line candidate starting from any pixel on the left edge of the first image, have. For example, the ground line candidate 931 initially advances in a direction toward a point 714, which is the first vanishing point of the first image, and changes its direction at the vertical boundary 911 to be the second vanishing point ( 724, respectively. And passes through each of the vertical boundary lines 912, 913, 914, and 915 to change the direction and ends at the right edge of the first image.

FIG. 10 is a diagram illustrating a result of determining a ground line of the first image among the plurality of ground line candidates of FIG. 9 according to an embodiment of the present invention.

(931, 932, 933, 934, etc.) among the ground line candidates (931, 932, 933, 934, etc.) based on at least one of the change of the color value in the y-axis direction and the change of the texture below the horizon line And detects the ground line candidate 933 having the largest value. The ground line candidate 933 is detected as the ground line 1010 of the first image.

The ground line 1010 differs from the actual ground line 520 of the first image shown in FIG. This is because the ground line of the left building and the ground line of the right building are not continuous in the x-axis direction with respect to the vertical boundary 912. That is, among the ground lines 931, 932, 933, 934, etc., the detected ground line 1010 may not exactly coincide with the actual ground line 520. These discrepancies are usually found in small parts that can be ignored.

11 is a diagram illustrating a method of 3D modeling according to an embodiment of the present invention.

The area 1101, which is the vertical part of the 2D image subjected to the tilt correction, is projected in the direction perpendicular to the paper surface, and becomes the area 1104 in the 3D model. An area 1103, which is a ground part of the 2D image below the ground line 1102, is projected in a direction parallel to the ground to become an area 1105 in the 3D model. That is, in the 3D model composed of the area 1104 and the area 1105, the buildings included in the area 1104 were modeled in the vertical direction.

The image processing method according to the present invention may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions recorded on the medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

While the invention has been shown and described with reference to certain preferred embodiments thereof, it will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. This is possible. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined by the equivalents of the claims, as well as the claims.

1 is a flowchart showing an image processing method according to an embodiment of the present invention.

2 is a flowchart illustrating a method of detecting a vanishing point in an image processing method according to an embodiment of the present invention.

3 is a flowchart illustrating a method of determining a plurality of ground line candidates in an image processing method according to an exemplary embodiment of the present invention.

4 is a diagram illustrating an input image according to an exemplary embodiment of the present invention.

FIG. 5 is a diagram illustrating a first image generated by performing skew correction on the image of FIG. 4 according to an embodiment of the present invention.

FIG. 6 is a diagram illustrating an image obtained by extracting a straight line in a horizontal direction from the first image of FIG. 5 according to an embodiment of the present invention.

FIG. 7 is a diagram illustrating an image in which a horizon of the first image is determined according to an embodiment of the present invention.

8 is a diagram illustrating an image in which a vertical boundary line of the first image is determined according to an embodiment of the present invention.

9 is a diagram illustrating an image in which a plurality of ground line candidates are determined according to an embodiment of the present invention.

FIG. 10 is a diagram illustrating a result of determining a ground line of the first image among the plurality of ground line candidates of FIG. 9 according to an embodiment of the present invention.

11 is a diagram illustrating a method of 3D modeling according to an embodiment of the present invention.

Claims (17)

A first step of generating a first image by performing an inclination correction on an input color image; A second step of determining a plurality of ground line candidates from the first image; And A third step of detecting a ground line of the input color image based on at least one of a color change in the y-axis direction and a texture change in the y-axis direction of the plurality of ground line candidates And an image processing method. 2. The method according to claim 1, Determining a horizon of the first image; Detecting at least one vanishing point of the first image from the horizon; And Determining a plurality of ground line candidates of the first image based on the vanishing point And an image processing method. 3. The method of claim 2, Wherein the determining of the horizon of the first image comprises: Extracting at least one horizontal straight line from the first image; And Determining a horizon of the first image based on a y-axis coordinate value of an intersection of the at least one horizontal straight line And an image processing method. The method of claim 3, Wherein the step of detecting at least one vanishing point of the first image comprises: And extracting at least one vanishing point of the first image from the horizon based on an x-axis coordinate value of an intersection of the at least one horizontal direction straight line. 5. The method of claim 4, Wherein the step of determining a plurality of ground line candidates of the first image based on the vanishing point comprises: Determining a boundary line of each group generated by grouping straight lines associated with the same vanishing point among the at least one horizontal straight line as a vertical boundary line of the first image; And Determining a plurality of ground line candidates composed of line segments having edges at a vertical boundary line of the first image And an image processing method. 6. The method of claim 5, Wherein the vertical boundary line of the first image is determined as a boundary line of each group generated by grouping straight lines related to the same vanishing point among the at least one horizontal straight line of the first area on the horizon of the first image. 6. The method of claim 5, Wherein the plurality of ground line candidates are determined in a second region below the horizon of the first image. 5. The method of claim 4, Wherein the step of determining a plurality of ground line candidates of the first image based on the vanishing point comprises: Determining a vertical boundary of the first image based on a color value change in an x-axis direction in an area above the horizon of the first image; And Determining a plurality of ground line candidates composed of line segments having edges at a vertical boundary line of the first image And an image processing method. 9. The method of claim 8, Wherein the plurality of ground line candidates are determined in a second region below the horizon of the first image. The method of claim 3, Wherein the determining of the horizon of the first image further comprises the step of clustering the y-axis coordinate values of the intersection of the at least one horizontal straight line to remove the outlier. 5. The method of claim 4, Wherein the step of detecting at least one vanishing point of the first image comprises: And clustering the x-axis coordinate values of the intersection of the at least one horizontal straight line to remove the outlier. The method according to claim 1, Wherein the first image is projected in a direction perpendicular to the ground on the basis of the ground line and the lower region of the ground line is projected in a direction parallel to the ground to perform 3D modeling of the first image Step 4 to Perform Further comprising the steps of: A first step of extracting at least one horizontal straight line from an input image; A second step of determining a horizon of the input image based on a y coordinate value of an intersection of the at least one horizontal straight line; And A third step of detecting at least one vanishing point from the horizon based on an x-coordinate value of an intersection of the at least one horizontal straight line, And an image processing method. 14. The method of claim 13, Wherein the input image is a skew corrected image by an iterative correction process. 15. The method of claim 14, A fourth step of determining a plurality of ground line candidates constituted by the at least one horizontal straight line within a second area below the horizon of the input image; And A fifth step of determining a ground line of the input image based on at least one of a change in the color value in the y-axis direction of the ground line candidate and a change in texture in the y- Further comprising the steps of: 16. The method of claim 15, And a sixth step of projecting the upper portion of the ground line in a direction perpendicular to the ground in the input image and projecting the lower portion of the ground line in a direction parallel to the ground to perform 3D modeling. A computer-readable recording medium storing a program for causing a computer to execute the method according to any one of claims 1 to 16.

Images