KR101439406B1 - Image Processing Apparatus and Method for obtaining depth information using an image - Google Patents

Abstract

An image processing method according to an exemplary embodiment of the present invention includes: forming a plane mesh composed of a plurality of vertices and a plurality of polygons corresponding to a plurality of pixels of an image; Forming a three-dimensional mesh by setting a height corresponding to a brightness value of each pixel corresponding to each of the vertices to the plurality of vertices; Simplifying the stereoscopic mesh by combining mesh regions in which the height difference is within a predetermined range; Matching a plurality of reference points having depth values of the image with the image at a view point of the image; And extracting depth values of the vertexes by associating vertices of the simplified stereoscopic mesh with the matched reference points.

Description

TECHNICAL FIELD The present invention relates to an image processing apparatus and an image processing method for acquiring depth information using an image,

The present invention relates to an image processing apparatus and an image processing method for acquiring depth information using an image, and more particularly, to an image processing apparatus and method for acquiring depth information of an object contained in an image from an image, And a method of processing the same.

3D restoration technology was used mainly by experts such as product design and inspection, reverse engineering, and image contents production. Recently, with the commercialization of low-cost small depth cameras such as Microsoft's Kinect, public interest has increased. Such 3D reconstruction techniques have been actively studied in the field of computer vision, and methods for reconstructing three-dimensional objects from images such as structure from motion, space carving, and shape from shading have been mainly developed and studied.

Among the above examples, the structure from motion acquires three-dimensional spatial information from two-dimensional images of various viewpoints by estimating camera motion occurring when photographing the surroundings of an object at various positions. The structure form motion finds the correspondence between images by using feature points that can be easily extracted from the image, such as corners, in order to find out camera movements from multiple images. Then, after the image matching process, the camera is automatically corrected to generate three-dimensional spatial information to complete the three-dimensional reconstruction. In this case, the three-dimensional spatial information is generally a set of reference points having three- The three-dimensional reference points acquired in this way are not dense. Therefore, it is difficult for the user to understand the shape of the reconstructed object in the three-dimensional form or to grasp the boundary between the objects.

In addition, the set of three-dimensional reference points has a limitation that it is difficult to use variously because it is difficult to represent three-dimensional information or to deform a shape through rendering, unlike a three-dimensional mesh, . However, this conversion process is generally not easy, and especially when the set of reference points is rough, it is more difficult to obtain the mesh of the desired shape.

Therefore, in order to solve the above problems, the embodiments of the present invention can easily acquire the dense three-dimensional depth information of the entire image using the brightness information of the image photographed at a specific point in time and the reference points extracted at the same point in time .

In addition, embodiments of the present invention aim to provide three-dimensional depth information that can clearly distinguish the boundaries between objects in order to convey clear depth information to a user.

According to an aspect of the present invention, there is provided an image processing method for forming a plane mesh composed of a plurality of vertices and a plurality of polygons corresponding to a plurality of pixels of an image, step; Forming a three-dimensional mesh by setting a height corresponding to a brightness value of each pixel corresponding to each of the vertices to the plurality of vertices; Simplifying the stereoscopic mesh by combining mesh regions in which the height difference is within a predetermined range; Matching a plurality of reference points having depth values of the image with the image at a view point of the image; Extracting depth values of the vertexes by matching vertices of the simplified stereoscopic mesh with the matched reference points; And a control unit.

Preferably, the method further comprises adjusting the depth value of the two vertices so that the depth value difference between the two neighboring vertices corresponds to the brightness value difference of the pixels corresponding to the two vertices.

The step of adjusting the depth values of the two vertices may include adjusting the depth values of the two vertices within a predetermined range if the difference in brightness values of two pixels corresponding to the two vertices is within a predetermined range, The depth value of each of the two vertices is adjusted.

The method further includes the step of forming another stereoscopic mesh by setting the height of the vertices of the simplified stereoscopic mesh on the basis of the extracted depth value.

Further, the step of forming the three-dimensional mesh is characterized by setting the height in the normal direction of the plane mesh.

The step of simplifying the stereoscopic mesh may further include reducing the number of the plurality of polygons by removing an edge shared by the neighboring polygons when the direction difference of the normal vector of the neighboring polygon is within a predetermined range .

The step of simplifying the stereoscopic mesh is characterized by using a quadric error metric (QEM) algorithm.

The step of matching the plurality of reference points to the image may include matching polygons of the simplified stereoscopic mesh with respect to each reference point.

The step of matching the plurality of reference points to the image may include ignoring the height value of the three-dimensional mesh, recognizing the three-dimensional mesh as another plane mesh, and matching the plurality of reference points with the image.

The step of extracting the depth values of the vertices may include projecting the matched reference point on the image so that the distance between the matched reference point and the plane of another plane mesh corresponds to the depth value of the matched reference point ; And extracting a depth value of the vertices of the polygon corresponding to the projected reference point from the depth value of the projected reference point.

The step of extracting the depth values of the vertices may include interpolating the extracted vertices to the reference point, wherein the extracted vertexes have the same depth value as the reference point .

The polygon of the plane mesh is a triangle.

According to another aspect of the present invention, there is provided an image display apparatus including: a mesh generation unit that forms a plane mesh composed of a plurality of vertices and a plurality of polygons corresponding to a plurality of pixels of a single image; Forming a three-dimensional mesh by setting a height corresponding to a brightness value of each pixel corresponding to each of the vertices at a plurality of vertices of the plane mesh, combining the mesh areas having a height difference within a predetermined range, A mesh simplification unit for simplifying the mesh size; A reference point matching unit for projecting a plurality of reference points having depth values of the image on the image so that a depth value of each reference point corresponds to a distance between the image and each reference point; And a depth information extraction unit for extracting depth values of the respective vertexes by associating vertices of the simplified stereoscopic mesh with the projected reference points.

Preferably, the image processing apparatus further includes a depth information adjusting unit that adjusts the depth values of the two vertices so that the depth value difference between the two neighboring vertices corresponds to the brightness value difference between the pixels corresponding to the two vertices. do.

The depth information extraction unit may extract a depth value of a vertex of a polygon matching the reference point from a depth value of one reference point.

The image processing apparatus and the image processing method according to an embodiment of the present invention configured as described above generate a plane mesh using brightness information of an image and provide depth information of the image for each vertex of the generated mesh It is possible to easily acquire three-dimensional depth information for all the vertices.

According to an embodiment of the present invention, when the depth information of the image is the same, the depth information of each extracted vertex is adjusted according to the brightness information of the pixel under the assumption that the brightness information in the image is similar, There is a part or the density is not uniform), it is possible to interpolate the (empty) part where information is empty.

1 is a structural block diagram of an image processing apparatus according to an embodiment of the present invention.
2 is a flowchart of an image processing method according to an embodiment of the present invention.
3 to 7 are schematic views for explaining a mesh generation step and a mesh simplification step according to an embodiment of the present invention.
8A is an input image according to an embodiment of the present invention.
FIG. 8B is a simplified image of the input image of FIG. 8A according to an embodiment of the present invention.
9 is a schematic diagram illustrating a plurality of reference points for an image in accordance with an embodiment of the present invention.
FIGS. 10 and 11 are cross-sectional views of I-II in FIG. 8 for explaining a reference point projecting step and a depth information extracting step according to an embodiment of the present invention.
12 is a schematic diagram for extracting depth values of three vertices from a reference point in the depth information extracting step according to an embodiment of the present invention.
13 is an image obtained by meshing the depth information finally adjusted according to an embodiment of the present invention.

Hereinafter, an image processing apparatus and an image processing method according to an embodiment of the present invention will be described in detail with reference to the drawings.

In the present specification, the same reference numerals are given to the same components in different embodiments, and the description thereof is replaced with the first explanation.

As used herein, the singular forms "a", "an" and "the" include plural referents unless the context clearly dictates otherwise.

It is to be understood that elements of the drawings attached hereto may be shown as being enlarged or reduced for convenience of description.

Also, terms including ordinals such as first, second, etc. used in this specification may be used to describe various elements, but since the terms are used only for the purpose of distinguishing one element from another, The elements are not limited by these terms.

First, a configuration of an image processing apparatus and an image processing method according to an embodiment of the present invention will be described with reference to FIGS. 1 and 2. FIG. The image processing apparatus according to an exemplary embodiment of the present invention may include a mesh generation unit 100, a mesh simplification unit 200, a reference point generation unit 300, a reference point matching unit 400, a depth information extraction unit 500, An adjustment unit 600, and a display unit 700.

The mesh generation unit 100 generates a mesh for an image captured at a specific view point. The image may be composed of a plurality of pixels each having brightness and color. The mesh generation unit 100 may generate a plane mesh composed of polygons matched with a plurality of pixels (S10). At this time, the polygon is composed of a plurality of vertices, preferably a triangle.

The mesh simplification unit 200 converts the plane mesh into a three-dimensional mesh and performs a simplification process.

First, the mesh simplifying unit 200 adds a height value to each vertex of the plane mesh to generate a three-dimensional mesh (S20). The height value may be a value corresponding to a brightness value of each pixel. Therefore, the generated stereoscopic mesh can be generated in proportion to the brightness value of the image. Then, the simplification process is performed on the generated stereoscopic mesh (S30). Since the mesh simplification process requires a long execution time to process finely structured three-dimensional data, It is a process of simplifying to reduce to an optimal threshold value with information. The mesh simplification process may be performed by eliminating a specific vertex or eliminating edges shared by neighboring two polygons.

The reference point generator 300 generates a plurality of reference points having the depth value at the same point in time as the image from the image. The reference point generating unit 300 may generate a plurality of reference points through a configuration such as a depth camera. However, the reference point generator 300 may not be included in the image processing apparatus according to an embodiment of the present invention. In this case, the depth value of the image is acquired from the outside by the user and input to the image processing apparatus. An embodiment of the invention may be implemented.

The reference point matching unit 400 matches the generated plurality of reference points to the image at a time point when the image is captured (S40). At this time, the reference point matching unit 400 recognizes the three-dimensional mesh as another plane mesh by omitting the height value of the simplified three-dimensional mesh, and matches the plurality of reference points to another plane mesh.

The depth information extraction unit 500 extracts the depth values of the vertices of the plane mesh (S50). Specifically, the depth information extracting unit 500 projects the reference point on the another plane mesh (image) so that the distance between the reference point and another plane mesh (plane of the image) corresponds to the depth value of the reference point. Then, the vertexes of the plane mesh correspond to the reference points, and the depth value of the vertices can be extracted through an operation of substituting the relation between the reference points and the corresponding vertices into a predetermined equation.

The depth information adjustment unit 600 adjusts the depth value of each vertex so that the difference in extracted depth values of the vertices is similar to the difference in brightness values of the pixels of the image corresponding to the vertices (S60). The display unit 700 can provide the adjusted depth value as data that can be recognized by the user. At this time, the mesh generation unit 100 generates a stereoscopic mesh by giving a height value to each vertex of the plane mesh using the adjusted depth value, and the display unit 700 displays the dense depth information of the image in one eye The generated mesh can be displayed to clearly recognize the three-dimensional mesh.

Hereinafter, a plane mesh generation step (S10), a three-dimensional mesh generation step (S20), and a mesh simplification step (S30) according to an embodiment of the present invention will be described in detail with reference to FIG. 3 to FIG.

The mesh generator 100 (see FIG. 1) converts the image 110 selected by the user into a plane mesh. Here, the image 110 may be a color image 110, and may include a plurality of pixels having brightness and color.

At this time, the mesh generation unit 100 divides all the pixels in the image 110 into a plurality of pixel groups composed of four adjacent pixels as shown in FIG. In Figure 3, three groups of pixels are shown enlarged. For example, the first pixel group G1 may include an eleventh pixel P11, a twelfth pixel P12, a thirteenth pixel P13, and a fourteenth pixel P14. The third pixel group G3 is composed of the twenty-first pixel P21, the twenty-second pixel P21, the twenty-second pixel P21, and the twenty-second pixel P22, (P22), a 23rd pixel (P23), and a 24th pixel (P24).

In this case, the brightness is high in the order of the twenty-first pixel P21, the thirteenth pixel P13, and the twelfth pixel P12, and the remaining pixels have the lowest brightness values.

Then, the mesh generation unit 100 determines pairs of pixels located diagonally to each other in the pixel group, and compares brightness values of the pairs of pixels. Then, the mesh generator 100 draws diagonal lines 112a and 112b in the direction of a pixel pair having a smaller brightness difference. As a result, the pixel pairs divided by the diagonal are the pixels having the largest difference in brightness value. At this time, one pixel group is divided into two triangles. For example, a first diagonal line 112a is generated across the eleventh pixel P11 and the fourteenth pixel P14 in the first pixel group G1, and the first diagonal line 112a is generated in the second pixel group, A second diagonal line 112b across the 22 pixels is generated and a third diagonal line 112c across the 22nd pixel P22 and the 23rd pixel P23 is generated in the third pixel group G3.

In this form, the plane mesh consists of a plurality of triangles and vertices. FIG. 4 shows a part of the plane mesh composed of the triangle and the vertices, and the first vertex v11 and the second vertex v12 are connected by one edge. Wherein each vertex of the plane mesh corresponds to each pixel of the image.

Then, the mesh simplifying unit 200 (see FIG. 1) allocates a height value to the plane mesh to form a three-dimensional mesh. First, as shown in FIG. 5, a brightness value corresponding to a brightness value of each pixel is assigned to each vertex of the plane mesh. As a result, a height is set at each vertex of the plane mesh. For example, in FIG. 4, the first vertex and the second vertex v11 and v12 of the plane mesh are assigned a first brightness value and a second brightness value according to corresponding pixels, and the first brightness value and the second brightness value The first height and the second height h1 and h2 are set. 5, all of the vertices have different heights, but the first height h1 of the reset first vertex v11 'is similar to the second height h2 of the second vertex v12'.

 Then, connecting the vertices having the height with all the edges creates a stereoscopic mesh having the shape as shown in FIG. The triangular meshes have normal vector components (n1 ~ n6) in different directions due to their height differences.

 Thereafter, the mesh simplification unit 200 simplifies the three-dimensional mesh. At this time, a method of simplifying the mesh may be selected by reducing the number of triangles by combining neighboring vertices, or by reducing the number of triangles by removing the edges shared by neighboring triangles and creating new vertices. At this time, an algorithm using a scheme such as QEM (Quadric Error Metic) or Variational Shape Approximation can be utilized for selection of an edge to be removed and selection of a new vertex to be generated. Hereinafter, a method of removing edges and generating new vertices will be described as an example of a mesh simplification method, and an example using a QEM algorithm will be described as an example.

The mesh simplification unit 200 simplifies the stereoscopic mesh based on the brightness value of each pixel. Therefore, the criterion of mesh simplification is related to the height value of each vertex. More specifically, how the directions of the normal vectors of neighboring triangles are similar is a basis of mesh simplification. That is, when the angular difference between the normal vector components of the neighboring triangles falls within the preset range, the mesh simplifying unit 200 removes the edges shared by the neighboring triangles, sets a new vertex, Simplify the stereoscopic mesh by setting new edges from other vertices.

For example, FIG. 6 shows the normal vector components of some triangles, which measure the angular difference of the vector components. The vectors neighboring the first vector n1 are the second vector n2, the third vector n3, and the fourth vector n4. Among them, the vector having the smallest angular difference from the first vector n1 is the second vector n2. The vector adjacent to the second vector n2 is a first vector n1, a fifth vector n5 and a sixth vector n6. Of these, the second vector n2 has the smallest angular difference The vector is the first vector (n1). Therefore, the first vector n1 and the second vector n2 become the most similar vector. At this time, the mesh simplifying unit 200 determines whether the angular difference between the first vector n1 and the second vector n2 is within a predetermined range. If it is determined that the angular difference is within the predetermined range, And removes the edge (E) shared by the triangle of the second vector (n2). Since the edge E includes the first vertex v11 'and the second vertex v12', the first and second vertices v11 'and v12' are also removed. As shown in FIG. 7, the stereoscopic mesh is simplified by generating the third vertices v13 and v13 'and creating new edges connecting the vertices and the third vertices v13 and v13'.

The removal of the edge during the mesh simplification process is carried out until the number of triangles becomes equal to or less than the specific number, as long as the angular difference of the normal vectors is within the predetermined range, It proceeds. This is because the new triangle is generated while removing the edge.

As a result, the simplified mesh has edges formed at the boundary of two regions having different brightness values.

Here, FIGS. 8A and 8B will be described by way of example. FIG. 8A shows the first input image, and FIG. 8B shows an input image in which the mesh simplification is performed. In FIG. 8A, the front of the drawer is ocher-colored and has a high brightness, while the side of the drawer is gray-based and low in brightness. Referring to FIG. 8B, the boundary between the side of the drawer and the front of the drawer is clearly defined by the edge, and the brightness of the side and front of the drawer is clearly distinguished from the edge.

In this manner, the mesh simplification unit 200 generates large triangles that include regions having small differences in brightness of images. As a result, since the triangles corresponding to the pixels with large brightness differences have a steep slope difference with respect to the edge between the triangles, the region having a large brightness difference can be clearly distinguished by the edge.

Next, a reference point projection step (S40) and a depth information extraction step (S50) according to an embodiment of the present invention will be described in detail with reference to FIG. 9 to FIG.

First, a reference point having depth information of an image is generated by a reference point generator 300 (see FIG. 1), and a plurality of reference points have a plurality of depth values of an image captured at one point in time. When a depth camera is used, a plurality of reference points can not have a uniform density when they are implemented by a method such as a motion from structure, and a limitation of a depth range in which a depth value can be measured. There can be a lot of gaps in part because of problems covered by objects. Therefore, it can not be the desired depth information in one embodiment of the present invention.

At this time, the reference point generator 300 may be implemented by a depth camera or the like, and a reference point may be obtained by a method such as a motion from structure. The depth camera measures the distance from the camera to the object through the infrared sensor and outputs it as an image. The reference point matching unit 400 (see FIG. 1) matches the reference points 120 with the image 110, as shown in FIG. At this time, the stereoscopic mesh generated corresponding to the image 110 is recognized as a plane mesh in which only the height value is ignored. At this time, the matching relation of the triangle of the plane mesh and the reference points 120 can be known by matching the reference points. Generally, the projected reference point 120 matches one triangle. However, since the reference point 120 has a non-uniform density, when the reference point 120 is projected on the plane mesh, a triangle that does not match the reference point may exist.

1) projects the reference point 120 such that the distance between the matched reference points 120 and the image 110 corresponds to the depth value of the reference point 120. In this case, Here, with reference to FIG. 10, a cross section of the regions I to II will be described. In FIG. 10, the reference points 120 are arranged at different intervals with respect to the surface of the image 110. For example, since the first reference point 120 has a first depth value, an interval between the first reference point 120a and the image 110 may be equal to the first depth value d1. Since the second reference point 120b has a second depth value, an interval between the second reference point 120 and the image 110 may be equal to the second depth value d2. At this time, the first depth value d1 and the second depth value d2 are different.

Then, the position of the reference points scattered randomly and scattered is line-drawn, and the depth information of the image is shown as the reference line 125.

12, the depth values d _i1 , d _i2 , and d _{i3 of the} three vertices v _i1 , v _i2 , and v _i3 from the triangle of the plane mesh matching one reference point 120 and the reference point 120 _i3 can be extracted. Specifically, the reference point 120 and the three vertices by computing the relationship between (v _i1, v _i2, v _i3), the depth value of the new three vertices with the (d _i1, d _i2, d _i3), (v _i1 ' , v _i2 can be obtained ', v _i3'). At this time, the three vertices (v _i1 ', v _i2', v depth values _{i3 ') (d i1, d} i2, d i3) are the three vertices (v _i1 ', v _i2 ', and v _i3 'are interpolated to the projected reference point 120, the calculation result becomes the same as the depth value d120 of the reference point 120.

An operation for extracting the depth value of the cleavage point is performed by the following equation (1).

은 상기 삼각형의 평면의 법선 벡터를 의미한다.Here, si denotes a reference point, f (s _i ) denotes a plane of a triangle matching s _i , p _i denotes an arbitrary point included in the plane of the triangle,

Represents the normal vector of the plane of the triangle.

E is an error value defined from the depth value of the vertex. E is derived by summing the squares of the distance between each reference point and the plane of the matching triangle.

Here, the depth value of the three vertices can be obtained by calculating E to be minimum. The extracted result can provide dense depth information to the user. However, since the reference points are not uniform in density, there can not exist matching reference points for all the triangles. Thus, if there is an arbitrary vertex that does not have a matching reference point, the depth value of the arbitrary vertex can not be extracted. Further, when noise is reflected in the depth value of the reference point, the extracted cleavage point reflects the noise as it is.

Therefore, an embodiment of the present invention includes a step of adjusting the depth information through the depth information adjuster 600 (see FIG. 1) to compensate for the above.

If the brightness values of two pixels corresponding to two vertices connected by one edge are similar to each other under the assumption that the depth values are similar if the brightness values of neighboring pixels included in the image are similar, And the depth value of the two vertices has a value similar to the brightness value.

Specifically, when the brightness difference between the two pixels is within a predetermined range, the depth information adjustment unit 600 calculates a depth value difference between the two vertices corresponding to the two pixels to be within the preset range .

In order to adjust the depth value, the depth information adjusting unit 600 calculates Equation (2) by adding one more factor in Equation (1).

The added factor is an item for calculating the depth value of the vertex to be similar to the brightness value of the pixel. The vertex (v _i1 ) constituting one edge (e _i ) of the three edges of the triangle and the other vertex (v _i2 ). < / RTI > Wherein _{XY (v i1) -XY (v} i2) is the two vertex distance difference between the mean plane (v _i1, v _{i2) and, I (v i1) -I (} v i2) is the two vertices (v _i1 , v _i2 ), and Z (v _i1 ) -Z (v _i2 ) represents a difference in depth value between the two vertices v _i1 and v _i2 .

And the added factor is a constant for determining whether E is to be computed based on which of the first factor or the second factor is multiplied by the custom value w. For example, if a user assigns a very large value to w, the first factor is almost ignored, and the added factor takes up most of the E value. As a result, almost all vertices have approximately the same depth value. On the other hand, if the user assigns a very small value to w, the added factor is almost ignored, and the first factor takes up most of the E value. As a result, substantially the same result as in the expression (1) is obtained.

Therefore, w should be assigned an appropriate value. When a proper w is used, not only a depth value can be obtained by interpolating a vertex having no matching reference point but also an effect that the vertices connected to the edge have a similar value to each other, so that the noise is less reflected. Through such a process, depth information that clearly distinguishes the brightness boundary between objects photographed on the image can be calculated by reflecting the brightness value of the pixel on the extracted depth information. For example, in FIG. 8B, the boundary between the side surface and the wall surface of the drawer is indefinitely displayed, but in FIG. 13, the boundary between the side surface and the wall surface of the drawer is clearly indicated by an edge. In FIG. 8B, the boundaries of the entire silhouette of the drawer are unclear, but in FIG. 13, the boundaries of the entire silhouette of the drawer are clearly marked.

Finally, the extracted depth values may be provided to the user as numerical data. However, by setting the final extracted depth values in the plane mesh as height, a stereoscopic mesh is created so that the user can intuitively find the three-dimensional information intuitively. May be provided.

As described above, according to an embodiment of the present invention, a mesh is generated and simplified using brightness values of an image, and depth information of a vertex of the mesh is extracted using a depth value of a reference point, It is possible to generate dense three-dimensional depth information. In addition, an embodiment of the present invention can provide three-dimensional depth information that clearly distinguishes two areas having a large brightness difference in an image by matching extracted depth information with brightness information of a pixel.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments.

Therefore, the scope of the present invention is not limited thereto, and various modifications and improvements of those skilled in the art using the basic concept of the present invention defined in the following claims are also within the scope of the present invention.

100: mesh generation unit 200: mesh simplification unit
300: reference point generating unit 400: reference point matching unit
500: depth information extracting unit 600: depth information adjusting unit
700:

Claims (15)

Forming a planar mesh comprising a plurality of vertices and a plurality of polygons corresponding to a plurality of pixels of an image;
Forming a three-dimensional mesh by setting a height corresponding to a brightness value of each pixel corresponding to each of the vertices to the plurality of vertices;
Reducing the number of polygons of the stereoscopic mesh by combining vertices whose difference in height is within a predetermined range;
Matching a plurality of reference points having depth values of the image with the image at a view point of the image; And
Extracting a depth value of the vertexes by associating vertices of the stereoscopic mesh with the number of the polygons decreased at the matched reference point;
Wherein the depth information is obtained by using an image. The method according to claim 1,
Adjusting the depth values of the two vertices so that the depth value difference between the two neighboring vertices corresponds to the brightness value difference between the pixels corresponding to the two vertices. An image processing method for acquiring information. 3. The method of claim 2,
The step of adjusting the depth values of the two vertices may include adjusting the depth values of the two vertices such that when the difference in brightness values of the two pixels corresponding to the two vertices is within a predetermined range, And adjusting the depth values of the two vertices to obtain depth information using the image.
The method according to claim 1,
Further comprising the step of forming another stereoscopic mesh by setting the height of the vertexes of the stereoscopic mesh whose number of polygons has been reduced based on the extracted depth value to acquire depth information / RTI > The method according to claim 1,
Wherein the step of forming the solid mesh sets a height in a normal direction of the plane mesh. The method according to claim 1,
Wherein the step of reducing the number of polygons of the three-
If the height difference is proportional to the direction difference of the normal vector of the polygon,
Wherein the number of polygons of the three-dimensional mesh is reduced by removing an edge shared by the neighboring polygons when a direction difference of a normal vector of the neighboring polygon is within a predetermined range, An image processing method for acquiring information. The method according to claim 1,
Wherein the step of reducing the number of polygons of the stereoscopic mesh uses a quadric error metric (QEM) algorithm. The method according to claim 1,
Wherein the step of matching the plurality of reference points with the image matches the polygons of the three-dimensional mesh whose number of polygons is reduced with respect to each reference point. The method according to claim 1,
Wherein the step of matching the plurality of reference points with the image further comprises recognizing the stereoscopic mesh as another plane mesh and matching the plurality of reference points with the image, wherein the depth information is obtained using the image. 10. The method of claim 9,
The step of extracting the depth values of the vertices includes:
Projecting the matched reference point onto the image such that a distance between the matched reference point and a plane of another plane mesh corresponds to a depth value of the matched reference point; And
Extracting a depth value of a vertex of a polygon corresponding to the projected reference point from a depth value of the projected reference point;
Wherein the depth information is obtained by using an image. 11. The method of claim 10,
Wherein the step of extracting the depth values of the vertices comprises interpolating the extracted vertices to the position of the reference point so that the depth value of the extracted vertices becomes equal to the depth value of the reference point. A method of image processing for acquiring depth information by utilizing the method. The method according to claim 1,
Wherein the polygon of the plane mesh is a triangle, and the depth information is obtained using the image. A mesh generation unit for forming a plane mesh composed of a plurality of vertices and a plurality of polygons corresponding to a plurality of pixels of a single image;
Forming a stereoscopic mesh by setting a height corresponding to a brightness value of each pixel corresponding to each of the vertices at a plurality of vertices of the plane mesh and combining the vertices whose height difference is within a predetermined range, A mesh simplification unit for reducing the number of polygons;
A reference point matching unit for projecting a plurality of reference points having depth values of the image on the image so that a depth value of each reference point corresponds to a distance between the image and each reference point; And
A depth information extracting unit for extracting a depth value of each of the vertexes by associating vertices of the stereoscopic mesh with the number of the polygons reduced at the projected reference point;
And acquiring depth information using the image. 14. The method of claim 13,
And adjusting a depth value of the two vertices so that a depth value difference between two neighboring vertices corresponds to a brightness value difference between pixels corresponding to the two vertices. To acquire depth information. 14. The method of claim 13,
Wherein the depth information extraction unit extracts depth values of vertices of a polygon matching the reference point from depth values of one reference point.

Images