WO2024043406A1

WO2024043406A1 - Point cloud compression system and method for streaming

Info

Publication number: WO2024043406A1
Application number: PCT/KR2022/019640
Authority: WO
Inventors: 박우출; 장준환; 양진욱; 최민수; 이준석; 구본재
Original assignee: 한국전자기술연구원
Priority date: 2022-08-22
Filing date: 2022-12-05
Publication date: 2024-02-29
Also published as: KR20240027182A

Abstract

Disclosed in the present technology are a point cloud compression system and method for streaming. According to a specific example of the present technology, a spherical point cloud is converted to hexahedral points, then converted to tree structure points, each converted point of the tree structure is set as a reference point, and a normal vector of a point, in which an operation distance between each point of a child node of the set corresponding reference point and the corresponding reference point is less than or equal to the reference distance, is derived, such that calculation complexity and calculation speed according to the normal vector calculation can be reduced, and as encoding is performed for a point in which the inner value of the derived normal vector of the point and a user viewing direction vector obtained through a camera is a positive number, the amount of data to be encoded can be reduced, and communication resources and memory resources can be secured accordingly.

Description

Point cloud compression system and method for streaming

The present invention relates to a point cloud compression system and method for streaming, and more specifically, to encoding points where the dot product of a normal vector generated from a spherical image of produced content and a direction vector of the user's field of view obtained by a camera is a positive number. It is about a technology that allows this to be done.

Recently, along with the demand for virtual reality (VR) content, the demand for real-time services of VR content is increasing.

VR content must be rendered in a spherical shape, but conversion and transmission of spherical images have the problem of not conforming to existing video conversion formats or broadcast transmission formats.

Accordingly, a technology was developed to reconstruct a spherical image into a square plane by applying the Mercator projection method (cylindrical projection method) among map projection methods. However, this technology has a problem in that communication delays occur when transmitting video data during the process of converting to a flat surface.

Accordingly, the present applicant converted the spherical point cloud of the produced content into a hexahedron and then converted it into points in a tree structure, and in the converted tree structure, the reference point of the first node and the second node and lower child nodes that are child nodes of the first node. The data is encoded by deriving the normal vectors of the points where the calculated distance between the points is less than the standard distance, and encoding the points where the dot product of the normal vector of the derived point and the user's viewing direction vector obtained through the camera is a positive number. We would like to propose a plan to reduce the number and fundamentally secure communication resources.

[Prior art literature]

[Patent Document]

Republic of Korea Patent No. 10-1820359 (registered on January 15, 2018)

Accordingly, the present invention can reduce the number of data to be encoded, reduce the computational complexity and speed of normal vector calculation, and fundamentally secure communication resources accordingly.

The object of the present invention is not limited to the object mentioned above, and other objects and advantages of the present invention that are not mentioned can be understood through the following description and will be more clearly understood through examples of the present invention. In addition, it will be readily apparent that the objects and advantages of the present invention can be realized by means and combinations thereof as indicated in the claims.

A point cloud compression system for streaming according to an embodiment of the present invention to achieve the above-described object,

A point structuring device that converts the spherical point cloud of the produced content into hexahedral points and then converts them into points in a tree structure;

a point normal vector estimation device for deriving a normal vector for each point based on a calculation distance and a predetermined reference distance between each point of the tree structure and the points of the child nodes of each point and storing the derived normal vector of the point in a temporary memory; and

One feature includes an encoding device that extracts points to be encoded using an inner product of the normal vector of each point and the direction vector of the user's field of view obtained by a camera, encodes the extracted points, and transmits them.

Preferably, the point normal vector estimation device is:

Each point in the tree structure is set as a reference point, the calculation distance between the point of the child node of each set reference point and the set reference point is output, and the reference points whose output calculation distance is less than the predetermined reference distance are stored in temporary memory. , a point sorting unit that sorts the reference points stored in the temporary memory in ascending order with respect to the calculation distance; and

Extract two adjacent points in the order of the calculated distance from the reference point, derive the normal vector of the plane formed by the two extracted adjacent points and the reference point, and set the derived normal vector to the location of the temporary memory where the reference points are arranged. It may include a normal vector derivation unit stored in .

Preferably, the point normal vector estimation device is

A normal vector optimization unit that extracts at least three points that have a short calculation distance from the reference point stored in the temporary memory, averages the normal vector of each extracted point, and then sets the average normal vector as the normal vector of the reference point. can do.

Preferably, the encoding device:

It may be provided to derive an inner product of the direction vector of the user's field of view obtained through a camera and the normal vector of each point, encode and transmit the points for which the derived inner product is a positive number.

The point cloud compression method for streaming of the present invention according to another embodiment,

A point structuring step of converting the spherical point cloud of the produced content into hexahedral points and then converting them into points in a tree structure;

A point normal vector estimation step of deriving a normal vector of each point based on a calculation distance and a predetermined reference distance between each point of the tree structure and the points of the child nodes of each point and storing the derived normal vector of the point in a temporary memory; and

One feature includes an encoding step of extracting points to be encoded using an inner product of the normal vector of each point and the direction vector of the user's field of view obtained by a camera, encoding the extracted points, and transmitting them.

Preferably, the point normal vector estimation step is,

Setting each point of the tree structure as a reference point and outputting a calculated distance between the point of the child node of each set reference point and the set reference point;

storing reference points whose output calculation distance is less than or equal to a predetermined reference distance in a temporary memory, and sorting the reference points stored in the temporary memory in ascending order with respect to the calculation distance; and

Extract two adjacent points in the order of the calculated distance from the reference point, derive the normal vector of the plane formed by the two extracted adjacent points and the reference point, and set the derived normal vector to the location of the temporary memory where the reference points are arranged. It may include the step of saving.

Preferably, the point normal vector estimation step is,

It may include extracting at least three points that have a short calculation distance from the reference point stored in temporary memory, averaging the normal vectors of each extracted point, and then setting the average normal vector as the normal vector of the reference point. .

Preferably, the encoding step is,

According to one embodiment, a spherical point cloud is converted to a hexahedral point and then converted to a tree structure point, each point of the converted tree structure is set as a reference point, each point of the child node of the set reference point and the above By deriving the normal vectors of each point where the calculation distance between the corresponding reference points is less than or equal to the reference distance, the calculation complexity and speed of the normal vector calculation can be reduced.

In addition, according to the present invention, by performing encoding on points where the inner product of the normal vector of the derived point and the user's viewing direction vector obtained through the camera is a positive number, the number of data to be encoded can be reduced, and the communication resources accordingly and memory resources can be secured.

The following drawings attached to this specification illustrate preferred embodiments of the present invention, and together with the detailed description of the invention described later, serve to further understand the technical idea of the present invention. Therefore, the present invention includes the matters described in such drawings. It should not be interpreted as limited to only .

1 is a configuration diagram of a point cloud compression system for streaming according to an embodiment.

Figure 2 is an exemplary diagram showing the process of the point structuring device of Figure 1.

Figure 3 is a detailed configuration diagram of the point normal vector estimation device of Figure 2.

Figure 4 is an example diagram showing the point arrangement process of the temporary memory of Figure 3.

Figure 5 is a diagram showing the operation process of the point array unit of Figure 3.

Figure 6 is an example diagram showing the point arrangement of the temporary memory based on Figure 5.

Figure 7 is a conceptual diagram of the normal vector of the normal vector derivation part of Figure 3.

Figure 8 is a diagram showing the operation process of the normal vector derivation unit of Figure 3.

Figure 9 is a diagram showing the operation process of the normal vector optimization unit of Figure 4.

Figure 10 is a diagram showing the normal vector for each point optimized based on Figure 9.

Figure 11 is a diagram explaining the concept of extracting points for encoding according to an embodiment.

Hereinafter, embodiments of the present invention will be described in more detail with reference to the drawings.

The advantages and features of the present invention and methods for achieving them will become clear with reference to the embodiments described below along with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms. The present embodiments are merely provided to ensure that the disclosure of the present invention is complete and to be understood by those skilled in the art. It is provided to fully inform those who have the scope of the invention, and the present invention is only defined by the scope of the claims.

The terms used in this specification will be briefly explained, and the present invention will be described in detail.

The terms used in the present invention are general terms that are currently widely used as much as possible while considering the function in the present invention, but this may vary depending on the intention or precedent of a person working in the art, the emergence of new technology, etc. In addition, in certain cases, there are terms arbitrarily selected by the applicant, and in this case, the meaning will be described in detail in the description of the relevant invention. Therefore, the terms used in the present invention should be defined based on the meaning of the term and the overall content of the present invention, rather than simply the name of the term.

When it is said that a part "includes" a certain element throughout the specification, this means that, unless specifically stated to the contrary, it does not exclude other elements but may further include other elements.

Accordingly, the functionality provided within the components and “parts” may be combined into a smaller number of components and “parts” or may be further separated into additional components and “parts”.

Below, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily implement the present invention. In order to clearly explain the present invention in the drawings, parts unrelated to the description are omitted.

An embodiment may include any number of each component to which it applies, in any suitable configuration. In general, computer and communication systems come in a wide range of configurations, and the drawings do not limit the scope of the disclosure to any particular configuration. Although the drawings illustrate one operating environment in which the various features disclosed in this patent document may be used, such features may be used in any other suitable system.

In the description in the specification, the subject performing the operation converts each point of the spherical point cloud into a hexahedral point, structures it in the form of a tree, sets each point of the tree structure as a reference point, and then sets the corresponding reference point and It may be a processor that derives the normal vector of each point through a neighboring point search and normal vector estimation algorithm based on the calculation distance between each point of the child node, which is a child node of the node of the corresponding reference point, and a predetermined reference distance. As another example, As such, it may be a recording medium on which a program that performs estimation and processing is recorded or a device containing the same.

The adjacent point search and normal vector estimation algorithm is sequentially performed for all points in the tree structure based on the tree structure, where the reference point refers to a point set as a reference point and a point set as a reference point among the points in the tree structure. .

FIG. 1 is a diagram showing the configuration of a point cloud compression system for streaming according to an embodiment, FIG. 2 is an example diagram explaining the point structuring process of FIG. 1, FIG. 3 is a conceptual diagram of a normal vector of a point, and FIG. 4 is a detailed configuration diagram of the point normal vector estimation device of FIG. 1, and FIG. 5 is an example diagram showing the point arrangement of the temporary memory of FIG. 4.

Referring to FIGS. 1 to 5, the point cloud compression system for streaming in one embodiment converts each point of a spherical point cloud into a hexahedron and then structures it into a tree form, and combines a reference point of a set tree structure and sub-points of the reference point. Normal vectors are derived for points where the calculation distance between each point of the child node is less than the standard distance, and encoding is performed on the extracted points as the inner product of the derived normal vector and the direction vector of the user's field of view obtained through the camera. The system may include a point structuring device (1), a point normal vector estimation device (2), and an encoding device (3).

The point structuring device 1 maps each point of the point cloud to a cube-shaped hexahedron and then performs the function of structuring each point of the mapped hexahedron into a tree shape.

In one embodiment, the series of processes for mapping each point of a spherical point cloud to each face of a cube-shaped hexahedron is not specifically specified, but this can be understood by those skilled in the art.

The point structuring device 1 structures each point of the hexahedron in a tree form. For example, referring to FIG. 2, the point structuring device 1 configures the first node with points 60, 70, and 60 located at the center of the X-axis of the cube as shown in (a), and then constructs the first node in (b) As shown, the point ((25, 50, 30), (50, 45, 50)) located at the center of the Y axis of each of the two areas divided by the X axis of the hexahedron is the second node as a child of the first node. It constitutes a node (hereinafter referred to as a child node), and as shown in (c), a point ((10, 70, 50) located at the center of the Z axis of each of the four areas divided by the X and Y axes of the hexahedron A child node is formed as a child node of the second node with (40, 25, 55), (65, 70, 60), (80, 30, 60)). The process of structuring these cube-shaped points into a tree shape is repeatedly performed for all points in the spherical point cloud. Here, points (25, 50, 30) represent the point location of the point's X, Y, and Z axis coordinate system.

Each point in the tree structure is transmitted to the point normal vector estimation device (2).

The point normal vector estimation device (2) sequentially sets a reference point at each point of the tree structure, searches for adjacent points based on the calculated distance d, which is the distance difference from the set corresponding reference point, and uses the two searched adjacent points and the reference point. It is equipped with a configuration for deriving the normal vector of a reference point perpendicular to the formed plane.

Accordingly, referring to FIG. 3, the point normal vector estimation device 2 may include a point alignment unit 21, a normal vector derivation unit 22, a temporary memory 23, and a normal vector optimization unit 24. .

Here, the point sorting unit 21 searches for points where the calculation distance between the reference point sequentially set as the point of each node of the point structuring device 1 and the point of the child node of the reference point is less than or equal to the set reference distance, and stores it in the temporary memory 23. Then, the points stored in the temporary memory 23 are sorted in ascending order with respect to the calculation distance.

For example, referring to FIG. 4, when the reference point is set as the point of the first node, the point alignment unit 21 calculates the distance between the point of the child node, which is a child node of the first node, and the reference point. , a point array is created with points whose calculation distance d is less than or equal to a predetermined standard distance.

That is, as shown in (a) of FIG. 4, if the point of point index 32 is set as the corresponding reference point, as shown in (b), the

point indices

2 and 83 of the child nodes that are lower nodes of point 32 The calculated distance d is derived by calculating the distance difference between and the set corresponding reference point 32. At this time, the calculated calculated distance d is 2.1 and 1.5, respectively, which is less than the standard distance r = 3, so the points at

point indices

2 and 83 are stored in temporary memory ( 23) Save it.

Meanwhile, as shown in (c) of FIG. 4, the point alignment unit 21 calculates the distance between the corresponding reference point of point index 2 and each point of

point indexes

93 and 7 of the child nodes of the reference point, and at this time, calculate The distance d is 3.1 and 2.4, respectively.

In addition, the point alignment unit 21 calculates the calculated distance d between the reference point of point index 83 and the reference point to be 1.7 and 2.3, respectively. Therefore,

point indices

45, 17, and 7 that are less than or equal to the reference distance r=3 are stored in the temporary memory 23. This sorting of points in the temporary memory 23 is performed for each point of all nodes.

FIG. 5 is a diagram for explaining the operation of the point arrangement unit 21 of FIG. 3, and FIG. 6 is an exemplary diagram showing the point arrangement of the temporary memory 23 of FIG. 3. Referring to FIGS. 5 and 6, The point arrangement unit 21 performs point arrangement in the temporary memory based on the distance between the reference node sequentially set as each point of the tree structure and each point of the child node, which is a lower node of the node to which the reference node belongs.

That is, referring to FIG. 5, the point array unit 21 sets the point array and reference distance r of the temporary memory formed by the number of points of the predetermined point cloud in

steps

110 and 120, and sets the point array of the set temporary memory. Insert the point of each node.

In this specification, the reference distance r is a result obtained through multiple experiments and can be set to reflect the resolution of the playback content.

Afterwards, in steps 131 to 133, the point array unit 21 calculates the left point A L and the set reference distance between the set reference point of the tree structure and the left point A _L among the child nodes that are lower nodes of the node to which the reference point belongs. Insert point A _L into the point array in temporary memory.

That is, the point array unit 21 inserts the left point into the left position of the temporary memory if the calculation distance is less than or equal to the reference distance (progressing in the “Yes” direction of the arrow).

If the calculation distance in step 132 is not less than the reference distance (proceeding in the “No” direction of the arrow), in steps 134 to 136, the point normal vector estimator 2 determines the right side of the child node of the first node. Based on the calculated distance between points A and _R and the set standard distance, the right point is inserted into the point array of temporary memory.

That is, the point array unit 21 inserts the right point A _R into the right position of the temporary memory when the calculation distance is less than or equal to the reference distance (proceeding in the “Yes” direction of the arrow).

In steps 137 to 139, the point array unit 21 determines whether the last adjacent point and the last node of the point tree structure have been reached, and if the last node has been reached (proceeding in the direction of the arrow “Yes”), it is stored in temporary memory. Sort in ascending order of the calculated distance between the saved point and the reference point.

Additionally, when the point array unit 21 proceeds in the direction of the arrow “No” in step 137, it repeatedly proceeds to step S31. Accordingly, the point array of the temporary memory is stored in the order of points A _L , BL _L , _C _L , D _L , DR , C _R , B _R , and A _R , as shown in FIG. 6 .

Afterwards, the point sorting unit 21 sorts the point array of the temporary memory 23 in order of calculation distance in step 138, and the point array of the temporary memory 23 is sorted in ascending order of calculation distance, Therefore, the temporary memory 23 stores points in order of the shortest calculation distance from the reference point.

Accordingly, in one embodiment, the number of points stored in the temporary memory 23 may be reduced compared to the number of input points to points whose calculation distance is less than the reference distance. Accordingly, the computational complexity of the normal vector for each point, which will be described later, is reduced.

The normal vector derivation unit 22 of FIG. 3 derives a normal vector as a perpendicular vector to a plane consisting of two adjacent points and one reference point retrieved as the first and second points of the temporary memory 23.

FIG. 7 is a conceptual diagram of the normal vector of the normal vector derivation unit of FIG. 3. Referring to FIG. 7, the normal vector is a reference point for a plane formed by an arbitrary reference point A and adjacent points A' and A'' among points in the temporary memory. It is a vertical vector.

At this time, the normal vector can be derived using a CUDA (Compute Unified Device Architecture) parallel processing algorithm, and the normal vector derivation unit 22 sets the CUDA threshold of the CUDA kernel function for each point and calculates the set CUDA threshold and two adjacent points and A normal vector is derived from one reference point.

In one embodiment, the process of deriving a normal vector through a CUDA (Compute Unified Device Architecture) parallel processing algorithm with a set CUDA threshold, two adjacent points, and one reference point as input is not specifically specified in the specification, but this is It can be understood at the level of those skilled in the art.

And the derived normal vector for each point is provided to the temporary memory 23, and each point below the standard distance stored in the temporary memory 23 includes information such as point index, point position coordinates of the hexahedron, and derived normal vector. may be included.

FIG. 8 is a diagram showing the operation process of the normal vector derivation unit of FIG. 4. Referring to FIG. 8, the normal vector derivation unit 22 converts the first point and the second point of the temporary memory into adjacent points in steps 141 to 144. After extraction, a plane is created based on the two extracted adjacent points and the reference point, a normal vector perpendicular to the created plane is derived, and the derived normal vector is inserted into the corresponding reference point in the temporary memory 23. Here, the first and second points of the temporary memory 23 are adjacent points having a short distance from the reference point. Afterwards, the normal vector derivation unit 22 proceeds to step 138.

Meanwhile, in the point cloud compression system for streaming, the normal vector derived using the vertical of the plane is accurate in the flat part of the object of the produced content, but the accuracy deteriorates in the curved part. Accordingly, the point normal vector estimation device 2 derives the average value of the normal vectors of at least three points in the temporary memory and then sets the derived average normal vector as the normal vector of the corresponding reference point. Accordingly, all points stored in the temporary memory 23 are normalized.

Meanwhile, the normal vector optimization unit 24 of FIG. 3 extracts at least 3 adjacent points to the corresponding point in the temporary memory 23, averages the normal vectors of the extracted adjacent points, and calculates the average normal vector as the normal of the reference point. Set as vector.

Therefore, in one embodiment, points whose calculation distance from the reference point is less than or equal to a predetermined reference distance are sorted and stored in ascending order of the calculation distance, making it easy to search for adjacent points and reducing the number of searches for adjacent points. You can.

FIG. 9 is a diagram showing the operation process of the normal vector optimization unit of FIG. 4, and FIG. 10 is a comparison diagram showing the optimized normal vector and the normal vector of FIG. 8. Referring to FIGS. 9 and 10, normal vector optimization of one embodiment is shown. In steps 211 to 212, the unit 24 sets a selection method for adjacent points stored in the temporary memory and extracts adjacent points in the temporary memory according to the set adjacent point selection method. For example, if the adjacent point selection method is a method of extracting the first N points, the first to Nth points are extracted as adjacent points. As another example, if the adjacent point selection method is a point extraction method within a predetermined calculation distance, points within the determined calculation distance for the point are extracted as adjacent points.

The normal vector optimization unit 24 calculates the average of the sail vector of each extracted point in

steps

213 and 214, sets the calculated average normal vector as the normal vector of the corresponding reference point, and stores it in temporary memory. After executing step (214), proceed to step (138).

Accordingly, referring to FIG. 10, it can be seen that the normal vector optimization unit 24 of one embodiment is an optimized normal vector compared to the normal vector derived based on FIG. 8.

In addition, the encoding device 3 of FIG. 1 compresses and transmits the point of the object location that matches the user's field of view obtained through a camera (not shown) using the derived normal vector.

The encoding device 3 encodes the point if the inner product value derived by dot producting the direction vector of the camera and the normal vector of the point of the content object is a positive number, and if the inner product value is a negative number, it determines the point to be out of the user's field of view. .

Figure 11 is an example diagram showing the encoding point for an object of produced content. As shown in Figure 10, it can be seen that the corresponding point present in the user's field of view obtained by a camera (not shown) is encoded. You can.

Therefore, in one embodiment, the computation complexity of the normal vector can be reduced by generating a normal vector of a point whose computation distance from the reference point of the tree structure is less than or equal to the reference distance among the point cloud of the object of the produced content. Encoding speed can be improved and versatility can be improved because it can be applied to lightweight devices.

As described above, although the embodiments have been described with limited examples and drawings, various modifications and variations can be made by those skilled in the art from the above description. For example, the described techniques are performed in a different order than the described method, and/or components of the described system, structure, device, circuit, etc. are combined or combined in a different form than the described method, or other components are used. Alternatively, appropriate results may be achieved even if substituted or substituted by an equivalent. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined by the claims and equivalents of this patent as well as the claims described below.

Claims

A point structuring device that converts the spherical point cloud of the produced content into hexahedral points and then converts them into points in a tree structure;

Each point of each node of the tree structure is sequentially set as a reference point, and each reference point is set based on the calculated distance and the determined reference distance between each set reference point and each point of the child node that is a child node of the node containing each reference point. A point normal vector estimation device that derives the normal vector of a point; and

A point cloud compression system for streaming, comprising an encoding device that extracts points using the inner product of the normal vector of each point and the direction vector of the user's field of view obtained by a camera, encodes the extracted points, and transmits them.
The method of claim 1, wherein the point normal vector estimation device is:

Each point of each node of the tree structure is sequentially set as a reference point, the calculation distance between the reference point and the point of the child node is output, and the points for which the output calculation distance is less than a predetermined standard distance are stored in temporary memory. and a point sorting unit that sorts the points stored in the temporary memory in ascending order with respect to the calculation distance. and

Deriving the normal vector of the reference point in the vertical direction of the plane formed by the adjacent points obtained as the first and second points in the temporary memory for the reference point and the corresponding reference point, and storing the normal vector in the corresponding reference point in the temporary memory. A point cloud compression system for streaming characterized by including wealth.
The method of claim 2, wherein the point normal vector estimation device is:

Characterized in that it further comprises a normal vector optimization unit that extracts at least 3 points as neighboring points of the corresponding reference point, averages the normal vector of each extracted point, and then sets the average normal vector as the normal vector of the reference point. A point cloud compression system for streaming.
The method of claim 3, wherein the encoding device:

A point cloud compression system for streaming, characterized in that it is provided to derive the inner product of the direction vector of the user's field of view acquired through a camera and the normal vector of each point, encode and transmit the points for which the derived inner product is a positive number.
The points of the hexahedron of the produced content in Paragraph 1 are converted into a tree structure, the points of the converted tree structure are sequentially set as reference points, and the distance between the set reference point and each point of the child node of the node containing the reference point is set. a point sorting unit that stores points whose calculation distance is less than or equal to a predetermined reference distance in a temporary memory, and sorts each point in the stored temporary memory in ascending order with respect to the calculation distance; and

Characterized by a normal vector derivation unit for deriving a normal vector in a vertical direction of a plane formed by two adjacent points of the temporary memory and the corresponding reference point and storing the derived normal vectors in corresponding reference points of the temporary memory, respectively. Point normal vector estimation device for point cloud compression system for streaming.
The method of claim 5, wherein the point normal vector estimation device is:

Characterized in that it further comprises a normal vector optimization unit that extracts at least 3 points as neighboring points of the corresponding reference point, averages the normal vector of each extracted point, and then sets the average normal vector as the normal vector of the reference point. Point normal vector estimation device for point cloud compression system for streaming.
In the point cloud compression method for streaming performed according to the streaming point cloud compression system of claim 1,

A point structuring step of converting the spherical point cloud of the produced content into hexahedral points and then converting them into points in a tree structure;

A point that sets each point of the tree structure as a reference point and derives the normal vector of each point based on the calculated distance between the set reference point and each point of the child node that is the child node of the node of the reference point and the determined reference distance. Normal vector estimation step; and

A point cloud compression method for streaming, comprising an encoding step of extracting points using an inner product of the normal vector of each point and the direction vector of the user's field of view obtained by a camera, encoding the extracted points, and transmitting them.
The method of claim 7, wherein the point normal vector estimation step is,

A step of setting each point of the tree structure as a reference point and outputting the calculated distance between the set reference point and each point of the child node;

Storing points in a temporary memory where the output calculation distance is less than or equal to a predetermined reference distance, and sorting the points in the temporary memory in ascending order with respect to the calculation distance; and

The normal vector of the reference point is derived in the vertical direction of the plane formed by the adjacent point and the reference point obtained as the first and second points of the temporary memory for the reference point, and the normal vector of the derived reference point is temporarily A point cloud compression method for streaming, comprising the step of storing corresponding reference points in memory.
The method of claim 8, wherein the point normal vector estimation step is,

Characterized in that it further comprises the step of extracting at least 3 points as neighboring points of the reference point, averaging the normal vectors of the at least 3 extracted points, and then setting the average normal vector as the normal vector of the reference point. Point cloud compression method for streaming.
The method of claim 9, wherein the encoding step,

A point cloud compression method for streaming, comprising deriving the inner product of the direction vector of the user's field of view acquired through a camera and the normal vector of each point, encoding and transmitting the points for which the derived inner product is a positive number.
A recording medium on which the point cloud compression method for streaming according to any one of claims 7 to 10 is recorded and performed by a computer.