WO2021197238A1

WO2021197238A1 - Point cloud attribute prediction method and device, coding method and device, and decoding method and device

Info

Publication number: WO2021197238A1
Application number: PCT/CN2021/083396
Authority: WO
Inventors: 李革; 何盈燊; 王静; 邵薏婷; 高文
Original assignee: 鹏城实验室; 北京大学深圳研究生院
Priority date: 2020-03-30
Filing date: 2021-03-26
Publication date: 2021-10-07
Also published as: CN113473153B; CN113473153A

Abstract

Disclosed are a point cloud attribute prediction method and device, a coding method and device, and a decoding method and device. The point cloud attribute prediction method comprises the steps of: adding an offset value to an original point cloud coordinate to obtain a new coordinate value; determining the offset Morton order according to the new coordinate value; and determining an attribute prediction value of a current node according to the offset Morton order. According to the present invention, an attribute prediction value of a current node is determined by using the offset Morton order to locate the closest neighboring point of the current node in a physical space, and finally, an attribute value of the current node is determined by performing attribute coding according to the attribute prediction value or according to the attribute prediction value of the current node and a point cloud attribute code stream. The point cloud attribute prediction method, the coding method, and the decoding method provided by the present invention can improve the utilization of the correlation between geometric information and attribute information of a point cloud, thereby improving the compressive property of the point cloud attribute.

Description

Point cloud attribute prediction method, encoding method, decoding method and equipment

priority

The PCT patent application requires that the filing date is March 30, 2020, and the application number is 202010239384.8 Chinese patent priority. This patent application combines the technical solutions of the above-mentioned patents.

Technical field

The present disclosure relates to the technical field of point cloud processing, and in particular to a point cloud attribute prediction method, encoding method, decoding method and equipment thereof.

Background technique

Three-dimensional point cloud is an important form of digital representation of the real world. With the rapid development of three-dimensional scanning equipment (such as lasers, radars, etc.), the accuracy and resolution of point clouds have become higher. High-precision point clouds are widely used in the construction of urban digital maps, and play a technical support role in many popular researches such as smart cities, unmanned driving, and cultural relics protection. The point cloud is obtained by sampling the surface of the object by a three-dimensional scanning device. The number of points in a frame of point cloud is generally in the order of one million. Each point contains geometric information, color, reflectivity and other attribute information, and the amount of data is very large. The huge data volume of 3D point cloud brings huge challenges to data storage and transmission, so it is very important to compress the point cloud.

Point cloud compression is mainly divided into geometric compression and attribute compression. At present, the point cloud attribute compression method described in the test platform PCEM provided by the Chinese AVS (Audio Video coding Standard) point cloud compression working group mainly uses the point cloud based on Morton order Prediction method, that is, the current node cloud is Morton sorted according to the position information of the point cloud, the attribute value of the previous point of the current node Morton order is selected as the attribute predicted value of the current node, and finally the actual attribute value of the current node is subtracted from the attribute The predicted value obtains the attribute residual value.

However, the above point cloud prediction method only considers the Morton order. There is a situation that the previous point of the Morton order cannot predict the attribute value of the current node well, which easily leads to low attribute prediction accuracy, which reduces the performance of encoding and decoding. .

Therefore, the existing technology needs to be improved and developed.

Public content

The present disclosure provides a point cloud attribute prediction method, encoding method, decoding method and equipment, and aims to solve the problem that in the prior art, the neighbors found in the attribute encoding of the point cloud are not close enough to affect the attribute prediction value, which leads to the point cloud attribute encoding And the problem of poor decoding performance.

In order to solve the above technical problems, the technical solutions adopted in the present disclosure are as follows:

A point cloud attribute prediction method, which includes the steps:

Add an offset value to the original coordinates of the point cloud to obtain a new coordinate value;

Determining a new sequence according to the new coordinate value;

Determine the attribute prediction value of the current node according to the new order.

In the point cloud attribute prediction method, the adding an offset value to the original coordinates of the point cloud to obtain a new coordinate value includes:

Add an offset value to the original coordinates of the point cloud, and the offset value added by different coordinates is the same or different.

The point cloud attribute prediction method, wherein the determining a new sequence according to the new coordinate value includes the steps:

Generating a Morton code corresponding to the point cloud according to the new coordinate value;

According to the Morton code sorting, the offset Morton order is obtained.

The point cloud attribute prediction method, wherein the determining the attribute prediction value of the current node according to the new order includes the steps:

Determine the prediction point of the current node according to the new sequence;

Determine the attribute predicted value of the current node according to the predicted point.

According to the offset Morton order, the first existing node is searched forward as the first prediction point;

Use the attribute value of the first prediction point as the attribute prediction value of the current node;

Or, searching forward K1 existing nodes as the first prediction point according to the offset Morton order, and finding the point with the smallest distance from the current node at the K1 first prediction points;

The attribute value of the point with the smallest distance is used as the attribute prediction value of the current node, or K1 existing nodes are searched forward according to the offset Morton order as the first prediction point,

The attribute values of the K1 first prediction points are weighted as the attribute prediction values of the current node.

According to the original Morton order under the original coordinates of the point cloud, the first existing node is searched forward as the second predicted point;

Calculating the distance d1 from the first prediction point to the current node;

Calculating the distance d2 from the second prediction point to the current node;

Compare the distance d1 and the distance d2, and select a point with a small distance as the third prediction point;

Using the attribute value of the third prediction point as the attribute prediction value of the current node;

Or, search forward K1 existing nodes as the first prediction point according to the offset Morton order; according to the original point cloud

The original Morton order under the initial coordinates looks forward to K2 existing nodes as the second prediction point;

Calculating the distance from the first prediction point and the second prediction point to the current node, and selecting one or more points with a smaller distance as the third prediction point;

Using the attribute value or attribute weighted value of the third prediction point as the attribute prediction value of the current node;

Or, search forward K1 existing nodes as the first prediction point according to the offset Morton order;

According to the original Morton order under the original coordinates of the point cloud, look forward to K2 existing nodes as the second predicted point;

Calculating a weighted distance d1 from the first prediction point to the current node;

Calculating a weighted distance d2 from the second prediction point to the current node;

The attribute weighted value of the third predicted point is used as the attribute predicted value of the current node.

In the point cloud attribute prediction method, the step of calculating the distance from the first prediction point to the current node includes:

Calculating the Euclidean distance between the current node and the first prediction point as the distance between the current node and the first prediction point;

Or, calculating the maximum value of the absolute value of the coordinate difference between the current node and the first prediction point in the three directions of X, Y, and Z as the distance between the current node and the first prediction point;

Alternatively, the sum of the difference between the coordinates of the current node and the first prediction point in the X, Y, and Z directions is calculated as the distance between the current node and the first prediction point.

A point cloud attribute prediction device, which includes a processor, a memory, and a communication bus; the memory stores a computer readable program that can be executed by the processor;

The communication bus realizes connection and communication between the processor and the memory;

When the processor executes the computer-readable program, the steps in the point cloud attribute prediction method as described in the present disclosure are implemented.

A point cloud attribute coding method, which includes the steps:

Using the point cloud attribute prediction method of the present disclosure to determine the attribute prediction value of the current node;

Determine the predicted residual value of the current node according to the difference between the attribute value of the current node and the attribute predicted value of the current node;

The prediction residual value of the current node is coded to obtain a point cloud code stream.

A point cloud attribute encoding device, which includes a processor, a memory, and a communication bus; the memory stores a computer readable program that can be executed by the processor;

When the processor executes the computer-readable program, the steps in the point cloud attribute encoding method as described in the present disclosure are implemented.

A point cloud attribute decoding method, which includes the steps:

Decode the point cloud attribute code stream to obtain the attribute residual value of the current node;

The attribute value of the current node is determined according to the sum of the attribute prediction value of the current node and the attribute residual value.

A point cloud attribute decoding device, which includes a processor, a memory, and a communication bus; the memory stores a computer readable program that can be executed by the processor;

When the processor executes the computer-readable program, the steps in the point cloud attribute decoding method as described in the present disclosure are implemented.

Beneficial effects: Compared with the prior art, the present disclosure uses the offset Morton order to find the nearest neighbor point to the current node in the physical space, thereby determining the attribute prediction value of the current node, and finally performs the process according to the attribute prediction value. The attribute coding or the attribute value of the current node is determined according to the attribute prediction value of the current node and the point cloud attribute code stream. The point cloud attribute prediction method, encoding method, and decoding method provided by the present disclosure can improve the utilization of the geometric information and the attribute information correlation of the point cloud, thereby improving the compression performance of the point cloud attribute.

Description of the drawings

FIG. 1 is a schematic diagram of the original Morton sequence of nodes in a point cloud under an octree in an embodiment of the disclosure.

Fig. 2 is a flowchart of a preferred embodiment of a point cloud attribute prediction method provided by the present disclosure.

FIG. 3 is a schematic diagram of the Morton sequence of offsetting nodes in the point cloud under the octree in an embodiment of the disclosure.

4 is a schematic diagram of the original Morton sequence of nodes in the point cloud under the quadtree in the embodiment of the disclosure.

FIG. 5 is a schematic diagram of the Morton sequence of offsetting nodes in a point cloud under a quadtree in an embodiment of the disclosure.

FIG. 6 is a flowchart of a preferred embodiment of a point cloud attribute encoding method provided by the present disclosure.

Fig. 7 is a schematic structural diagram of a point cloud attribute encoding device provided by the present disclosure.

FIG. 8 is a flowchart of a preferred embodiment of a point cloud attribute decoding method provided by the present disclosure.

Detailed ways

The present disclosure provides a point cloud attribute prediction method, encoding method, decoding method, and equipment. In order to make the objectives, technical solutions, and effects of the present disclosure clearer and clearer, the present disclosure will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present disclosure, and not used to limit the present disclosure.

Those skilled in the art can understand that, unless specifically stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present disclosure refers to the presence of the described features, integers, steps, operations, elements and/or components, but does not exclude the presence or addition of one or more other features, Integers, steps, operations, elements, components, and/or groups of them. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element, or intervening elements may also be present. In addition, “connected” or “coupled” used herein may include wireless connection or wireless coupling. The term "and/or" as used herein includes all or any unit and all combinations of one or more associated listed items.

Those skilled in the art can understand that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meanings as those commonly understood by those of ordinary skill in the art to which this disclosure belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have a meaning consistent with the meaning in the context of the prior art, and unless specifically defined as here, they will not be idealized or overly Explain the formal meaning.

In the following, the disclosure will be further explained through the description of the embodiments in conjunction with the accompanying drawings.

The present invention provides a point cloud attribute prediction method, which includes the steps of: adding an offset value to the original coordinates of the point cloud to obtain a new coordinate value; determining a new order according to the new coordinate value; The sequence determines the predicted value of the attributes of the current node.

In this embodiment, an offset value is added to the original coordinates of the point cloud, and the offset value added by different coordinates may be the same or different. When the new order is obtained, the prediction point of the current node is determined according to the new order, and finally the attribute prediction value of the current node is determined according to the prediction point.

In some embodiments, the determining a new order according to the new coordinate value includes the steps of: generating a Morton code corresponding to the point cloud according to the new coordinate value; Pause order.

Specifically, the Morton order after Morton sorting of each point in the point cloud can be specifically expressed as: the position coordinates (Xk, Yk, Zk) of the k-th point are expressed as:

Then the Morton code corresponding to the k-th point is expressed as:

Or use octal numbers to represent every three bits

Then the Morton code corresponding to the k-th point can be expressed as:

Figure 1 is a partial schematic diagram of a three-dimensional point cloud. The numbers represent the Morton order in the three-dimensional point cloud. It is assumed that there are points in the

Morton order

0, 2, 10, 16, and the three points are numbered with letters for the convenience of expression. It is A, B, C, D. In the attribute coding of the PCEM encoder, the method for searching the predicted point of the point cloud according to the Morton order is to find the first point in the Morton order as the current predicted point. Assuming that the current node is D(16), according to the PCEM search method, the first point C(10) in the Morton order is found as the predicted point of D(16), so there will be a problem, B(2) is The point closer to D(16) but the predicted point of D(16) is farther C(10). Using the attribute value of C(10) to predict D(16) will affect the point cloud compression performance.

Based on the problems of the above-mentioned technology, the present disclosure provides a flow chart of a preferred embodiment of a point cloud attribute prediction method, as shown in FIG. 2, which includes the steps:

S1. Add an offset value to the original coordinates of the point cloud to obtain a new coordinate value;

S2. Determine the offset Morton order according to the new coordinate value;

S3. Determine the attribute prediction value of the current node according to the offset Morton order.

Specifically, this embodiment adds a fixed offset value (j1, j2, j3) to the original point cloud coordinates (x, y, z), and uses the new coordinates (x+j1, y+j2, z). +j3) Generate the Morton code corresponding to the point cloud, and obtain the offset Morton order according to the Morton order, and the value of the offset values j1, j2, and j3 is greater than or equal to 1. This embodiment uses the offset Morton order to find the physically closest neighbor to the current node, thereby determining the attribute prediction value of the current node, and finally performing attribute encoding according to the attribute prediction value or according to the current node The attribute prediction value of and the point cloud attribute code stream determine the attribute value of the current node, thereby improving the coding and decoding performance of the point cloud attribute.

In some specific implementations, the offset value (1, 1, 1) is added to the 0, 2, 10, and 16 points in the origin point cloud to obtain 7, 21, 29, and 29 in the new offset point cloud. 23 points, as shown in Figure 3, that is, A, B, C, and D correspond to the points numbered 0, 2, 10, 16 in the original Morton order, and correspond to the points numbered 7, 21, 29 in the new Morton order. , 23, this new Morton order is called the offset Morton order.

In some embodiments, the first existing node is searched forward as the first prediction point according to the offset Morton order; the attribute value of the first prediction point is used as the attribute prediction value of the current node.

As an example, according to the offset Morton order to find the first existing point as the predicted point, in the offset Morton order shown in Figure 3, the current node D (23) looks for the offset Morton order The first point is used as the predicted point, the predicted point found is B(21), and B(21) is used as the predicted point of the current node D(23).

In some embodiments, K1 existing nodes are searched forward as the first prediction point according to the offset Morton order, and the point with the smallest distance from the current node is found at the K1 first prediction points; The attribute value of the point is used as the attribute predicted value of the current node.

As an example, suppose K1 is set to 2. In the offset Morton order shown in Figure 3, the current node D (23) looks for the first 2 points in the offset Morton order as prediction points, and the predicted point found is B (21), A(7), the first-order distance of the current nodes D(23) and B(21) is 1, and the first-order distance of the current nodes D(23) and A(7) is 2, choose the smaller distance B(21) is used as the prediction point of the current node D(23).

In some embodiments, K1 existing nodes are searched forward as the first prediction point according to the offset Morton order, and the attribute values of the K1 first prediction points are weighted as the attribute prediction value of the current node.

As an example, suppose K1 is set to 2. In the offset Morton order shown in Figure 3, the current node D (23) looks for the first 2 points in the offset Morton order as prediction points, and the predicted point found is B (21), A(7), use the average of the attribute values of B(21) and A(7) as the predicted value of the current node D(23).

In some embodiments, the first existing node is searched forward according to the offset Morton order as the first prediction point; the first existing node is searched forward according to the original Morton order under the original coordinates of the point cloud as the second prediction point. Prediction point; calculate the distance d1 from the first prediction point to the current node; calculate the distance d2 from the second prediction point to the current node; compare the distance d1 and the distance d2, and select the smaller distance Point as the third predicted point; use the attribute value of the third predicted point as the attribute predicted value of the current node.

As an example, look for the first existing node of the current node D(16) in the original Morton order as shown in Figure 1. The found point is C(10), and C(10) is used as the predicted point 1. The order distance is 3. Find the first existing point of the current node D (23) in the offset Morton order shown in Figure 3, and the current node D (23) finds the first point in the Morton order as the predicted point, and the predicted point found It is B(21), using B(21) as the predicted point 2, and the first-order distance is 1. Compare the distance between predicted point 1 and predicted point 2 to the current node, select predicted point 2 with a smaller distance as the predicted point of current node D (23), and use the attribute value of predicted point 2 to perform attribute prediction on the current node.

In some embodiments, K1 existing nodes are searched forward as the first prediction point according to the offset Morton order; K2 existing nodes are searched forward as the second prediction point according to the original Morton order under the original coordinates of the point cloud Calculate the distance between the first prediction point and the second prediction point to the current node, and select the point with a small distance as the third prediction point; use the attribute value of the third prediction point as the current node The predicted value of the attribute.

As an example, find K1 existing points forward in the offset Morton order as shown in Figure 3, find K2 existing points forward in the original Morton order as shown in Figure 1, and choose from K1+K2 and The point with the smallest distance from the current point is used as the predicted point. In this example, set K1 to 2 and K2 to 2. In the original Morton order, the current point D(16) looks for the first 2 points in the original Morton order as prediction points, and the predicted points found are C(10), B (2). In the offset Morton order, the current point D(23) looks for the first 2 points in the offset Morton order as the predicted points, and the predicted points found are B(21) and A(7). The set of points found in the two Morton sequences is A, B, and C. The first-order distances from the current point D are 2, 1, and 4 respectively. The B with the smallest distance is selected as the predicted point of the current point D.

In some embodiments, K1 existing nodes are searched forward as the first prediction point according to the offset Morton order; K2 existing nodes are searched forward as the second prediction point according to the original Morton order under the original coordinates of the point cloud Calculate the weighted distance d1 from the first predicted point to the current node; calculate the weighted distance d2 from the second predicted point to the current node; compare the distance d1 and the distance d2, and select the smaller distance Point as the third predicted point; use the attribute weighted value of the third predicted point as the attribute predicted value of the current node.

As an example, find K1 existing points forward in the offset Morton order shown in Figure 3, find K2 existing points forward in the original Morton order shown in Figure 1, and calculate K1 points to The weighted distance of the current point and the weighted distance of K2 points to the current point, from which K1 points or K2 points with a smaller distance are selected as predicted points, and their attribute weighted values are used as predicted values. In this example, K1 is set to 2, K2 is set to 2, and the distance weighting method is the mean value. In the original Morton order, the current point D(16) finds the first 2 points in the original Morton order as the predicted points. The predicted points found are C(10), B(2), one distance from the current point D(16) The order distance is 4, 1, and the average value of the distance is 2.5. In the offset Morton order, the current point D(23) looks for the first 2 points in the offset Morton order as the predicted points, and the predicted points found are B(21), A(7) and the current point D(16) The first-order distance of is 1, 2, and the average value of the distance is 1.5. Select B(21) and A(7) in the offset Morton order with the smaller weighting distance as the predicted point, and the attribute weighted value of B(21) and A(7) as the predicted value.

In some embodiments, K1 existing points are found forward in the offset Morton order, K2 existing points are found forward in the original Morton order, and the KX points with the closest distance are selected from the K1+K2 points as For predicted points, KX attribute weighted values are used as predicted values. In this example, K1 is set to 2, K2 is set to 2, KX is set to 2, and the distance weighting method is average. In the original Morton order, the current point D(23) finds the first 2 points in the original Morton order as prediction points, and the predicted points found are C(10) and B(2). In the offset Morton order, the current point D(23) looks for the first 2 points in the offset Morton order as the predicted points, and the predicted points found are B(21) and A(7). The collection of the points found in the two Morton sequences is A, B, C, and the first-order distances from the current point D are 2, 1, and 4 respectively. The two points A and B with the smaller first-order distance are selected as the current The predicted point of point D, the attribute weighted value of A and B are used as the predicted value.

In some embodiments, the step of calculating the distance between the first prediction point and the current node includes: calculating the Euclidean distance between the current node and the first prediction point as the current node and the first prediction point. The distance between a predicted point; or, calculate the maximum absolute value of the coordinate difference between the current node and the first predicted point in the X, Y, and Z directions as the current node and the first predicted The distance of the point; or, the sum of the difference between the coordinates of the current node and the first predicted point in the X, Y, and Z directions is calculated as the distance between the current node and the first predicted point.

In some embodiments, in the case of a quadtree, in the original Morton sequence as shown in FIG. 4, the point cloud has a total of 5 points, respectively

Morton numbers

8, 14, 19, 24, 25, in order to It is convenient to express these points as A, B, C, D, E. In the original Morton order, the previous Morton order prediction point found at the current point E(25) is D(24), the first-order distance between D(24) and the current point E(25) is 4, D(24) For the first prediction point. Add the coordinates of the point in Figure 4 to the offset (1, 1) to obtain the point in Figure 5, which forms the offset Morton sequence. A, B, C, D, and E correspond to the offset Morton sequence in Figure 5 The points numbered 25, 27, 26, 30, 28 are then searched for the predicted point of the current point E(28) from the offset Morton order, and then B(27), C( 26). The previous Morton order prediction point found by A(25) is B(27), the first-order distance from the current point is 1, and B(27) is used as the second prediction point. According to the principle of finding the nearest predicted point, select the nearest B(27) as the predicted point of the current point, which is closer than the predicted point D(24) found in the original Morton order, so point B is selected as the current point E Point of prediction.

In some embodiments, a point cloud attribute prediction device is also provided, which includes a processor, a memory, and a communication bus; the memory stores a computer readable program that can be executed by the processor; the communication bus The connection and communication between the processor and the memory are realized; when the processor executes the computer-readable program, the steps in the point cloud attribute prediction method as described in the present disclosure are realized.

In some embodiments, a point cloud attribute coding method is also provided, as shown in FIG. 6, which includes the steps:

S10. Using the point cloud attribute prediction method to determine the attribute prediction value of the current node;

S20: Determine the predicted residual value of the current node according to the difference between the attribute value of the current node and the predicted attribute value of the current node;

S30. Encode the prediction residual value of the current node to obtain a point cloud code stream.

In this embodiment, an offset Morton order is used to find the physically nearest neighbor to the current node, thereby determining the attribute prediction value of the current node, and finally performing attribute encoding according to the attribute prediction value. The point cloud attribute coding method provided in this embodiment can improve the utilization of the geometric information and the attribute information correlation of the point cloud, thereby improving the coding and decoding performance of the point cloud attribute.

Based on the above point cloud attribute encoding method, the present disclosure also provides a point cloud attribute encoding device, where, as shown in FIG. 7, it includes at least one processor (processor) 20; a display screen 21; and a memory (memory) 22, It may also include a communication interface (Communications Interface) 23 and a bus 24. Among them, the processor 20, the display screen 21, the memory 22, and the communication interface 23 can communicate with each other through the bus 24. The display screen 21 is set to display a user guide interface preset in the initial setting mode. The communication interface 23 can transmit information. The processor 20 can call the logic instructions in the memory 22 to execute the method in the foregoing embodiment.

In addition, the aforementioned logic instructions in the memory 22 can be implemented in the form of software functional units and when sold or used as independent products, they can be stored in a computer readable storage medium.

As a computer-readable storage medium, the memory 22 can be configured to store software programs and computer-executable programs, such as program instructions or modules corresponding to the methods in the embodiments of the present disclosure. The processor 20 executes functional applications and data processing by running software programs, instructions, or modules stored in the memory 22, that is, implements the methods in the foregoing embodiments.

The memory 22 may include a program storage area and a data storage area, where the program storage area may store an operating system and an application program required by at least one function; the data storage area may store data created according to the use of the electronic device, and the like. In addition, the memory 22 may include a high-speed random access memory, and may also include a non-volatile memory. For example, U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or CD-ROM and other media that can store program code, or temporary State storage medium.

In addition, the specific process of loading and executing the multiple instruction processors in the foregoing storage medium and the point cloud attribute encoding device has been described in detail in the foregoing method, and will not be described here.

In some embodiments, a point cloud attribute decoding method is also provided, as shown in FIG. 8, which includes the steps:

S100. Decode the point cloud attribute code stream to obtain the attribute residual value of the current node;

S200. Use the point cloud attribute prediction method described in the present disclosure to determine the attribute prediction value of the current node;

S300. Determine the attribute value of the current node according to the sum of the attribute prediction value of the current node and the attribute residual value.

This embodiment uses the offset Morton order to find the physically closest neighbor to the current node, thereby determining the attribute predicted value of the current node, and finally determines the location based on the attribute predicted value of the current node and the point cloud attribute code stream. Describe the attribute value of the current node. The point cloud attribute decoding method provided by the present disclosure can improve the utilization of the geometric information of the point cloud and the correlation of the attribute information, thereby improving the point cloud attribute decoding performance.

Based on the foregoing point cloud attribute decoding method, the present disclosure also provides a point cloud attribute decoding device, which includes a processor, a memory, and a communication bus; the memory stores a computer readable program that can be executed by the processor; The communication bus realizes the connection and communication between the processor and the memory; when the processor executes the computer-readable program, the steps in the point cloud attribute decoding method as described in the present disclosure are realized.

In some embodiments, the point cloud attribute encoding method and decoding method provided in the present disclosure are used to compare the results obtained by point cloud compression with the benchmark results of the test platform PCEM. The shift amount is (3, 3, 3), the search range is N=8, M=8:

Table 1 is a comparison table of rate-distortion data of brightness, chromaticity and reflectance under the condition of finite loss geometry and lossy attributes

Table 2 is a comparison table of rate-distortion data of brightness, chromaticity, and reflectivity under lossless geometry and lossy attribute conditions

Table 3 is a comparison table of rate-distortion data of luminance, chromaticity, and reflectivity under the condition of lossless geometry and limited loss properties

Table 4 is a comparison table of rate-distortion data of luminance, chromaticity, and reflectivity under the condition of lossless geometry and lossless attributes

From the data in Table 1 to Table 4, it can be seen that compared with the benchmark results of the test platform PCEM, for the reflectivity properties, in the condition of finite loss geometry, lossy attribute, lossless geometry, lossy attribute condition, lossless geometry, finite Under the condition of loss attribute, the end-to-end rate distortion of the present disclosure saves 8.5%, 5.0%, and 5.5% respectively; for the brightness attribute, the code stream size is 94.7% of the original under the condition of lossless geometry and lossless attribute. , Lossy attribute conditions, lossless geometry, lossy attribute conditions, lossless geometry, finite loss attribute conditions, the end-to-end rate distortion of the present disclosure saves 4.9%, 3.2%, and 6.9%, respectively, in the lossless geometry, lossless attribute conditions The size of the lower code stream is 90% of the original; for the chrominance attribute, the end-to-end rate distortion of the present disclosure is under the condition of finite loss geometry, lossy attribute condition, lossless geometry, lossy attribute condition, lossless geometry, finite loss attribute condition The highest savings were 5.2%, 4.0%, and 6.9%.

In summary, the present disclosure uses the offset Morton order to find the nearest neighbor point to the current node in the physical space, thereby determining the attribute prediction value of the current node, and finally performing attribute encoding or according to the attribute prediction value. The attribute prediction value of the current node and the point cloud attribute code stream determine the attribute value of the current node. The point cloud attribute prediction method, encoding method, and decoding method provided by the present disclosure can improve the utilization of the geometric information and the attribute information correlation of the point cloud, thereby improving the compression performance of the point cloud attribute.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure, not to limit them; although the present disclosure has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: The technical solutions recorded in the foregoing embodiments are modified, or some of the technical features are equivalently replaced; these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

A point cloud attribute prediction method is characterized in that it comprises the steps:

Add an offset value to the original coordinates of the point cloud to obtain a new coordinate value;

Determining a new sequence according to the new coordinate value;

Determine the attribute prediction value of the current node according to the new order.
The point cloud attribute prediction method according to claim 1, wherein the adding an offset value to the original coordinates of the point cloud to obtain a new coordinate value comprises:

Add an offset value to the original coordinates of the point cloud, and the offset value added by different coordinates is the same or different.
The point cloud attribute prediction method according to claim 1, wherein the determining a new order according to the new coordinate value comprises the steps of:

Generating a Morton code corresponding to the point cloud according to the new coordinate value;

According to the Morton code sorting, the offset Morton order is obtained.
The point cloud attribute prediction method according to claim 1, wherein the determining the attribute prediction value of the current node according to the new sequence comprises the steps of:

Determine the prediction point of the current node according to the new sequence;

Determine the attribute predicted value of the current node according to the predicted point.
The point cloud attribute prediction method according to claim 1 or 3, wherein the determining the attribute prediction value of the current node according to the new order comprises the steps of:

According to the offset Morton order, the first existing node is searched forward as the first prediction point;

Use the attribute value of the first prediction point as the attribute prediction value of the current node;

Or, searching forward K1 existing nodes as the first prediction point according to the offset Morton order, and finding the point with the smallest distance from the current node at the K1 first prediction points;

The attribute value of the point with the smallest distance is used as the attribute prediction value of the current node, or K1 existing nodes are searched forward according to the offset Morton order as the first prediction point,

The attribute values of the K1 first prediction points are weighted as the attribute prediction values of the current node.
The point cloud attribute prediction method according to claim 1 or 3, wherein the determining the attribute prediction value of the current node according to the new order comprises the steps of:

According to the offset Morton order, the first existing node is searched forward as the first prediction point;

According to the original Morton order under the original coordinates of the point cloud, the first existing node is searched forward as the second predicted point;

Calculating the distance d1 from the first prediction point to the current node;

Calculating the distance d2 from the second prediction point to the current node;

Compare the distance d1 and the distance d2, and select a point with a small distance as the third prediction point;

Using the attribute value of the third prediction point as the attribute prediction value of the current node;

Or, search forward K1 existing nodes as the first prediction point according to the offset Morton order; search forward K2 existing nodes as the second prediction point according to the original Morton order under the original coordinates of the point cloud;

Calculating the distance from the first prediction point and the second prediction point to the current node, and selecting one or more points with a smaller distance as the third prediction point;

Using the attribute value or attribute weighted value of the third prediction point as the attribute prediction value of the current node;

Or, search forward K1 existing nodes as the first prediction point according to the offset Morton order;

According to the original Morton order under the original coordinates of the point cloud, look forward to K2 existing nodes as the second predicted point;

Calculating a weighted distance d1 from the first prediction point to the current node;

Calculating a weighted distance d2 from the second prediction point to the current node;

Compare the distance d1 and the distance d2, and select a point with a small distance as the third prediction point;

The attribute weighted value of the third predicted point is used as the attribute predicted value of the current node.
The point cloud attribute prediction method according to claim 6, wherein the step of calculating the distance from the first prediction point to the current node comprises:

Calculating the Euclidean distance between the current node and the first prediction point as the distance between the current node and the first prediction point;

Or, calculating the maximum value of the absolute value of the coordinate difference between the current node and the first prediction point in the three directions of X, Y, and Z as the distance between the current node and the first prediction point;

Alternatively, the sum of the difference between the coordinates of the current node and the first prediction point in the X, Y, and Z directions is calculated as the distance between the current node and the first prediction point.
A point cloud attribute prediction device, which is characterized by comprising a processor, a memory, and a communication bus; the memory stores a computer readable program that can be executed by the processor;

The communication bus realizes connection and communication between the processor and the memory;

When the processor executes the computer-readable program, the steps in the point cloud attribute prediction method according to any one of claims 1-7 are realized.
A point cloud attribute coding method, which is characterized in that it comprises the steps:

Using the point cloud attribute prediction method of any one of claims 1-7 to determine the attribute prediction value of the current node;

Determine the predicted residual value of the current node according to the difference between the attribute value of the current node and the attribute predicted value of the current node;

The prediction residual value of the current node is coded to obtain a point cloud code stream.
A point cloud attribute encoding device, which is characterized by comprising a processor, a memory, and a communication bus; the memory stores a computer readable program that can be executed by the processor;

The communication bus realizes connection and communication between the processor and the memory;

When the processor executes the computer-readable program, the steps in the point cloud attribute encoding method according to claim 9 are implemented.
A point cloud attribute decoding method, which is characterized in that it comprises the steps:

Decode the point cloud attribute code stream to obtain the attribute residual value of the current node;

Using the point cloud attribute prediction method of any one of claims 1-7 to determine the attribute prediction value of the current node;

The attribute value of the current node is determined according to the sum of the attribute prediction value of the current node and the attribute residual value.
A point cloud attribute decoding device, characterized by comprising a processor, a memory, and a communication bus; the memory stores a computer readable program that can be executed by the processor;

The communication bus realizes connection and communication between the processor and the memory;

When the processor executes the computer-readable program, the steps in the point cloud attribute decoding method according to claim 11 are implemented.