WO2023169007A1 - 点云预测处理方法、装置、计算机、存储介质 - Google Patents

点云预测处理方法、装置、计算机、存储介质 Download PDF

Info

Publication number
WO2023169007A1
WO2023169007A1 PCT/CN2022/135899 CN2022135899W WO2023169007A1 WO 2023169007 A1 WO2023169007 A1 WO 2023169007A1 CN 2022135899 W CN2022135899 W CN 2022135899W WO 2023169007 A1 WO2023169007 A1 WO 2023169007A1
Authority
WO
WIPO (PCT)
Prior art keywords
point cloud
point
target
group
points
Prior art date
Application number
PCT/CN2022/135899
Other languages
English (en)
French (fr)
Inventor
朱文婕
Original Assignee
腾讯科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 腾讯科技（深圳）有限公司 filed Critical 腾讯科技（深圳）有限公司
Priority to US18/236,279 priority Critical patent/US20230394711A1/en
Publication of WO2023169007A1 publication Critical patent/WO2023169007A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T9/00—Image coding
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/103—Selection of coding mode or of prediction mode
    • H04N19/105—Selection of the reference unit for prediction within a chosen coding or prediction mode, e.g. adaptive choice of position and number of pixels used for prediction
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/124—Quantisation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/48—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using compressed domain processing techniques other than decoding, e.g. modification of transform coefficients, variable length coding [VLC] data or run-length data
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding

Definitions

  • the present application relates to the field of computer technology, and in particular, to a point cloud prediction processing method, device, computer, and storage medium.
  • point cloud data types can be divided into point cloud coding based on geometric structures and point cloud coding based on projection.
  • attribute prediction is performed on the point cloud.
  • the initial block size is determined.
  • the spatial structure of the current point is determined, and then in this spatial structure Obtain the neighbor points of the current point and perform attribute prediction. That is to say, the acquisition of the neighbor point is based on the current point (which can be called a local point), which makes the acquisition of neighbor points more limited, thus reducing the efficiency of point cloud attribute prediction. performance.
  • Embodiments of the present application provide a point cloud prediction processing method, device, computer, and storage medium.
  • embodiments of the present application provide a point cloud prediction processing method, which method includes:
  • the candidate point set of the target point cloud group where the target point cloud point is located belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; the point cloud points included in each point cloud group
  • the coordinate codewords are the same after moving through the group movement bits corresponding to the point cloud group; M is a positive integer;
  • Prediction processing is performed on the target point cloud points based on the prediction reference points to obtain target attribute prediction values of the target point cloud points.
  • embodiments of the present application provide a point cloud prediction processing method, which method includes:
  • the candidate point set of the target point cloud group where the target point cloud point is located belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; the point cloud points included in each point cloud group
  • the coordinate codewords are the same after moving through the group movement bits corresponding to the point cloud group; M is a positive integer;
  • Obtain the code stream corresponding to the target point cloud point decode the code stream corresponding to the target point cloud point, and obtain the target attribute residual of the target point cloud point. Based on the target attribute prediction value and the target attribute residual, determine the target point The target attribute reconstruction value of the cloud point.
  • embodiments of the present application provide a point cloud prediction processing device, which includes:
  • the alternative set acquisition module is used to obtain the alternative point set of the target point cloud group where the target point cloud point is located; the alternative point set belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; each point The coordinate codes of the point cloud points included in the cloud group are the same after being moved by the group movement bits corresponding to the point cloud group; M is a positive integer;
  • the reference point acquisition module is used to obtain the predicted reference point associated with the target point cloud point from the candidate point set;
  • the attribute prediction module is used to predict the target point cloud points based on the prediction reference points, and obtain the target attribute prediction value of the target point cloud points.
  • embodiments of the present application provide a point cloud prediction processing device, which includes:
  • the alternative set acquisition module is used to obtain the alternative point set of the target point cloud group where the target point cloud point is located; the alternative point set belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; each point The coordinate codes of the point cloud points included in the cloud group are the same after being moved by the group movement bits corresponding to the point cloud group; M is a positive integer;
  • the reference point acquisition module is used to obtain the predicted reference point associated with the target point cloud point from the candidate point set;
  • the attribute prediction module is used to predict the target point cloud points based on the prediction reference points, and obtain the target attribute prediction value of the target point cloud points;
  • the code stream acquisition module is used to obtain the code stream corresponding to the target point cloud point
  • the code stream decoding module is used to decode the code stream corresponding to the target point cloud point and obtain the target attribute residual of the target point cloud point;
  • the attribute reconstruction module is used to determine the target attribute reconstruction value of the target point cloud point based on the target attribute prediction value and the target attribute residual.
  • embodiments of the present application provide a computer device, including one or more processors, memories, and input and output interfaces;
  • the processor is connected to the memory and the input/output interface respectively, where the input/output interface is used to receive data and output data, the memory is used to store computer readable instructions, and the processor is used to call the computer readable instructions so that the computer including the processor
  • the computer device executes the point cloud prediction processing method in one aspect of the embodiment of the present application.
  • embodiments of the present application provide one or more computer-readable storage media.
  • the computer-readable storage media stores computer-readable instructions.
  • the computer-readable instructions are adapted to be loaded and executed by a processor, so that the processor has
  • the computer device executes the point cloud prediction processing method in one aspect of the embodiment of the present application.
  • embodiments of the present application provide a computer program product.
  • the computer program product includes computer-readable instructions, and the computer-readable instructions are stored in one or more computer-readable storage media.
  • One or more processors of the computer device read the computer-readable instructions from the computer-readable storage medium, and the processor executes the computer-readable instructions, causing the computer device to perform various optional methods in one aspect of the embodiments of the present application.
  • methods provided in In other words, when the computer readable instructions are executed by the processor, the methods provided in various optional ways in one aspect of the embodiments of the present application are implemented.
  • Figure 1 is a network interaction architecture diagram of point cloud prediction processing provided by an embodiment of the present application
  • Figure 2 is a schematic diagram of a point cloud prediction processing scenario provided by an embodiment of the present application.
  • Figure 3 is a flow chart of a method for point cloud prediction processing in the encoding process provided by an embodiment of the present application
  • Figure 4 is a schematic diagram of point cloud point distribution provided by an embodiment of the present application.
  • Figure 5 is an optional point cloud point grouping process provided by the embodiment of the present application.
  • Figure 6 is a flow chart of a method for point cloud prediction processing in the decoding process provided by an embodiment of the present application
  • Figure 7 is a data interaction architecture diagram provided by an embodiment of the present application.
  • Figure 8 is a schematic diagram of a point cloud prediction processing device provided by an embodiment of the present application.
  • Figure 9 is a schematic diagram of a point cloud prediction processing device provided by an embodiment of the present application.
  • Figure 10 is a schematic structural diagram of a computer device provided by an embodiment of the present application.
  • Big data refers to a collection of data that cannot be captured, managed and processed with conventional software tools within a certain time range. It requires a new processing model to have stronger decision-making power. Massive, high-growth and diverse information assets with insight discovery and process optimization capabilities. With the advent of the cloud era, big data has also attracted more and more attention. Big data requires special technologies to effectively handle large amounts of data within a tolerable time. Technologies applicable to big data include massively parallel processing databases, data mining, distributed file systems, distributed databases, cloud computing platforms, the Internet, and scalable storage systems. For example, big data processing technology and data computing technology in the field of big data can be used to group, predict, encode and decode point cloud points to improve the efficiency of data processing.
  • Figure 1 is a network interaction architecture diagram of point cloud prediction processing provided by this embodiment of the present application.
  • the computer device 101 can obtain the point cloud points that need to be encoded from the computer device 101, and encode the obtained point cloud points; or, obtain the code stream that needs to be decoded from the computer device 101, and encode the obtained code stream. stream to decode.
  • the computer device 101 can obtain the point cloud points that need to be encoded from other associated devices, and encode the obtained point cloud points; or, obtain the code stream that needs to be decoded from the associated device, and encode the obtained code stream. Decode etc.
  • the number of the associated devices is one or at least two. For example, taking the number 3 in Figure 1 as an example, such as the associated device 102a, the associated device 102b, or the associated device 102c.
  • Figure 2 is a schematic diagram of a point cloud prediction processing scenario provided by an embodiment of the present application.
  • the computer device can acquire at least two point cloud points 201, group the at least two point cloud points 201 based on the coordinate code words corresponding to the at least two point cloud points 201, and obtain M point cloud groups.
  • M is a positive integer.
  • M is a positive integer greater than or equal to 3, such as point cloud group 1, point cloud group 2, point cloud group M, etc.
  • the computer device can obtain the target point cloud group in which the target point cloud point is located.
  • the computer device can obtain the target point cloud from the candidate point set.
  • the predicted reference point associated with the point can perform prediction processing on the target point cloud points based on the prediction reference points to obtain target attribute prediction values of the target point cloud points. That is to say, when performing point cloud prediction, the point cloud points that need to be encoded will be grouped and processed, and when predicting a certain point cloud point, the business group that needs to be predicted will be obtained from the grouped business group.
  • the candidate point set of the point cloud group where the point cloud point is located takes into account the spatial correlation between each point cloud group, so that the attribute prediction of the point cloud point can include partial information of the spatial correlation between the groups, thereby improving Subsequent encoding and decoding performance and encoding and decoding efficiency.
  • the associated device mentioned in the embodiment of the present application may be a computer device, and the computer device in the embodiment of the present application includes but is not limited to a terminal device or a server.
  • the computer device can be a server or a terminal device, or a system composed of a server and a terminal device.
  • the terminal device mentioned above can be an electronic device, including but not limited to mobile phones, tablet computers, desktop computers, notebook computers, handheld computers, vehicle-mounted equipment, augmented reality/virtual reality (AR) /VR) equipment, helmet displays, smart TVs, wearable devices, smart speakers, digital cameras, cameras and other mobile Internet devices (MID) with network access capabilities, or in scenarios such as trains, ships, flights, etc. terminal equipment, etc.
  • AR augmented reality/virtual reality
  • MID mobile Internet devices
  • the server mentioned above can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or it can provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, Cloud servers for basic cloud computing services such as network services, cloud communications, middleware services, domain name services, security services, vehicle-road collaboration, Content Delivery Network (CDN), and big data and artificial intelligence platforms.
  • cloud services such as network services, cloud communications, middleware services, domain name services, security services, vehicle-road collaboration, Content Delivery Network (CDN), and big data and artificial intelligence platforms.
  • CDN Content Delivery Network
  • the data involved in the embodiment of the present application can be stored in a computer device, or the data can be stored based on cloud storage technology, which is not limited here.
  • FIG. 3 is a flow chart of a method for point cloud prediction processing in the encoding process provided by an embodiment of the present application.
  • the point cloud prediction process includes the following steps:
  • Step S301 Obtain the candidate point set of the target point cloud group where the target point cloud point is located.
  • the candidate point set belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; the coordinate codeword of the point cloud point included in each point cloud group is passed through the point cloud where it is located.
  • the number of group movement bits corresponding to the group is the same after movement; M is a positive integer.
  • the computer device can obtain M point cloud groups grouped by k point cloud points, k is a positive integer, and the computer device can obtain the target point cloud group in which the target point cloud point is located from the M point cloud groups.
  • the candidate point set that is, the candidate point set is essentially a point cloud group, and the number of the candidate point set can be one or at least two.
  • the computer device can obtain the point cloud to be encoded, and obtain k point cloud points that make up the point cloud to be encoded; or, the computer device can obtain the space filling curve, and obtain the k point cloud points included in the space filling curve. Point cloud points etc.
  • the specific method of obtaining the k point cloud points that need to be encoded is not limited here.
  • the computer equipment can obtain k point cloud points and the coordinate codewords corresponding to each point cloud point, and divide the point cloud points with the same codeword sequence obtained after the group shift bit shift into a group, and obtain M point cloud group.
  • the number of packet shift bits may be a fixed value or a variable value.
  • the number of digits to move in a group is recorded as L.
  • the L can be a quantitative value, that is, the value does not change.
  • the L can also be a variable, that is, the value will change as the grouping proceeds.
  • L can be considered as a Positive integer.
  • the coordinate codeword of the r-th point cloud point can be recorded as H r , assuming that the number of coordinate dimension digits corresponding to the coordinate codeword is dig, and the codeword length of the coordinate codeword in each coordinate dimension is s, r is a positive integer less than or equal to k, dig is a positive integer, and s is a positive integer. Then the coordinate codeword of the r-th point cloud point can be expressed as:
  • coor is used to represent coordinates
  • coor_1 is used to represent the first coordinate dimension
  • coor_dig is used to represent the dig-th coordinate dimension
  • subscript is used to represent the number of codewords corresponding to each coordinate dimension.
  • "s-1" represents the encoding of the (s-1)th bit in each coordinate dimension, in other words, constitutes the encoding of the r-th point cloud point in the first coordinate dimension, It constitutes the encoding of the r-th point cloud point in the second coordinate dimension, etc.
  • dig is 3 and includes three coordinate dimensions: x, y, and z
  • the coordinate codeword of the r-th point cloud point can be expressed as:
  • the coordinate codeword of the r-th point cloud point can be expressed as wait.
  • Figure 4 is a schematic diagram of point cloud point distribution provided by an embodiment of the present application.
  • the coordinate distribution of point cloud points is equivalent to the distribution of a three-dimensional space.
  • the three point cloud points shown in Figure 4 these three point cloud points Belongs to a point cloud group 401.
  • the point cloud points located in the point cloud group 401 all belong to the point cloud group 401.
  • the coordinate code words of each point cloud point located in the point cloud group 401 have been moved.
  • the codeword range included in the point cloud group 401 can be obtained. That is, the range of the point cloud group 401 in Figure 4 can represent the point cloud group sequence of the point cloud group 401.
  • the number of grouping movement digits corresponding to the M point cloud groups are all the default grouping movement digits, that is, the grouping movement digits can be a fixed value, where the default grouping movement digits can be obtained based on experience, Or it can be provided by the user, or it can be the number of historical grouping movements, etc., and the movement grouping is carried out through fixed values, so that the coordinate code words of each point cloud point are grouped on the same basis. In this case, the point cloud point can be improved. The efficiency of grouping and subsequent acquisition of alternative point sets.
  • the number of packet shift bits can be a variable value. Specifically, when the average number of point cloud points contained in M 1 adjacent point cloud groups adjacent to the target point cloud group is greater than the first point threshold, the number of group movement bits corresponding to the target point cloud group is less than M The number of group movement bits corresponding to 1 adjacent point cloud group; when the average number of point cloud points contained in M 1 adjacent point cloud groups adjacent to the target point cloud group is less than the second point threshold, the target point cloud The number of group movement bits corresponding to the group is greater than the number of group movement bits corresponding to M 1 adjacent point cloud groups; M 1 is a positive integer less than M; when there are M 1 adjacent point cloud groups adjacent to the target point cloud group When the average number of point cloud points contained in is greater than or equal to the second point threshold and less than or equal to the first point threshold, the number of group movement digits of the target point cloud group is the same as that of the previous point cloud group of the target point cloud group.
  • the number of group movement bits is the same. That is to say, when the number of point cloud points contained in the adjacent point cloud group is too large, you can reduce the number of point cloud points contained in the subsequently generated point cloud group by reducing the number of group movement bits; when the adjacent point cloud group When the number of point cloud points contained in a point cloud group is too small, you can increase the number of point cloud points contained in the subsequently generated point cloud group by increasing the number of group movement bits, so that the number of point cloud points contained in each point cloud group can be increased. The number of point cloud points is balanced as much as possible to improve the effect of point cloud grouping.
  • the M point cloud groups may include one or at least two inter-group sets.
  • Point cloud groups included in the same inter-group set have the same number of group movement bits.
  • Point clouds in different inter-group sets The number of grouping movement bits of the group is different, the number of point cloud groups included in an inter-group set is less than or equal to the grouping unit threshold, and, when M 1 is adjacent to the first point cloud group in the j-th inter-group set
  • the number of group movement bits of the point cloud group included in the j-th inter-group set is less than M 1 adjacent point cloud groups respectively.
  • the number of corresponding group movement bits when the mean number of point cloud points contained in the M 1 adjacent point cloud groups adjacent to the first point cloud group in the j-th inter-group set is less than the second point number threshold, The number of group movement bits of the point cloud group included in the j-th inter-group set is greater than the number of group movement bits corresponding to M 1 adjacent point cloud groups; when compared with the first point cloud in the j-th inter-group set When the average number of point cloud points contained in M 1 adjacent point cloud groups is greater than or equal to the second point threshold and less than or equal to the first point threshold, the points included in the j-th inter-group set
  • the number of group movement bits of the cloud group can be the same as the number of group movement bits of the previous inter-group set of the jth inter-group set.
  • the grouping movement bits are updated once number, reducing the amount of data that needs to be processed. For example, assuming that the grouping unit threshold is 5, obtain the grouping movement number, and obtain the first point cloud group to the fifth point cloud group based on the grouping movement number; obtain the next point cloud point of the fifth point cloud group The average number of point cloud points contained in M 1 adjacent point cloud groups. Based on the average number of points, the number of group movement bits is updated. Based on the updated group movement number, the 6th point cloud group to the 10th point cloud are obtained. Group until k point cloud points are grouped.
  • the inter-group set may be a concept used to describe the change in the number of packet movement bits.
  • the candidate point set when obtaining the candidate point set of the target point cloud group in the M point cloud groups, in an alternative point set acquisition method, includes the candidate point set located in the target point cloud group in the M point cloud group.
  • the point cloud group before the point cloud group and adjacent to the target point cloud group; the total number of point cloud points included in the candidate point set is less than or equal to the third point number threshold.
  • the third point number threshold can be Denoted as maxNumofNeighbor.
  • the computer device can obtain the point cloud groups sequentially from the M point cloud groups, using the target point cloud group as the reference group, until an alternative point set is obtained, and the point cloud group corresponding to the alternative point set is
  • the total number of point cloud points is less than or equal to the third point number threshold, and the sum of the number of point cloud points included in the candidate point set and the point cloud group located before the candidate point set is greater than the third point number threshold.
  • point cloud group 1 point cloud group 2, point cloud group 3, ... and point cloud group M
  • the target point cloud group is point cloud group 5
  • the third point number threshold is 10
  • point cloud group 4 includes 3 point cloud points, and 3 is less than the third point number threshold; continue to obtain point cloud group 3, assuming that point cloud group 3 includes 5 points.
  • Cloud points at this time, point cloud group 4 and point cloud group 3 include a total of 8 point cloud points, 8 is less than the third point number threshold; continue to obtain point cloud group 2, assuming that point cloud group 2 includes 4 point cloud points,
  • point cloud group 4, point cloud group 3 and point cloud group 2 include a total of 12 point cloud points, and 12 is greater than the third point threshold, then point cloud group 4 and point cloud group 3 are determined as the target point cloud group.
  • a collection of alternative points are determined as the target point cloud group.
  • the accuracy of point cloud prediction can be further improved.
  • the alternative point set is a point cloud group located before the target point cloud group among the M point cloud groups; the number of point cloud points included in each alternative point set is greater than or equal to The threshold for the number of points in the group.
  • the number of candidate point sets is less than or equal to the set number selection threshold.
  • the computer device can use the target point cloud group as the reference group and sequentially obtain candidate point cloud groups from the M point cloud groups, wherein the number of point cloud points included in the candidate point cloud group is greater than or equal to The threshold of the number of points in the group. When the traversal of M point cloud groups is completed, the obtained candidate point cloud group will be determined as the candidate point set of the target point cloud group.
  • point cloud group 1 point cloud group 2, point cloud group 3, ... and point cloud group M.
  • the threshold value of the number of points in the group is 4
  • the target point cloud group is point cloud group 5
  • point cloud group 1 , point cloud group 2 and point cloud group 3 respectively include the number of point cloud points greater than or equal to 4
  • point cloud group 1, point cloud group 2 and point cloud group 3 are determined as candidates for the target point cloud group Point collection.
  • the computer device may use the target point cloud group as the reference group and sequentially obtain candidate point cloud groups among the M point cloud groups, wherein the number of point cloud points included in the candidate point cloud group is greater than or equal to the group Midpoint number threshold.
  • the obtained candidate point cloud group will be determined as the candidate point set of the target point cloud group. For example, there are (point cloud group 1, point cloud group 2, point cloud group 3, ... and point cloud group M). Assume that the threshold of the number of points in the group is 4, the target point cloud group is point cloud group 5, and the threshold for selecting the number of groups is 2. Assume that point cloud group 4 includes 3 point cloud points, and 3 is less than the threshold of the number of points in the group; assuming that point cloud group 3 includes 5 point cloud points, and 5 is greater than the threshold of the number of points in the group, determine point cloud group 3 as a candidate point cloud.
  • point cloud group 2 at this time, there is a candidate point cloud group, and 1 is less than the group number selection threshold; assuming that point cloud group 2 includes 4 point cloud points, and 4 is equal to the point cloud group threshold, point cloud group 2 is determined as the candidate point cloud group , at this time, there are 2 candidate point cloud groups, and 2 is the group number selection threshold, then point cloud group 2 and point cloud group 3 are determined as the candidate point sets of the target point cloud group.
  • the alternative point set is N point cloud groups located before the target point cloud group among the M point cloud groups; N is a positive integer, and N is the default adjacency group threshold.
  • the default adjacency group threshold is the threshold of the number of point cloud groups that precede and are adjacent to the target point cloud group.
  • the computer device can use the target point cloud group as the reference group, acquire N point cloud groups forward among the M point cloud groups, and determine the acquired N point cloud groups as candidate points of the target point cloud group. gather.
  • point cloud group 5 is used as the base group and three point cloud groups are obtained in sequence, namely, point cloud group 4, point cloud group 3 and Point cloud group 2, point cloud group 4, point cloud group 3 and point cloud group 2 are determined as candidate point sets of the target point cloud group.
  • a point cloud group corresponds to a point cloud group sequence.
  • the point cloud group sequence refers to the coordinate codes of the point cloud points included in the corresponding point cloud group.
  • the group's grouping movement is obtained by shifting the number of bits. For example, taking the r-th point cloud point mentioned above as an example, assuming that the grouping movement number of the point cloud group where the r-th point cloud point is located is dig, then the codeword sequence obtained after the r-th point cloud point is moved It can be written as:
  • the coordinate code words of all point cloud points in the point cloud group where the r-th point cloud point is located after the group movement bits are moved, will move with the coordinate code words of the r-th point cloud point.
  • the codeword sequence obtained is the same.
  • the order of the candidate point cloud groups corresponding to the candidate point set is obtained after moving through the first multiple of the number of digits in the coordinate dimension.
  • the alternative movement sequence is the same as the target movement sequence obtained by moving the target point cloud group sequence corresponding to the target point cloud group through the first multiple of the number of coordinate dimension digits; the number of coordinate dimension digits refers to the number of digits in each point cloud group.
  • the number of dimensions corresponding to the coordinate codewords of the included point cloud points that is, the number of dimensions of the coordinate dimensions corresponding to the coordinate codewords.
  • the target point cloud group sequence corresponding to the target point cloud group is recorded as H K1
  • the target point cloud group is recorded as point cloud group K1
  • H K2 >>dig*mul 1 H K1 >>dig*
  • the point cloud group of mul 1 is the movement sequence obtained by moving the point cloud group sequence through the first multiple of the number of coordinate dimension digits.
  • the target point cloud group sequence corresponding to the target point cloud group is obtained after passing through the coordinate dimension digit number.
  • the point cloud group obtained after moving the first multiple of the target point cloud group with the same target movement sequence is determined as the candidate point set of the target point cloud group.
  • mul 1 is used to represent the first multiple, and mul 1 is a positive integer.
  • the sequence of the candidate point cloud groups corresponding to the candidate point set is moved through the number of digits in the coordinate dimension.
  • the alternative movement sequence obtained after moving through the first multiple of the coordinate dimension is the same as the target movement sequence obtained after the target point cloud group sequence corresponding to the target point cloud group is moved through the first multiple of the coordinate dimension digits. In this way, Further improve the accuracy of point cloud prediction.
  • the alternative movement of the candidate point cloud group sequence corresponding to the candidate point set is obtained by moving the number of digits in the supplementary dimension.
  • the sequence is the same as the target movement sequence obtained after the target point cloud group sequence corresponding to the target point cloud group is moved by the number of supplementary dimension bits; the number of supplementary dimension bits refers to the remainder of the number of bits of movement of the target group and the number of coordinate dimension bits.
  • the difference in digits from the coordinate dimension digits; alternatively, the supplementary dimension digits is the sum of the difference in digits plus the second multiple of the coordinate dimension digits.
  • the second multiple can be recorded as mul 2 , and mul 2 is a positive integer.
  • v is a positive integer
  • the target point cloud group sequence corresponding to the target point cloud group is recorded as H K1
  • the target point cloud group It is recorded as point cloud group K1.
  • the number of supplementary dimension digits is (dig-1) or ⁇ (dig-1)+dig*mul 2 ⁇
  • the point cloud group K2 can be considered as a neighbor node of the point cloud group K1.
  • H K2 >> 1 is satisfied H K1 >>1
  • ">>" means movement. For example, assuming that H K1 is 001101, then after H K1 >>1, 00110 is obtained.
  • the sequence of the candidate point cloud groups corresponding to the candidate point set is moved through the number of digits in the coordinate dimension.
  • the alternative movement sequence obtained after moving through the first multiple of the coordinate dimension is the same as the target movement sequence obtained after the target point cloud group sequence corresponding to the target point cloud group is moved through the first multiple of the coordinate dimension digits. In this way, Further improve the accuracy of point cloud prediction.
  • the obtained target movement sequence can be represented by area 402, and the point cloud group located in this area 402 can be determined as a point cloud group.
  • the area 402 also includes 6 co-planar neighbor nodes, such as the area indicated by the dotted box in Figure 4. It may also include 12 co-linear neighbor nodes, 8 co-point neighbor nodes, etc.
  • the computer device may obtain a set of candidate points of the target point cloud group from coplanar neighbor nodes, collinear neighbor nodes, or common point neighbor nodes of the parent node of the target point cloud group based on needs.
  • the number of coplanar neighbor nodes, the number of collinear neighbor nodes, and the number of common point neighbor nodes are determined by the number of coordinate dimension dig, and are not limited here.
  • the acquisition process of the candidate point set may be implemented through any one of the above-mentioned alternative point set acquisition methods or any combination of multiple methods.
  • Step S302 Obtain the prediction reference point associated with the target point cloud point from the candidate point set.
  • the computer device can obtain the predicted reference point associated with the target point cloud point from the candidate point set.
  • the number of candidate point sets is one or at least two.
  • the point cloud points included in the candidate point set can be directly determined as the prediction associated with the target point cloud point.
  • Reference point or, obtain the inter-point distance between the point cloud points included in the candidate point set and the target point cloud point, based on the point between the point cloud point included in the candidate point set and the target point cloud point distance between each other, and obtain the predicted reference point associated with the target point cloud point from the candidate point set, etc.
  • the distance between points can be the distance between the coordinates of the point cloud points included in the candidate point set and the coordinates of the target point cloud point, which can also be called geometric distance, or it can be the distance between the coordinates of the point cloud points included in the candidate point set.
  • the number of candidate point sets is at least two.
  • the computer device can select d point cloud points respectively from at least two candidate point sets, and combine the d point cloud points corresponding to the at least two candidate point sets, Determine the prediction reference point associated with the target point cloud point; d is a positive integer. For example, assuming that d is 1, obtain one point cloud point from at least two candidate point sets respectively as the prediction reference point associated with the target point cloud point. In one embodiment, in this manner, it can be considered that the prediction reference points are the same for the point cloud points in the same point cloud group. Therefore, the computer device can search for the prediction reference corresponding to the target point cloud group.
  • the prediction reference point corresponding to the target point cloud group is obtained, then the prediction reference point corresponding to the target point cloud group is determined as the prediction reference point of the target point cloud point; if the prediction reference point of the target point cloud group is not obtained, Corresponding prediction reference points, then select d point cloud points respectively from at least two candidate point sets, and determine the d point cloud points corresponding to the at least two candidate point sets as the target point cloud points associated Predicting reference points, at the same time, the d point cloud points corresponding to the at least two candidate point sets can be determined as the prediction reference points corresponding to the target point cloud group; through the above process, the same point cloud group can be , the prediction reference point only needs to be obtained once, thereby reducing the amount of data that needs to be processed and improving the efficiency of obtaining the prediction reference point.
  • d point cloud points are selected from at least two candidate point sets as candidate point cloud points, and the first point between the candidate point cloud point and the target point cloud point is obtained.
  • Distance sort the candidate point cloud points based on the distance between the first points, and obtain the predicted reference points associated with the target point cloud points from the sorted candidate point cloud points.
  • the distance between the first points can be the distance between the coordinates of the candidate point cloud point and the coordinates of the target point cloud point, which can also be called the geometric distance, or it can be the distance between the candidate point cloud point and the target point cloud point within M The number of interval points in the point cloud group, etc.
  • the candidate point cloud points are sorted based on the first inter-point distance between the candidate point cloud points and the target point cloud point, and the information associated with the target point cloud point is directly obtained from the sorted candidate point cloud points.
  • the prediction reference point can improve the efficiency of obtaining the prediction reference point.
  • the second inter-point distance between the point cloud points included in at least two candidate point sets and the target point cloud point is obtained, and the at least two candidate points are compared based on the second inter-point distance.
  • the point cloud points included in the set are sorted, and the prediction reference points associated with the target point cloud points are obtained from the point cloud points included in the sorted at least two candidate point sets.
  • the second inter-point distance may be the distance between the coordinates of the point cloud points included in the at least two candidate point sets and the coordinates of the target point cloud point, which may also be called a geometric distance, or may be at least two The number of intervals between the point cloud points included in the candidate point sets and the target point cloud points in the M point cloud groups, etc.
  • the second inter-point distance may be the geometric distance between the coordinates of point cloud point 1 and the coordinates of the target point cloud point, or may be the point cloud point The number of interval points between 1 and the positions of the target point cloud points in the M point cloud groups, etc.
  • the point cloud points included in the at least two candidate point sets are sorted according to the second inter-point distance between the point cloud points included in the at least two candidate point sets and the target point cloud point, And directly obtain the prediction reference points associated with the target point cloud points from the point cloud points included in the sorted at least two candidate point sets, which can improve the efficiency of obtaining the prediction reference points.
  • the set priorities corresponding to at least two candidate point sets are obtained, and the at least two candidate sets are sorted based on the set priorities.
  • the computer device can obtain the inter-group association between at least two alternative point sets and the target point cloud group, and determine the set priority corresponding to the at least two alternative point sets respectively based on the inter-group association.
  • the inter-group association includes but is not limited to neighbor association, parent association and interval association.
  • the neighbor association can be considered as the point cloud group adjacent to the target point cloud group and the target point cloud group.
  • the inter-group association relationship between; the parent association relationship can be considered as the inter-group association relationship between point cloud groups with the same parent node, that is to say, the point cloud group sequence of two point cloud groups with parent-level association relationship It is the same after the second move; the interval association relationship can be considered as the inter-group association relationship between non-adjacent point cloud groups located in M point cloud groups, etc.
  • the set priorities corresponding to at least two candidate point sets may also be preset in advance. Alternatively, if at least two candidate point sets are obtained through multiple candidate point set acquisition methods, the at least two candidate point sets can be determined based on the candidate point set acquisition methods corresponding to the at least two candidate point sets.
  • the at least two candidate point sets are sorted according to their corresponding set priorities, and the target point cloud point association is directly obtained from the sorted at least two candidate sets.
  • the prediction reference point can improve the efficiency of obtaining the prediction reference point.
  • Step S303 Perform prediction processing on the target point cloud points based on the prediction reference points to obtain target attribute prediction values of the target point cloud points.
  • the computer device can obtain the predicted reference coordinates of the predicted reference point and obtain the target coordinates of the target point cloud point. Based on the coordinate distance between the predicted reference coordinates and the target coordinates, the reference weight of the predicted reference point is determined. Specifically, the computer device can obtain the reciprocal of the coordinate distance and determine the reference weight of the predicted reference point corresponding to the coordinate distance. In one embodiment, the computer device may determine the sum of coordinate differences between the predicted reference coordinates and the target coordinates in each coordinate dimension as the coordinate distance between the predicted reference coordinates and the target coordinates. At this time, the reference weight of the prediction reference point can be seen in formula 1:
  • w iu is used to represent the reference weight between the u-th prediction reference point of the i-th point cloud point and the i-th point cloud point
  • the subscript iu is used to represent the reference weight of the i-th point cloud point.
  • the u-th prediction reference point for example, coor_2 iu is used to represent the coordinate value of the u-th prediction reference point of the i-th point cloud point in the second coordinate dimension.
  • the u-th prediction reference point is in the second coordinate
  • the absolute value of the difference between the coordinate value in the second dimension and the coordinate value of the i-th point cloud point in the second coordinate dimension can be the difference between the u-th prediction reference point and the i-th point cloud point in the second coordinate dimension.
  • the coordinate difference in coordinate dimensions that is,
  • i is a positive integer less than or equal to k.
  • dig is 3, that is, the number of digits in the coordinate dimension is 3, including the three coordinate dimensions of x, y, and z
  • the reference weight of the prediction reference point can be seen in formula 2:
  • the computer device can obtain the dimension weights corresponding to the dig coordinate dimensions, and perform a weighted calculation of the coordinate differences between the predicted reference coordinates and the target coordinates in each coordinate dimension based on the dimension weights corresponding to the dig coordinate dimensions. and, get the coordinate distance between the predicted reference coordinates and the target coordinates.
  • dig 3
  • the number of digits in the coordinate dimension is 3, including the three coordinate dimensions of x, y, and z
  • the reference weight of the prediction reference point can be seen in formula 3:
  • a is used to represent the dimension weight corresponding to the x coordinate dimension
  • b is used to represent the coordinate dimension corresponding to the y coordinate dimension
  • c is used to represent the dimension weight corresponding to the z coordinate dimension.
  • the coordinate distance between the predicted reference point and the target point cloud point is not limited to the above calculation method.
  • the coordinate distance between the i-th point cloud point and the u-th prediction reference point of the i-th point cloud point is obtained, and there is no limit here.
  • the reference attribute reconstruction value of the prediction reference point can be obtained, and weighted processing is performed based on the reference attribute reconstruction value and the reference weight to obtain the target attribute prediction value of the target point cloud point.
  • the prediction value of the target attribute can be obtained as shown in formula 4:
  • the The reference attribute reconstruction value used to represent the u-th predicted reference point of the i-th point cloud point w iu is used to represent the reference weight between the u-th predicted reference point of the i-th point cloud point and the i-th point cloud point, and num is used to represent the total number of predicted reference points of the i-th point cloud point.
  • the attribute prediction value of any one of the k point cloud points can be obtained through the above formula 1 to formula 4.
  • the target point cloud point is any point cloud point among the k point cloud points. Therefore, the target attribute prediction value of the target point cloud point can be obtained through the above formula 1 to formula 4.
  • the optimization parameters can also be used to optimize the reference weight to obtain the optimized weight based on the reference attribute.
  • the reconstruction value and the optimization weight are weighted to obtain the target attribute prediction value of the target point cloud point.
  • the optimization parameters may include but are not limited to the attribute quantization step size, or the reference number of predicted reference points with the largest coordinate distance from the target point cloud point, etc.
  • the computer device may determine the attribute quantization step size as the optimization parameter.
  • a reference number of the predicted reference points with the largest coordinate distance from the target point cloud point is obtained among the predicted reference points, and the reference number is determined as the optimization parameter.
  • a smaller number can be obtained as the optimization parameter from the attribute quantization step size and reference quantity obtained above. That is, when the attribute quantification step size is greater than the reference quantity, the reference quantity is determined as the optimization parameter; when the attribute quantification step size is When it is less than the reference quantity, the attribute quantification step size is determined as the optimization parameter; when the two are equal, the attribute quantification step size or the reference quantity is determined as the optimization parameter, etc.
  • the reference weight of the predicted reference point is determined by the coordinate distance between the predicted reference coordinate of the predicted reference point and the target coordinate of the target point cloud point, and is weighted by the reference attribute reconstruction value of the predicted reference point and the reference weight. Processing to obtain the target attribute prediction value of the target point cloud point, which can improve the prediction accuracy of the target attribute prediction value.
  • the computer device can obtain the actual value of the target attribute of the target point cloud point, and obtain the target attribute residual of the target point cloud point based on the difference between the actual value of the target attribute and the predicted value of the target attribute of the target point cloud point.
  • the target attribute residual is quantified and converted to obtain the target transformation coefficient of the target point cloud point.
  • This quantification conversion method is not limited here.
  • the target attribute residuals can be subjected to Discrete Cosine Transform (DCT) to obtain the target transformation coefficients of the target point cloud points, or the target point cloud can be obtained by constructing a binary tree.
  • the target transformation coefficient of the point wherein, the target transform coefficient includes a first transform coefficient and a second transform coefficient.
  • the computer device can also obtain the attribute residuals of the point cloud points included in the target point cloud group, perform a quantitative conversion on the attribute residuals of the point cloud points included in the target point cloud group, and obtain the target point.
  • the transformation coefficients of the point cloud points included in the cloud group include the target transformation coefficients of the target point cloud points.
  • the target attribute residual of the target point cloud point is obtained through the difference between the actual value of the target attribute of the target point cloud point and the predicted value of the target attribute of the target point cloud point, and the target attribute residual is quantified. Convert to obtain the target transformation coefficient of the target point cloud point to better encode the target point cloud point and further improve the coding performance and coding efficiency.
  • the computer device can perform encoding on a group-by-group basis. Specifically, the transformation coefficients of the point cloud points included in the target point cloud group are obtained, the transformation coefficients of the point cloud points included in the target point cloud group are encoded, and the target code stream corresponding to the target point cloud group is obtained.
  • the target code stream corresponding to the target point cloud group is obtained, thereby improving the acquisition efficiency of the encoding results, thereby improving the encoding performance and encoding efficiency.
  • the group number limit threshold can be obtained, and g point cloud groups to be encoded containing the target point cloud group are obtained based on the group number limit threshold, and the transformation coefficients of the point cloud points included in the g point cloud groups to be encoded are obtained.
  • the transformation coefficients of the point cloud points included in the point cloud group to be encoded are encoded to obtain group code streams corresponding to g point cloud groups to be encoded; g is a positive integer, and g is less than or equal to the group number limit threshold.
  • a candidate point set of the target point cloud group where the target point cloud point is located is obtained; the candidate point set belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; each point cloud The coordinate codewords of the point cloud points included in the group are the same after being moved by the group movement bits corresponding to the point cloud group; M is a positive integer; the prediction reference associated with the target point cloud point is obtained from the candidate point set point, perform prediction processing on the target point cloud points based on the prediction reference points, and obtain the target attribute prediction value of the target point cloud points.
  • encoding or decoding can be performed based on the predicted value of the target attribute, that is, encoding and decoding of the point cloud is implemented, where the candidate point set is obtained from the point cloud group, which is based on the point cloud group.
  • the coordinate codes of the cloud points are obtained by grouping the point cloud points, and the predicted reference points associated with the target point cloud points are obtained from the obtained set of candidate points, and the spatial correlation between each point cloud group can be taken into account.
  • corresponding attribute predictions can be made based on the spatial correlation between each point cloud group, thereby improving the accuracy of point cloud prediction.
  • encoding and decoding on this basis can improve the performance and efficiency of encoding and decoding.
  • Figure 5 is a point cloud point grouping process in one embodiment provided by the embodiment of the present application. As shown in Figure 5, the point cloud point grouping process includes the following steps:
  • Step S501 Obtain k point cloud points and the coordinate codeword of each point cloud point.
  • the computer device can obtain k point cloud points and the coordinate codeword of each point cloud point.
  • the computer device can obtain the space-filling curve codeword of the space-filling curve, and obtain the space-filling curve codeword from the space-filling curve codeword. Obtain the coordinate codewords corresponding to k point cloud points.
  • Step S502 Obtain the number of packet movement bits.
  • the number of packet movement bits L can be obtained.
  • L is a variable value.
  • Step S503 Move the coordinate codeword of the i-th point cloud point based on the number of grouping movement bits to obtain the codeword sequence of the i-th point cloud point.
  • the computer device can move the coordinate codeword of the i-th point cloud point to obtain the codeword sequence of the i-th point cloud point. For example, assuming that the coordinate codeword of the i-th point cloud point is "0010101110", and the number of group movement digits at this time is 3, then by moving the coordinate codeword of the i-th point cloud point, the i-th point cloud point can be obtained The codeword sequence of point cloud points is "0010101".
  • Step S504 Check whether the i-th codeword sequence is the same as the (i-1)-th codeword sequence.
  • the computer device can detect whether the i-th codeword sequence is the same as the (i-1)th codeword sequence, where, when i is 1, the (i-1)th point cloud point The codeword sequence defaults to empty or special identifier. If the codeword sequence of the i-th point cloud point is the same as the codeword sequence of the (i-1)th point cloud point, step S505 is executed; if the codeword sequence of the i-th point cloud point is the same as the codeword sequence of the (i-1)th point cloud point ) point cloud points have different codeword sequences, then step S508 is executed.
  • Step S505 Add the i-th point cloud point to the initial point cloud group where the (i-1)-th point cloud point is located.
  • the codeword sequence of the i-th point cloud point is the same as the codeword sequence of the (i-1)th point cloud point, it means that the i-th point cloud point can be the same as the (i-1)th point cloud point.
  • point cloud points are divided into a point cloud group, and the i-th point cloud point can be added to the initial point cloud group where the (i-1)-th point cloud point is located.
  • an initial point cloud group can be created based on the i-th point cloud point and the (i-1)th point cloud point; if There is an initial point cloud group for (i-1) point cloud points, then the i-th point cloud point is added to the initial point cloud group where the (i-1)-th point cloud point is located. For example, if i is 4, if the codeword sequence of the 4th point cloud point is the same as the codeword sequence of the 3rd point cloud point, then the 4th point cloud point is added to the initial point where the 3rd point cloud point is located. Cloud group.
  • step S513 is executed; if i is not k, step S507 is executed.
  • the value of i is increased by one, that is, the next point cloud point is processed.
  • Step S508 Generate the current point cloud group, i++.
  • the initial point cloud group where the (i-1)th point cloud point is located can be obtained, and the initial point cloud group where the (i-1)th point cloud point is located can be determined as a point cloud group .
  • the (i-1)th point cloud point is formed into a point cloud group. Further, it can be detected whether i is k. If i is k, step S513 is executed; if i is not k, i is updated, that is, i++, and step S509 is executed.
  • Step S509 Check whether the group statistical value e is the grouping unit threshold.
  • step S511 it is detected whether the group statistical value e is the grouping unit threshold. If e is the grouping unit threshold, step S511 is executed; if e is not the grouping unit threshold, step S510 is executed. Among them, the initial value of e is 0.
  • step S503 add one to the value of e and return to step S503.
  • Step S511 reset the group statistical value e.
  • the group statistical value e is reset, that is, the group statistical value e is reset to the initial value, and step S512 is performed.
  • Step S512 When the average number of point cloud points contained in the M1 adjacent point cloud groups of the current point cloud group is greater than the first point threshold, reduce the number of group movement bits; the M1 adjacent point cloud groups of the current point cloud group include When the average number of point cloud points is less than the second point threshold, increase the number of group movement digits.
  • the number of packet movement bits is updated, that is, the value of L is updated, and the execution returns to step S503.
  • Step S513 Generate the current point cloud group and end the grouping.
  • the current point cloud group is generated based on the initial point cloud group where the i-th point cloud point is located; in one embodiment, if there is no initial point cloud group for the i-th point cloud point, the current point cloud group is generated based on the i-th point cloud point. i point cloud points generate the current point cloud group. At this time, M point cloud groups composed of k point cloud points are obtained, and the grouping is completed.
  • FIG. 6 is a flow chart of a method for point cloud prediction processing in the decoding process provided by an embodiment of the present application.
  • the point cloud prediction process includes the following steps:
  • Step S601 Obtain the candidate point set of the target point cloud group where the target point cloud point is located.
  • the candidate point set belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; the coordinate codeword of the point cloud point included in each point cloud group is passed through the point cloud where it is located.
  • the number of group movement bits corresponding to the group is the same after movement; M is a positive integer.
  • Step S602 Obtain the prediction reference point associated with the target point cloud point from the candidate point set.
  • the specific implementation process may refer to the specific description shown in step S302 in Figure 3 .
  • Step S603 Predict the target point cloud point based on the prediction reference point to obtain the target attribute prediction value of the target point cloud point.
  • the specific implementation process may refer to the specific description shown in step S303 in Figure 3 .
  • the attribute prediction value of any one of the k point cloud points can be obtained.
  • k is a positive integer
  • the target point cloud point refers to any point cloud point among the k point cloud points.
  • Step S604 Obtain the code stream corresponding to the target point cloud point, decode the code stream corresponding to the target point cloud point, and obtain the target attribute residual of the target point cloud point.
  • the code stream corresponding to the target point cloud point is decoded to obtain the target transformation coefficient of the target point cloud point. Perform inverse transformation processing on the target transformation coefficient to obtain the target attribute residual of the target point cloud point.
  • the target attribute residual of the target point cloud point is obtained, which can improve the acquisition efficiency of the target attribute residual.
  • the code stream corresponding to the target point cloud point refers to the code stream of multiple point cloud groups including the target point cloud group in which the target point cloud point is located. stream; or, when the code stream is encoded group by group, the code stream corresponding to the target point cloud point refers to the code stream of the target point cloud group, etc.
  • the computer device can decode the code stream corresponding to the target point cloud point to obtain the transformation coefficient sequence corresponding to the code stream, and reverse the sequence of transformation coefficients corresponding to the code stream to obtain the transformation coefficient sequence corresponding to the code stream.
  • the reverse sorting is a sorting method based on restoring the sorting method of the transformation coefficients during the encoding process.
  • Step S605 Determine the target attribute reconstruction value of the target point cloud point based on the target attribute prediction value and the target attribute residual.
  • the target attribute prediction value and the target attribute residual can be attribute fused to obtain the target attribute reconstruction value of the target point cloud point.
  • the attribute fusion can be attribute addition, etc.
  • step S604 may be executed first, and then steps S601 to S603, etc. may be executed.
  • a candidate point set of the target point cloud group where the target point cloud point is located is obtained; the candidate point set belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; each point cloud The coordinate codewords of the point cloud points included in the group are the same after being moved by the group movement bits corresponding to the point cloud group; M is a positive integer; the prediction reference associated with the target point cloud point is obtained from the candidate point set point, perform prediction processing on the target point cloud points based on the prediction reference points, and obtain the target attribute prediction value of the target point cloud points.
  • encoding or decoding can be performed based on the predicted value of the target attribute, that is, encoding and decoding of the point cloud is implemented, where the candidate point set is obtained from the point cloud group, which is based on the point cloud group.
  • the coordinate codes of the cloud points are obtained by grouping the point cloud points, and the predicted reference points associated with the target point cloud points are obtained from the obtained set of candidate points, and the spatial correlation between each point cloud group can be taken into account.
  • corresponding attribute predictions can be made based on the spatial correlation between each point cloud group, thereby improving the accuracy of point cloud prediction.
  • encoding and decoding on this basis can improve the performance and efficiency of encoding and decoding.
  • the above encoding process and decoding process can be implemented in the same computer device, or can be implemented in different computer devices.
  • Figure 7 is a data interaction architecture diagram provided by an embodiment of the present application.
  • the computer device 701 can encode k point cloud points to obtain an encoded code stream, where the number of code streams corresponding to the k point cloud points can be one or at least two.
  • the computer device 701 can send the encoded code stream to the computer device 702, and the computer device 702 can decode the obtained code stream, so that k point cloud points can be obtained.
  • the attributes corresponding to the k point cloud points can be obtained. Rebuild value.
  • the computer device 701 can obtain k point cloud points from the computer device 701, can also obtain k point cloud points from the computer device 702, or can obtain k point cloud points from other associated devices. point, there is no restriction here.
  • FIG. 8 is a schematic diagram of a point cloud prediction processing device provided by an embodiment of the present application.
  • the point cloud prediction processing device may be a computer readable instruction (including program code, etc.) running in a computer device.
  • the point cloud prediction processing device may be an application software; the device may be used to execute the provisions of the embodiments of the present application. corresponding steps in the method.
  • the point cloud prediction processing device 800 can be used in the computer equipment in the embodiment corresponding to Figure 3.
  • the device can include: a candidate set acquisition module 11, a reference point acquisition module 12 and attribute prediction. Module 13.
  • the candidate set acquisition module 11 is used to obtain the candidate point set of the target point cloud group where the target point cloud point is located; the candidate point set belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; each The coordinate codes of the point cloud points included in the point cloud group are the same after being moved by the group movement bits corresponding to the point cloud group; M is a positive integer;
  • the reference point acquisition module 12 is used to obtain the predicted reference point associated with the target point cloud point from the candidate point set;
  • the attribute prediction module 13 is used to perform prediction processing on the target point cloud point based on the prediction reference point, and obtain the target attribute prediction value of the target point cloud point.
  • the number of group movement digits corresponding to the M point cloud groups are all the default group movement digits; or,
  • the number of group movement bits corresponding to the target point cloud group is less than M 1 adjacent points
  • M 1 adjacent points The number of group movement bits corresponding to each point cloud group; when the average number of point cloud points contained in the M 1 adjacent point cloud groups adjacent to the target point cloud group is less than the second point threshold, the target point cloud group corresponding to The number of grouping movements is greater than the number of grouping movements corresponding to M 1 adjacent point cloud groups; M 1 is a positive integer less than M.
  • the candidate point set is a point cloud group that is located in front of the target point cloud group and adjacent to the target point cloud group among the M point cloud groups; the total number of point cloud points included in the candidate point set is less than or equal to the Three point threshold.
  • the candidate point set is a point cloud group located before the target point cloud group among the M point cloud groups; the number of point cloud points included in each candidate point set is greater than or equal to the point number threshold in the group.
  • the candidate point set is the N point cloud groups located before the target point cloud group among the M point cloud groups; N is a positive integer, and N is the default adjacent group threshold.
  • a point cloud group corresponds to a point cloud group sequence.
  • the point cloud group sequence refers to the coordinate codeword of the point cloud points included in the corresponding point cloud group. After the group movement bits of the corresponding point cloud group are moved, owned;
  • the order of the candidate point cloud groups corresponding to the candidate point set is obtained after moving through the first multiple of the number of digits in the coordinate dimension.
  • the alternative movement sequence is the same as the target movement sequence obtained by moving the target point cloud group sequence corresponding to the target point cloud group through the first multiple of the number of coordinate dimension digits; the number of coordinate dimension digits refers to the number of digits in each point cloud group.
  • the alternative movement of the candidate point cloud group sequence corresponding to the candidate point set is obtained by moving the number of digits in the supplementary dimension.
  • the sequence is the same as the target movement sequence obtained after the target point cloud group sequence corresponding to the target point cloud group is moved by the number of supplementary dimension bits; the number of supplementary dimension bits refers to the remainder of the target group movement number and the coordinate dimension, which is the same as the coordinate dimension.
  • the difference in digits between dimensions; alternatively, the supplementary dimension digits is the sum of the difference in digits plus the second multiple of the digits in the coordinate dimension.
  • the reference point acquisition module 12 includes:
  • the candidate acquisition unit 121 is configured to select d point cloud points respectively from at least two candidate point sets, and determine the d point cloud points corresponding to the at least two candidate point sets as target point cloud point associations.
  • prediction reference point; d is a positive integer.
  • the reference point acquisition module 12 includes:
  • the candidate acquisition unit 122 is used to select d point cloud points respectively from at least two candidate point sets as candidate point cloud points;
  • the distance selection unit 123 is used to obtain the first inter-point distance between the candidate point cloud point and the target point cloud point, sort the candidate point cloud points based on the first inter-point distance, and obtain from the sorted candidate point cloud points.
  • the reference point acquisition module 12 includes:
  • the candidate sorting unit 124 is used to obtain the second inter-point distance between the point cloud points included in the at least two candidate point sets and the target point cloud point, and sort the at least two candidate point sets based on the second inter-point distance.
  • the included point cloud points are sorted, and the predicted reference points associated with the target point cloud points are obtained from the point cloud points included in the sorted at least two candidate point sets.
  • the reference point acquisition module 12 includes:
  • the priority acquisition unit 125 is used to obtain the set priorities corresponding to at least two candidate point sets, sort the at least two candidate sets based on the set priorities, and obtain the set priorities of the at least two candidate sets after sorting.
  • P prediction reference points associated with the target point cloud point; P is a positive integer.
  • the attribute prediction module 13 includes:
  • the coordinate acquisition unit 131 is used to obtain the predicted reference coordinates of the predicted reference point and obtain the target coordinates of the target point cloud point;
  • the weight acquisition unit 132 is configured to determine the reference weight of the predicted reference point based on the coordinate distance between the predicted reference coordinates and the target coordinates;
  • the attribute prediction unit 133 is used to obtain the reference attribute reconstruction value of the prediction reference point, perform weighting processing based on the reference attribute reconstruction value and the reference weight, and obtain the target attribute prediction value of the target point cloud point.
  • the device 800 also includes:
  • the residual acquisition module 14 is used to obtain the actual value of the target attribute of the target point cloud point, and obtain the target attribute residual of the target point cloud point based on the difference between the actual value of the target attribute and the predicted value of the target attribute of the target point cloud point. ;
  • the quantization processing module 15 is used to perform quantification conversion on the target attribute residual to obtain the target transformation coefficient of the target point cloud point.
  • the device 800 also includes:
  • the first encoding module 16 is used to obtain the transformation coefficients of the point cloud points included in the target point cloud group, encode the transformation coefficients of the point cloud points included in the target point cloud group, and obtain the target code corresponding to the target point cloud group. flow; or,
  • the second encoding module 17 is used to obtain the group number limit threshold, obtain g point cloud groups to be encoded including the target point cloud group based on the group number limit threshold, and obtain the transformation of point cloud points included in the g point cloud groups to be encoded. Coefficient, encode the transformation coefficients of the point cloud points included in the g point cloud groups to be encoded, and obtain the group code stream corresponding to the g point cloud groups to be encoded; g is a positive integer, and g is less than or equal to the group number limit threshold .
  • FIG. 9 is a schematic diagram of a point cloud prediction processing device provided by an embodiment of the present application.
  • the point cloud prediction processing device may be a computer readable instruction (including program code, etc.) running in a computer device.
  • the point cloud prediction processing device may be an application software; the device may be used to execute the provisions of the embodiments of the present application. corresponding steps in the method.
  • the point cloud prediction processing device 900 can be used in the computer equipment in the embodiment corresponding to Figure 6.
  • the device can include: a candidate set acquisition module 21, a reference point acquisition module 22, and attribute prediction. Module 23, code stream acquisition module 24, code stream decoding module 25 and attribute reconstruction module 26.
  • the candidate set acquisition module 21 is used to obtain the candidate point set of the target point cloud group where the target point cloud point is located; the candidate point set belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; each The coordinate codes of the point cloud points included in the point cloud group are the same after being moved by the group movement bits corresponding to the point cloud group; M is a positive integer;
  • the reference point acquisition module 22 is used to obtain the predicted reference point associated with the target point cloud point from the candidate point set;
  • the attribute prediction module 23 is used to perform prediction processing on the target point cloud point based on the prediction reference point, and obtain the target attribute prediction value of the target point cloud point;
  • the code stream acquisition module 24 is used to obtain the code stream corresponding to the target point cloud point;
  • the code stream decoding module 25 is used to decode the code stream corresponding to the target point cloud point and obtain the target attribute residual of the target point cloud point;
  • the attribute reconstruction module 26 is used to determine the target attribute reconstruction value of the target point cloud point based on the target attribute prediction value and the target attribute residual.
  • the code stream decoding module 25 includes:
  • the initial decoding unit 251 is used to decode the code stream corresponding to the target point cloud point and obtain the target transformation coefficient of the target point cloud point;
  • the coefficient transformation unit 252 is used to perform inverse transformation processing on the target transformation coefficient to obtain the target attribute residual of the target point cloud point.
  • the embodiment of the present application provides a point cloud prediction processing device, which can obtain a set of candidate points of the target point cloud group where the target point cloud point is located; the set of candidate points belongs to M point cloud groups, and the M point cloud groups Including the target point cloud group; the coordinate codewords of the point cloud points included in each point cloud group are the same after moving through the group movement bits corresponding to the point cloud group; M is a positive integer; from the candidate point set Obtain the prediction reference point associated with the target point cloud point, perform prediction processing on the target point cloud point based on the prediction reference point, and obtain the target attribute prediction value of the target point cloud point.
  • encoding or decoding can be performed based on the predicted value of the target attribute, that is, encoding and decoding of the point cloud is implemented, where the candidate point set is obtained from the point cloud group, which is based on the point cloud group.
  • the coordinate codes of the cloud points are obtained by grouping the point cloud points, and the predicted reference points associated with the target point cloud points are obtained from the obtained set of candidate points, and the spatial correlation between each point cloud group can be taken into account.
  • corresponding attribute predictions can be made based on the spatial correlation between each point cloud group, thereby improving the accuracy of point cloud prediction.
  • encoding and decoding on this basis can improve the performance and efficiency of encoding and decoding.
  • Figure 10 is a schematic structural diagram of a computer device provided by an embodiment of the present application.
  • the computer device in this embodiment of the present application may include: one or more processors 1001, a memory 1002, and an input and output interface 1003.
  • the processor 1001, the memory 1002 and the input/output interface 1003 are connected through a bus 1004.
  • the memory 1002 is used to store computer readable instructions, which include program instructions.
  • the input and output interface 1003 is used to receive data and output data, such as for data interaction; the processor 1001 is used to execute the program instructions stored in the memory 1002.
  • the processor 1001 can perform the following operations when encoding:
  • the candidate point set of the target point cloud group where the target point cloud point is located belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; the point cloud points included in each point cloud group
  • the coordinate codewords are the same after moving through the group movement bits corresponding to the point cloud group; M is a positive integer;
  • the processor 1001 can perform the following operations:
  • the candidate point set of the target point cloud group where the target point cloud point is located belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; the point cloud points included in each point cloud group
  • the coordinate codewords are the same after moving through the group movement bits corresponding to the point cloud group; M is a positive integer;
  • Obtain the code stream corresponding to the target point cloud point decode the code stream corresponding to the target point cloud point, and obtain the target attribute residual of the target point cloud point. Based on the target attribute prediction value and the target attribute residual, determine the target point The target attribute reconstruction value of the cloud point.
  • the processor 1001 can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), special-purpose integrated Circuit (application specific integrated circuit, ASIC), off-the-shelf programmable gate array (field-programmable gate array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • DSP digital signal processors
  • ASIC application specific integrated circuit
  • FPGA off-the-shelf programmable gate array
  • a general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.
  • the memory 1002 may include read-only memory and random access memory, and provides instructions and data to the processor 1001 and the input-output interface 1003. A portion of memory 1002 may also include non-volatile random access memory. For example, memory 1002 may also store device type information.
  • the computer device can execute the implementation provided by each step in Figure 3 or Figure 6 through its built-in functional modules.
  • Embodiments of the present application provide a computer device, including: a processor, an input and output interface, and a memory.
  • the processor obtains computer-readable instructions in the memory, executes each step of the method shown in Figure 3, and performs point cloud prediction. processing operations.
  • the embodiment of the present application realizes the acquisition of the candidate point set of the target point cloud group where the target point cloud point is located; the candidate point set belongs to M point cloud groups, and the M point cloud groups include the target point cloud group; each point cloud group The coordinate codewords of the included point cloud points are the same after being moved by the group movement bits corresponding to the point cloud group; M is a positive integer; the predicted reference point associated with the target point cloud point is obtained from the candidate point set , perform prediction processing on the target point cloud points based on the prediction reference points, and obtain the target attribute prediction values of the target point cloud points.
  • encoding or decoding can be performed based on the predicted value of the target attribute, that is, encoding and decoding of the point cloud is implemented, where the candidate point set is obtained from the point cloud group, which is based on the point cloud group.
  • the coordinate codes of the cloud points are obtained by grouping the point cloud points, and the predicted reference points associated with the target point cloud points are obtained from the obtained set of candidate points, and the spatial correlation between each point cloud group can be taken into account.
  • corresponding attribute predictions can be made based on the spatial correlation between each point cloud group, thereby improving the accuracy of point cloud prediction.
  • encoding and decoding on this basis can improve the performance and efficiency of encoding and decoding.
  • Embodiments of the present application also provide a computer-readable storage medium that stores computer-readable instructions.
  • the computer-readable instructions are suitable for the processor to load and execute the steps in Figure 3 or Figure 6.
  • the provided point cloud prediction processing method please refer to the implementation provided by each step in Figure 3 or Figure 6 for details, and will not be described again here.
  • the description of the beneficial effects of using the same method will not be described again.
  • technical details not disclosed in the computer-readable storage medium embodiments involved in this application please refer to the description of the method embodiments in this application.
  • computer readable instructions may be deployed to execute on one computer device, or on multiple computer devices located at one location, or on multiple computers distributed at multiple locations and interconnected by a communications network. executed on the device.
  • the computer-readable storage medium may be the point cloud prediction processing device provided in any of the foregoing embodiments or an internal storage unit of the computer device, such as a hard disk or memory of the computer device.
  • the computer-readable storage medium can also be an external storage device of the computer device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card equipped on the computer device, Flash card, etc.
  • the computer-readable storage medium may also include both an internal storage unit of the computer device and an external storage device.
  • the computer-readable storage medium is used to store the computer-readable instructions and other programs and data required by the computer device.
  • the computer-readable storage medium can also be used to temporarily store data that has been output or is to be output.
  • Embodiments of the present application also provide a computer-readable instruction product or computer-readable instructions.
  • the computer-readable instruction product or computer-readable instructions include computer-readable instructions.
  • the computer-readable instructions are stored in a computer-readable storage medium. .
  • the processor of the computer device reads the computer-readable instructions from the computer-readable storage medium, and the processor executes the computer-readable instructions, so that the computer device performs the method provided in various optional ways in Figure 3 or Figure 6 , can be encoded or decoded based on the predicted value of the target attribute, that is, the encoding and decoding processing of the point cloud is implemented, where the candidate point set is obtained from the point cloud group, and the point cloud group is based on the coordinates of the point cloud point
  • the code words are obtained by grouping point cloud points, and the prediction reference points associated with the target point cloud points are obtained from the obtained set of candidate points.
  • the spatial correlation between each point cloud group can be taken into consideration, and further, based on The spatial correlation between each point cloud group is used to predict the corresponding attributes, thereby improving the accuracy of point cloud prediction. And encoding and decoding on this basis can improve the performance and efficiency of encoding and decoding.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)

Abstract

本申请实施例公开了一种点云预测处理方法，涉及大数据领域，该方法包括：获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；从备选点集合中获取目标点云点关联的预测参考点，基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。

Description

点云预测处理方法、装置、计算机、存储介质
本申请要求于2022年03月11日提交中国专利局，申请号为2022102434989、发明名称为“点云预测处理方法、装置、计算机、存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及计算机技术领域，尤其涉及一种点云预测处理方法、装置、计算机、存储介质。
背景技术
现在主流的点云编码技术中，针对不同的点云数据类型，可以分为基于几何结构的点云编码以及基于投影的点云编码。其中，在进行点云编码的过程中，会对点云进行属性预测，一般情况下，会确定初始块大小，基于当前点及初始块大小，确定当前点的空间结构，进而在该空间结构中获取当前点的邻居点，进行属性预测，也就是说，该邻居点的获取是基于当前点(可以称为局部点)所获取的，使得邻居点的获取较为局限，从而降低点云属性预测的性能。
发明内容
本申请实施例提供了一种点云预测处理方法、装置、计算机、存储介质。
本申请实施例一方面提供了一种点云预测处理方法，该方法包括：
获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
从备选点集合中获取目标点云点关联的预测参考点；
基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。
本申请实施例一方面提供了一种点云预测处理方法，该方法包括：
获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
从备选点集合中获取目标点云点关联的预测参考点；
基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值；
获取目标点云点所对应的码流，对目标点云点所对应的码流进行解码处理，得到目标点云点的目标属性残差，基于目标属性预测值与目标属性残差，确定目标点云点的目标属性重建值。
本申请实施例一方面提供了一种点云预测处理装置，该装置包括：
备选集合获取模块，用于获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
参考点获取模块，用于从备选点集合中获取目标点云点关联的预测参考点；
属性预测模块，用于基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。
本申请实施例一方面提供了一种点云预测处理装置，该装置包括：
备选集合获取模块，用于获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
参考点获取模块，用于从备选点集合中获取目标点云点关联的预测参考点；
属性预测模块，用于基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值；
码流获取模块，用于获取目标点云点所对应的码流；
码流解码模块，用于对目标点云点所对应的码流进行解码处理，得到目标点云点的目标属性残差；
属性重建模块，用于基于目标属性预测值与目标属性残差，确定目标点云点的目标属性重建值。
本申请实施例一方面提供了一种计算机设备，包括一个或多个处理器、存储器、输入输出接口；
处理器分别与存储器和输入输出接口相连，其中，输入输出接口用于接收数据及输出数据，存储器用于存储计算机可读指令，处理器用于调用该计算机可读指令，以使包含该处理器的计算机设备执行本申请实施例一方面中的点云预测处理方法。
本申请实施例一方面提供了一个或多个计算机可读存储介质，计算机可读存储介质存储有计算机可读指令，该计算机可读指令适于由处理器加载并执行，以使得具有该处理器的计算机设备执行本申请实施例一方面中的点云预测处理方法。
本申请实施例一方面提供了一种计算机程序产品，该计算机程序产品包括计算机可读指令，该计算机可读指令存储在一个或多个计算机可读存储介质中。计算机设备的一个或多个处理器从计算机可读存储介质读取该计算机可读指令，处理器执行该计算机可读指令，使得该计算机设备执行本申请实施例一方面中的各种可选方式中提供的方法。换句话说，该计算机可读指令被处理器执行时实现本申请实施例一方面中的各种可选方式中提供的方法。
本申请的一个或多个实施例的细节在下面的附图和描述中提出。本申请的其它特征、目的和优点将从说明书、附图以及权利要求书变得明显。
附图说明
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1是本申请实施例提供的一种点云预测处理的网络交互架构图；
图2是本申请实施例提供的一种点云预测处理场景示意图；
图3是本申请实施例提供的一种编码过程中的点云预测处理的方法流程图；
图4是本申请实施例提供的一种点云点分布示意图；
图5是本申请实施例提供的一种可选的点云点分组过程；
图6是本申请实施例提供的一种解码过程中的点云预测处理的方法流程图；
图7是本申请实施例提供的一种数据交互架构图；
图8是本申请实施例提供的一种点云预测处理装置示意图；
图9是本申请实施例提供的一种点云预测处理装置示意图；
图10是本申请实施例提供的一种计算机设备的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
其中，本申请涉及大数据领域，大数据(Big data)是指无法在一定时间范围内用常规软件工具进行捕捉、管理和处理的数据集合，是需要新处理模式才能具有更强的决策力、洞察发现力和流程优化能力的海量、高增长率和多样化的信息资产。随着云时代的来临，大数据也吸引了越来越多的关注，大数据需要特殊的技术，以有效地处理大量的容忍经过时间内的数据。适用于大数据的技术，包括大规模并行处理数据库、数据挖掘、分布式文件系统、分布式数据库、云计算平台、互联网和可扩展的存储系统。例如，可以通过大数据领域中的大数据处理技术及数据计算技术等，对点云点进行分组、预测及编解码等，以提高数据处理的效率。
在本申请实施例中，请参见图1，图1是本申请实施例提供的一种点云预测处理的网络交互架构图。其中，计算机设备101可以从计算机设备101中获取需要进行编码的点云点，对获取到的点云点进行编码；或者，从计算机设备101中获取需要进行解码的码流，对获取到的码流进行解码。或者，计算机设备101可以从其他关联设备中获取需要进行编码的点云点，对获取到的点云点进行编码；或者，从关联设备中获取需要进行解码的码流，对获取到的码流进行解码等。该关联设备的数量为一个 或至少两个，例如，以图1中数量为3为例，如关联设备102a、关联设备102b或关联设备102c等。
具体的，请参见图2，图2是本申请实施例提供的一种点云预测处理场景示意图。如图2所示，计算机设备可以获取至少两个点云点201，基于至少两个点云点201分别对应的坐标码字，对至少两个点云点201进行分组，得到M个点云组，M为正整数，例如图2中，以M为大于或等于3的正整数为例，如点云组1、点云组2及点云组M等。计算机设备可以获取目标点云点所在的目标点云组，在M个点云组中，获取该目标点云组所对应的备选点集合，计算机设备可以从备选点集合中获取目标点云点关联的预测参考点。进一步地，计算机设备可以基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。也就是说，在进行点云预测时，会对需要进行编码的点云点进行分组处理，并在对某一个点云点进行预测时，会从分组得到的业务组中，获取需要进行预测的点云点所在的点云组的备选点集合，从而考虑到各个点云组之间的空间关联性，使得对点云点的属性预测可以包括组间的空间关联性的部分信息，进而提高后续的编解码性能及编解码效率。
可以理解的是，本申请实施例中所提及的关联设备可以是一种计算机设备，本申请实施例中的计算机设备包括但不限于终端设备或服务器。换句话说，计算机设备可以是服务器或终端设备，也可以是服务器和终端设备组成的系统。其中，以上所提及的终端设备可以是一种电子设备，包括但不限于手机、平板电脑、台式电脑、笔记本电脑、掌上电脑、车载设备、增强现实/虚拟现实(Augmented Reality/Virtual Reality，AR/VR)设备、头盔显示器、智能电视、可穿戴设备、智能音箱、数码相机、摄像头及其他具备网络接入能力的移动互联网设备(mobile internet device，MID)，或者火车、轮船、飞行等场景下的终端设备等。其中，以上所提及的服务器可以是独立的物理服务器，也可以是多个物理服务器构成的服务器集群或者分布式系统，还可以是提供云服务、云数据库、云计算、云函数、云存储、网络服务、云通信、中间件服务、域名服务、安全服务、车路协同、内容分发网络(Content Delivery Network，CDN)、以及大数据和人工智能平台等基础云计算服务的云服务器。
在一个实施例中，本申请实施例中所涉及的数据可以存储在计算机设备中，或者可以基于云存储技术对该数据进行存储，在此不做限制。
进一步地，请参见图3，图3是本申请实施例提供的一种编码过程中的点云预测处理的方法流程图。如图3所示，该点云预测处理过程包括如下步骤：
步骤S301，获取目标点云点所在的目标点云组的备选点集合。
在本申请实施例中，备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数。具体的，计算机设备可以获取由k个点云点分组得到的M个点云组，k为正整数，计算机设备可以从M个点云组中获取该目标点云点所在的目标点云组的备选点集合，即，该备选点集合本质上是点云组，该备选点集合的数量可以为一个或至少两个。在一个实施例中，计算机设备可以获取待编码点云，获取组成该待编码点云的k个点云点；或者，计算机设备可以获取空间填充曲线，获取该空间填充曲线中所包括的k个点云点等。具体的需要编码的k个点云点的获取方式在此不做限制。其中，计算机设备可以获取k个点云点以及每个点云点分别对应的坐标码字，将在经过分组移动位数移动后得到的码字序列相同的点云点划分为一组，得到M个点云组。在一个实施例中，该分组移动位数可以是固定的值，也可以是可变的值。例如，将分组移动位数记作L，该L可以是一个定量，即取值不变，该L也可以是一个变量，即取值会随着分组的进行发生变化等，L可以认为是一个正整数。
其中，该第r个点云点的坐标码字可以记作H r，假定坐标码字所对应的坐标维度位数为dig，该坐标码字在每一个坐标维度下的码字长度为s，r为小于或等于k的正整数，dig为正整数，s为正整数，则该第r个点云点的坐标码字可以表示为：
Figure PCTCN2022135899-appb-000001
。其中，coor用于表示坐标，“coor_1”用于表示第一个坐标维度，“coor_dig”用于表示第dig个坐 标维度，下角标用于表示在各个坐标维度下所对应的码字的位数，例如，“s-1”表示在各个坐标维度下的第(s-1)位的编码，换句话说，
Figure PCTCN2022135899-appb-000002
组成了第r个点云点在第一个坐标维度下的编码，
Figure PCTCN2022135899-appb-000003
组成了第r个点云点在第二个坐标维度下的编码等。在一个实施例中，假定dig为3，包括x、y、z三个坐标维度，则第r个点云点的坐标码字可以表示为：
Figure PCTCN2022135899-appb-000004
其中，
Figure PCTCN2022135899-appb-000005
组成了第r个点云点在x这一坐标维度下的编码；
Figure PCTCN2022135899-appb-000006
组成了第r个点云点在y这一坐标维度下的编码；
Figure PCTCN2022135899-appb-000007
组成了第r个点云点在z这一坐标维度下的编码。在一个实施例中，假定dig为4，包括x、y、z、t四个坐标维度，则第r个点云点的坐标码字可以表示为
Figure PCTCN2022135899-appb-000008
等。
举例来说，参见图4，图4是本申请实施例提供的一种点云点分布示意图。如图4所示，以坐标维度位数为3为例，点云点的坐标分布相当于是一个三维空间的分布，例如，图4中所示的三个点云点，该三个点云点属于一个点云组401，也可以说，位于该点云组401中的点云点，均属于该点云组401，其中，位于点云组401中的各个点云点的坐标码字经过移动后，可以得到点云组401所包括的码字范围，即，图4中点云组401的范围可以表示该点云组401的点云组序。
具体的，M个点云组分别对应的分组移动位数均为默认分组移动位数，即，分组移动位数可以是固定的值，其中，该默认分组移动位数可以是基于经验得到的，或者由用户提供的，或者可以是历史分组移动位数等，通过固定的值进行移动分组，使得各个点云点的坐标码字的分组依据相同，在这一情况下，可以提高点云点的分组及后续进行备选点集合的获取的效率。
或者，分组移动位数可以是可变的值。具体的，当与目标点云组相邻的M 1个邻接点云组中分别包含的点云点的数量均值大于第一点数阈值时，目标点云组所对应的分组移动位数，小于M 1个邻接点云组分别对应的分组移动位数；当与目标点云组相邻的M 1个邻接点云组中分别包含的点云点的数量均值小于第二点数阈值时，目标点云组所对应的分组移动位数，大于M 1个邻接点云组分别对应的分组移动位数；M 1为小于M的正整数；当与目标点云组相邻的M 1个邻接点云组中分别包含的点云点的数量均值，大于或等于第二点数阈值，且小于或等于第一点数阈值时，目标点云组的分组移动位数，与目标点云组的前一个点云组的分组移动位数相同。也就是说，当邻接点云组所包含的点云点的数量过大时，可以通过减小分组移动位数，以减少后续生成的点云组中所包含的点云点的数量；当邻接点云组所包含的点云点的数量过少时，可以通过增大分组移动位数，以增多后续生成的点云组中所包含的点云点的数量，可以使得各个点云组所包含的点云点的数量尽可能地达到平衡，以提高点云组分组的效果。
在一个实施例中，M个点云组中可以包括一个或至少两个组间集合，同一个组间集合所包括的点云组的分组移动位数相同，不同的组间集合中的点云组的分组移动位数不同，一个组间集合所包括的点云组的数量小于或等于分组单位阈值，且，当与第j个组间集合中的第一个点云组相邻的M 1个邻接点云组中分别包含的点云点的数量均值大于第一点数阈值时，该第j个组间集合所包括的点云组的分组移动位数，小于M 1个邻接点云组分别对应的分组移动位数；当与第j个组间集合中的第一个点云组相邻的M 1个邻接点云组中分别包含的点云点的数量均值小于第二点数阈值时，该第j个组间集合所包括的点云组的分组移动位数，大于M 1个邻接点云组分别对应的分组移动位数；当与第j个组间集合中的第一个点云组相邻的M 1个邻接点云组中分别包含的点云点的数量均值，大于或等于第二点数阈值，且小于或等于第一点数阈值时，第j个组间集合所包括的点云组的分组移动位数，可以与第j个组间集合的前一个组间集合的分组移动位数相同。使得即可以对分组移动位数进行变更，以提高分组后的点 云点的分布平衡性，同时限制分组移动位数的变更次数，即每得到分组单位阈值个点云组，更新一次分组移动位数，减少需要处理的数据量。举例来说，假定分组单位阈值为5，获取分组移动位数，基于分组移动位数得到第一个点云组至第5个点云组；获取第5个点云组的下一个点云点的M 1个邻接点云组中包含的点云点的数量均值，基于该数量均值更新分组移动位数，基于更新后的分组移动位数，得到第6个点云组至第10个点云组，直至将k个点云点分组完成。在一个实施例中，组间集合可以是一种为了描述分组移动位数变化的一个概念。
在一个实施例中，在M个点云组中获取目标点云组的备选点集合时，一种备选点集合获取方式下，该备选点集合包括在M个点云组中位于目标点云组之前，且与目标点云组相邻的点云组；备选点集合所包括的点云点的总数小于或等于第三点数阈值，在一个实施例中，可以将第三点数阈值记作maxNumofNeighbor。具体的，计算机设备可以在M个点云组中，以目标点云组为基准组，依次向前获取点云组，直至得到备选点集合，该备选点集合所对应的点云组中的点云点的总数小于或等于第三点数阈值，且，该备选点集合与位于该备选点集合之前的点云组所包括的点云点的数量之和大于第三点数阈值。例如，存在(点云组1、点云组2、点云组3、…及点云组M)，假定目标点云组为点云组5，假定第三点数阈值为10，则以目标点云组为基准组，依次获取点云组4，假定该点云组4包括3个点云点，3小于第三点数阈值；继续获取点云组3，假定该点云组3包括5个点云点，此时，点云组4与点云组3一共包括8个点云点，8小于第三点数阈值；继续获取点云组2，假定该点云组2包括4个点云点，此时，点云组4、点云组3及点云组2一共包括12个点云点，12大于第三点数阈值，则将点云组4及点云组3确定为目标点云组的备选点集合。
上述实施例中，通过将M个点云组中位于目标点云组之前，且与目标点云组相邻的至少一个点云组作为备选点集合，可以进一步提升点云预测的准确性。
一种备选点集合获取方式下，备选点集合为在M个点云组中位于目标点云组之前的点云组；每个备选点集合所包括的点云点的数量大于或等于组中点数阈值。在一个实施例中，备选点集合的数量小于或等于组数选取阈值。具体的，计算机设备可以以目标点云组为基准组，在M个点云组中依次向前获取候选点云组，其中，该候选点云组中所包括的点云点的数量大于或等于组中点数阈值，当M个点云组遍历完成时，将获取到的候选点云组，确定为目标点云组的备选点集合。例如，存在(点云组1、点云组2、点云组3、…及点云组M)，假定组中点数阈值为4，目标点云组为点云组5，假定点云组1、点云组2及点云组3分别包括的点云点的数量均大于或等于4，则将点云组1、点云组2及点云组3，确定为目标点云组的备选点集合。或者，计算机设备可以以目标点云组为基准组，在M个点云组中依次向前获取候选点云组，其中，该候选点云组中所包括的点云点的数量大于或等于组中点数阈值，当候选点云组的数量为组数选取阈值，或M个点云组遍历完成时，将获取到的候选点云组，确定为目标点云组的备选点集合。例如，存在(点云组1、点云组2、点云组3、…及点云组M)，假定组中点数阈值为4，目标点云组为点云组5，组数选取阈值为2，假定点云组4包括3个点云点，3小于组中点数阈值；假定点云组3包括5个点云点，5大于组中点数阈值，将点云组3确定为候选点云组，此时，存在一个候选点云组，1小于组数选取阈值；假定该点云组2包括4个点云点，4等于组中点数阈值，将点云组2确定为候选点云组，此时，存在2个候选点云组，2为组数选取阈值，则将点云组2及点云组3，确定为目标点云组的备选点集合。
上述实施例中，通过将M个点云组中位于目标点云组之前的、且包括的点云点的数量大于或等于组中点数阈值的点云组，作为备选点集合，可以进一步提升点云预测的准确性。
一种备选点集合获取方式下，备选点集合为在M个点云组中位于目标点云组之前的N个点云组；N为正整数，N为默认邻接组阈值。默认邻接组阈值，是目标点云组之前、且与目标点云组相邻的点云组的数量阈值。具体的，计算机设备可以以目标点云组为基准组，在M个点云组中向前获取N个点云组，将获取到的N个点云组确定为目标点云组的备选点集合。例如，假定默认邻接组阈值为3，目标点云组为点云组5，则以点云组5为基准组，依次获取3个点云组，即，点云组4、点云组3及点云组2，将点云组4、点云组3及点云组2确定为目标点云组的备选点集合。
上述实施例中，通过将M个点云组中位于目标点云组之前的、且与目标点云组相邻的N个点云组 作为备选点集合，可以进一步提升点云预测的准确性。一种备选点集合获取方式下，一个点云组对应一个点云组序，点云组序是指对应的点云组中所包括的点云点的坐标码字，在经过对应的点云组的分组移动位数移动后得到的。例如，以上述的第r个点云点为例，假定该第r个点云点所在的点云组的分组移动位数为dig，则第r个点云点经过移动后得到的码字序列可以记作：
Figure PCTCN2022135899-appb-000009
也就是说，该第r个点云点所在的点云组中的所有点云点的坐标码字，在经过该分组移动位数移动后，均与第r个点云点的坐标码字移动后得到的码字序列相同。
当目标点云组所对应的目标分组移动位数为坐标维度位数的倍数时，备选点集合所对应的备选点云组序在经过坐标维度位数的第一倍数进行移动后得到的备选移动序列，与目标点云组所对应的目标点云组序在经过坐标维度位数的第一倍数进行移动后得到的目标移动序列相同；坐标维度位数是指每个点云组所包括的点云点的坐标码字所对应的维度数量，即，坐标码字所对应的坐标维度的维度数量。例如，当目标分组移动位数L＝dig*v时，v为正整数，将该目标点云组对应的目标点云组序记作H K1，将该目标点云组记作点云组K1，将该目标点云组序H K1经过坐标维度位数的第一倍数进行移动，记作H K1>>dig*mul 1，获取满足H K2>>dig*mul 1＝H K1>>dig*mul 1的点云组，即获取点云组序在经过坐标维度位数的第一倍数进行移动后得到的移动序列，与目标点云组所对应的目标点云组序在经过坐标维度位数的第一倍数进行移动后得到的目标移动序列相同的点云组，将获取到的点云组确定为目标点云组的备选点集合。其中，mul 1用于表示第一倍数，mul 1为正整数。例如，假定坐标维度位数dig为3，则以上过程可以表示为当L＝3v时，获取H K2>>3＝H K1>>3，则点云组K2为点云组K1的备选点集合，其中，可以认为该点云组K2是点云组K1所在父节点的邻居节点；或者，当H K2>>6＝H K1>>6时，6为3的2倍，点云组K2为点云组K1的备选点集合等。
上述实施例中，在目标点云组所对应的目标分组移动位数为坐标维度位数的倍数的情况下，通过限定备选点集合所对应的备选点云组序在经过坐标维度位数的第一倍数进行移动后得到的备选移动序列，与目标点云组所对应的目标点云组序在经过坐标维度位数的第一倍数进行移动后得到的目标移动序列相同，这样，可以进一步提升点云预测的准确性。
当目标点云组所对应的目标分组移动位数不为坐标维度位数的倍数时，备选点集合所对应的备选点云组序在经过补充维度位数进行移动后得到的备选移动序列，与目标点云组所对应的目标点云组序在经过补充维度位数进行移动后得到的目标移动序列相同；补充维度位数是指目标分组移动位数与坐标维度位数的余数，与坐标维度位数之间的位数差值；或者，补充维度位数是指位数差值，与坐标维度位数的第二倍数之和。其中，可以将第二倍数记作mul 2，mul 2为正整数。例如，当目标分组移动位数L＝dig*v-(dig-1)时，v为正整数，将该目标点云组对应的目标点云组序记作H K1，将该目标点云组记作点云组K1，此时，补充维度位数为(dig-1)或{(dig-1)+dig*mul 2}，获取满足H K2>>(dig-1)＝H K1>>(dig-1)的点云组，或者，获取满足H K2>>{(dig-1)+dig*mul 2}＝H K1>>{(dig-1)+dig*mul 2}的点云组，将获取到的点云组K2，确定为目标点云组的备选点集合；当目标分组移动位数L＝dig*v-(dig-2)时，此时，补充维度位数为(dig-2)或{(dig-2)+dig*mul 2}，获取满足H K2>>(dig-2)＝H K1>>(dig-2)的点云组，，或者，获取满足H K2>>{(dig-2)+dig*mul 2}＝H K1>>{(dig-2)+dig*mul 2}的点云组，将获取到的点云组K2，确定为目标点云组的备选点集合；…；当目标分组移动位数L＝dig*v-1时，此时，补充维度位数为1，获取满足H K2>>1＝H K1>>1的点云组，或者，获取满足H K2>>{1+dig*mul 2}＝H K1>>{1+dig*mul 2}的点云组，将获取到的点云组K2，确定为目标点云组的备选点集合。此时，该点云组K2可以认为是点云组K1的邻居节点。例如，假定坐标维度位数dig为3，目标分组移动位数L为2，则补充维度位数可以是1，也可以是(1+3＝4)等，例如，当满足H K2>>1＝H K1>>1时，可以认为点云组K2为点云组K1的备选点集合；或者，当满足H K2>>4＝H K1>>4时，可以认为点云组K2为点云组K1的备选点集合等。其中，“>>”表示移动，例如，假定H K1为001101，则经过H K1>>1后，得到00110。
上述实施例中，在目标点云组所对应的目标分组移动位数为坐标维度位数的倍数的情况下，通过 限定备选点集合所对应的备选点云组序在经过坐标维度位数的第一倍数进行移动后得到的备选移动序列，与目标点云组所对应的目标点云组序在经过坐标维度位数的第一倍数进行移动后得到的目标移动序列相同，这样，可以进一步提升点云预测的准确性。
如图4所示，假定目标点云组为点云组401，经过二次移动后，得到的目标移动序列可以通过区域402所示，可以将位于该区域402中的点云组，确定为点云组401的备选点集合。也就是说，该区域402可以认为是点云组401的父节点。在一个实施例中，该区域402还包括6个共面邻居节点，如图4中虚线框所指示的区域，还可以包括12个共线邻居节点以及8个共点邻居节点等。在一个实施例中，计算机设备可以基于需要，从目标点云组的父节点的共面邻居节点、共线邻居节点或共点邻居节点中，获取该目标点云组的备选点集合。具体的，该共面邻居节点的数量、共线邻居节点的数量及共点邻居节点的数量，是由坐标维度位数dig所确定的，在此不做限制。在一个实施例中，备选点集合的获取过程可以是通过上述各个备选点集合获取方式中的任意一个方式或任意多个方式组合实现的。
步骤S302，从备选点集合中获取目标点云点关联的预测参考点。
在本申请实施例中，计算机设备可以从备选点集合中获取目标点云点关联的预测参考点。其中，备选点集合的数量为一个或至少两个，当备选点集合的数量为一个时，可以直接将备选点集合中所包括的点云点，确定为目标点云点关联的预测参考点；或者，获取备选点集合中所包括的点云点与目标点云点之间的点间距离，基于备选点集合中所包括的点云点与目标点云点之间的点间距离，从备选点集合中获取目标点云点关联的预测参考点等。其中，点间距离可以是备选点集合中所包括的点云点的坐标与目标点云点的坐标之间的距离，也可以称为几何距离，或者，可以是备选点集合中所包括的点云点与目标点云点在M个点云组中的间隔点数量等。
其中，备选点集合的数量为至少两个。具体的，一种预测参考点获取方式下，计算机设备可以在至少两个备选点集合中，分别选择d个点云点，将至少两个备选点集合分别对应的d个点云点，确定为目标点云点关联的预测参考点；d为正整数。例如，假定d为1，则从至少两个备选点集合中分别获取一个点云点，作为该目标点云点关联的预测参考点。在一个实施例中，在该方式下，可以认为对于同一个点云组中的点云点来说，预测参考点是相同的，因此，计算机设备可以查找该目标点云组所对应的预测参考点，若获取到目标点云组所对应的预测参考点，则将该目标点云组所对应的预测参考点，确定为目标点云点的预测参考点；若未获取到目标点云组所对应的预测参考点，则在至少两个备选点集合中，分别选择d个点云点，将至少两个备选点集合分别对应的d个点云点，确定为目标点云点关联的预测参考点，同时，可以将该至少两个备选点集合分别对应的d个点云点，确定为目标点云组所对应的预测参考点；通过以上过程，可以使得对于同一个点云组，只需要获取一次预测参考点，从而减少需要处理的数据量，提高预测参考点的获取效率。
一种预测参考点获取方式下，在至少两个备选点集合中，分别选择d个点云点，作为候选点云点，获取候选点云点与目标点云点之间的第一点间距离，基于第一点间距离对候选点云点进行排序，从排序后的候选点云点中获取目标点云点关联的预测参考点。其中，第一点间距离可以是候选点云点的坐标与目标点云点的坐标之间的距离，也可以称为几何距离，或者，可以是候选点云点与目标点云点在M个点云组中的间隔点数量等。
上述实施例中，通过候选点云点与目标点云点之间的第一点间距离，对候选点云点进行排序，并从排序后的候选点云点中直接获取目标点云点关联的预测参考点，可以提高预测参考点的获取效率。
一种预测参考点获取方式下，获取至少两个备选点集合所包括的点云点与目标点云点之间的第二点间距离，基于第二点间距离对至少两个备选点集合所包括的点云点进行排序，从排序后的至少两个备选点集合所包括的点云点中，获取目标点云点关联的预测参考点。其中，第二点间距离可以是至少两个备选点集合中所包括的点云点的坐标与目标点云点的坐标之间的距离，也可以称为几何距离，或者，可以是至少两个备选点集合中所包括的点云点与目标点云点在M个点云组中的间隔点数量等。例如，对于备选点集合中的点云点1来说，该第二点间距离可以是点云点1的坐标与目标点云点的坐标之间的几何距离，或者，可以是点云点1与目标点云点分别在M个点云组中的位置之间的间隔点数量等。
上述实施例中，通过至少两个备选点集合所包括的点云点与目标点云点之间的第二点间距离，对至少两个备选点集合所包括的点云点进行排序，并从排序后的至少两个备选点集合所包括的点云点中，直接获取目标点云点关联的预测参考点，可以提高预测参考点的获取效率。
一种预测参考点获取方式下，获取至少两个备选点集合分别对应的集合优先级，基于集合优先级对至少两个备选集合进行排序，在排序后的至少两个备选集合中，获取目标点云点关联的P个预测参考点；P为正整数。在一个实施例中，计算机设备可以获取至少两个备选点集合分别与目标点云组之间的组间关联关系，基于组间关联关系确定至少两个备选点集合分别对应的集合优先级。其中，组间关联关系包括但不限于邻居关联关系、父级关联关系及间隔关联关系，例如，该邻居关联关系可以认为是与目标点云组相邻的点云组与该目标点云组之间的组间关联关系；父级关联关系可以认为是父节点相同的点云组之间的组间关联关系，也就是说，对具有父级关联关系的两个点云组的点云组序进行二次移动后是相同的；间隔关联关系可以认为是位于M个点云组中，不相邻的点云组之间的组间关联关系等。在一个实施例中，至少两个备选点集合分别对应的集合优先级也可以是提前预设的。或者，若至少两个备选点集合是通过多个备选点集合获取方法所得到的，则可以基于至少两个备选点集合分别对应的备选点集合获取方法，确定至少两个备选点集合分别对应的集合优先级等。即，集合优先级的获取方式在此不做限制。
上述实施例中，通过至少两个备选点集合分别对应的集合优先级，对至少两个备选集合进行排序，并在排序后的至少两个备选集合中，直接获取目标点云点关联的预测参考点，可以提高预测参考点的获取效率。
步骤S303，基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。
在本申请实施例中，计算机设备可以获取预测参考点的预测参考坐标，获取目标点云点的目标坐标。基于预测参考坐标与目标坐标之间的坐标距离，确定预测参考点的参考权重。具体的，计算机设备可以获取将该坐标距离的倒数，确定为该坐标距离所对应的预测参考点的参考权重。在一个实施例中，计算机设备可以将预测参考坐标与目标坐标在各个坐标维度下的坐标差值之和，确定为预测参考坐标与目标坐标之间的坐标距离。此时，该预测参考点的参考权重可以参见公式①所示：
Figure PCTCN2022135899-appb-000010
如公式①所示，w iu用于表示第i个点云点的第u个预测参考点与第i个点云点之间的参考权重，下角标iu用于表示第i个点云点的第u个预测参考点，例如，coor_2 iu用于表示第i个点云点的第u个预测参考点在第二个坐标维度下的坐标值，该第u个预测参考点在第二个坐标维度下的坐标值，与第i个点云点在第二个坐标维度下的坐标值之间的差值的绝对值，可以为第u个预测参考点与第i个点云点在第二个坐标维度下的坐标差值，即|coor_2 i-coor_2 iu|。同理，可以得到第i个点云点，与该第i个点云点的第u个预测参考点在各个坐标维度下的坐标差值。其中，i为小于或等于k的正整数。例如，假定dig为3，即坐标维度位数为3，包括x、y、z三个坐标维度，则该预测参考点的参考权重可以参见公式②所示：
Figure PCTCN2022135899-appb-000011
在一个实施例中，计算机设备可以获取dig个坐标维度分别对应的维度权重，基于dig个坐标维度分别对应的维度权重，对预测参考坐标与目标坐标在各个坐标维度下的坐标差值进行加权求和，得到预测参考坐标与目标坐标之间的坐标距离。此时，假定dig为3，即坐标维度位数为3，包括x、y、z三个坐标维度，则该预测参考点的参考权重可以参见公式③所示：
Figure PCTCN2022135899-appb-000012
如公式③所示，a用于x坐标维度对应的维度权重，b用于表示y坐标维度对应的坐标维度，c用于表示z坐标维度对应的维度权重。
在一个实施例中，预测参考点与目标点云点之间的坐标距离并不仅限于上述计算方式，例如，以三个坐标维度(x坐标维度、y坐标维度及z坐标维度)为例，也可以通过公式
Figure PCTCN2022135899-appb-000013
得到第i个点云点与第i个点云点的第u个预测参考点之间的坐标距离等，在此不做限制。
进一步地，可以获取预测参考点的参考属性重建值，基于参考属性重建值与参考权重进行加权处理，得到目标点云点的目标属性预测值。其中，该目标属性预测值的获取可以参见公式④所示：
Figure PCTCN2022135899-appb-000014
如公式④所示，该
Figure PCTCN2022135899-appb-000015
用于表示第i个点云点的第u个预测参考点的参考属性重建值。w iu用于表示第i个点云点的第u个预测参考点与第i个点云点之间的参考权重，num用于表示第i个点云点的预测参考点的总数。
其中，可以通过上述公式①至公式④，得到k个点云点中的任意一个点云点的属性预测值。而目标点云点是k个点云点中的任意一个点云点，因此，可以通过上述公式①至公式④，得到目标点云点的目标属性预测值。
在一个实施例中，在基于参考属性重建值与参考权重进行加权处理，得到目标点云点的目标属性预测值时，也可以采用优化参数对参考权重进行优化处理，得到优化权重，基于参考属性重建值与优化权重进行加权处理，得到目标点云点的目标属性预测值。其中，该优化参数可以包括但不限于属性量化步长，或者，与目标点云点之间的坐标距离最大的预测参考点的参考数量等。例如，计算机设备可以将属性量化步长确定为优化参数。或者，在预测参考点中获取与目标点云点之间的坐标距离最大的预测参考点的参考数量，将该参考数量确定为优化参数。或者，可以从上述得到的属性量化步长及参考数量中，获取较小的数作为优化参数，即，当属性量化步长大于参考数量时，将参考数量确定为优化参数；当属性量化步长小于参考数量时，将属性量化步长确定为优化参数；当两者相等时，将属性量化步长或参考数量，确定为优化参数等。
上述实施例中，通过预测参考点的预测参考坐标与目标点云点的目标坐标之间的坐标距离，确定预测参考点的参考权重，并通过预测参考点的参考属性重建值与参考权重进行加权处理，得到目标点云点的目标属性预测值，可以提升目标属性预测值的预测准确率。
进一步地，计算机设备可以获取目标点云点的目标属性实际值，基于目标属性实际值与目标点云点的目标属性预测值之间的差值，得到目标点云点的目标属性残差。对目标属性残差进行量化转换，得到目标点云点的目标变换系数。该量化转换方式在此不做限制，例如，可以对目标属性残差进行离散余弦变换(Discrete Cosine Transform，DCT)，得到目标点云点的目标变换系数，或者，可以通过构建二叉树得到目标点云点的目标变换系数。其中，该目标变换系数包括第一变换系数及第二变换系数。在一个实施例中，计算机设备也可以获取到目标点云组所包括的点云点的属性残差，对该目标点云组所包括的点云点的属性残差进行量化转换，得到目标点云组所包括的点云点的变换系数，包括目标点云点的目标变换系数。
上述实施例中，通过目标点云点的目标属性实际值与目标点云点的目标属性预测值之间的差值，得到目标点云点的目标属性残差，并对目标属性残差进行量化转换，得到目标点云点的目标变换系数，以更好的对目标点云点进行编码，进一步提高编码性能及编码效率。
进一步地，计算机设备可以逐组进行编码。具体的，获取目标点云组所包括的点云点的变换系数，对目标点云组所包括的点云点的变换系数进行编码，得到目标点云组所对应的目标码流。
上述实施例中，通过直接对目标点云组所包括的点云点的变换系数进行编码，得到目标点云组所 对应的目标码流，提高编码结果的获取效率，进而提高编码性能及编码效率。
或者，可以获取组数限制阈值，基于组数限制阈值获取包含目标点云组的g个待编码点云组，获取g个待编码点云组所包括的点云点的变换系数，对g个待编码点云组所包括的点云点的变换系数进行编码，得到g个待编码点云组所对应的组码流；g为正整数，g小于或等于组数限制阈值。通过对一次需要编码的点云组的数量进行限制，可以不需要等到所有的点云点编码完成，便可以得到部分的编码结果，从而减少编码过程中出现异常情况(如异常中断等)造成的编码失败情况，提高编码结果的获取效率，进而提高编码性能及编码效率。
在本申请实施例中，获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；从备选点集合中获取目标点云点关联的预测参考点，基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。在此基础上，可以基于目标属性预测值进行编码或解码，即，实现对点云的编解码处理，其中，备选点集合是从点云组中所获取的，该点云组是基于点云点的坐标码字对点云点进行分组得到的，而从获取到的备选点集合中获取目标点云点关联的预测参考点，可以考虑到各个点云组之间的空间关联性，进而，可以基于各个点云组之间的空间关联性进行相应的属性预测，从而提高点云预测的准确性。并且在此基础上进行编解码，可以提高编解码的性能及效率。
在一个实施例中，当分组移动位数为可变的值时，可以参见图5，图5是本申请实施例提供的一种在一个实施例中点云点分组过程。如图5所示，该点云点分组过程包括如下步骤：
步骤S501，获取k个点云点及每个点云点的坐标码字。
在本申请实施例中，计算机设备可以获取k个点云点及每个点云点的坐标码字，例如，计算机设备可以获取空间填充曲线的空间填充曲线码字，从空间填充曲线码字中获取k个点云点分别对应的坐标码字。
步骤S502，获取分组移动位数。
在本申请实施例中，可以获取分组移动位数L，在该实施例中，L为可变的值。
步骤S503，基于分组移动位数，对第i个点云点的坐标码字进行移动，得到第i个点云点的码字序列。
在本申请实施例中，计算机设备可以对第i个点云点的坐标码字进行移动，得到第i个点云点的码字序列。举例来说，假定该第i个点云点的坐标码字为“0010101110”，此时的分组移动位数为3，则对第i个点云点的坐标码字进行移动，可以得到第i个点云点的码字序列，为“0010101”。
步骤S504，第i个码字序列与第(i-1)个码字序列是否相同。
在本申请实施例中，计算机设备可以检测第i个码字序列与第(i-1)个码字序列是否相同，其中，当i为1时，将第(i-1)个点云点的码字序列默认为空或特殊标识。若第i个点云点的码字序列与第(i-1)个点云点的码字序列相同，则执行步骤S505；若第i个点云点的码字序列与第(i-1)个点云点的码字序列不同，则执行步骤S508。
步骤S505，将第i个点云点添加至第(i-1)个点云点所在的初始点云组。
在本申请实施例中，若第i个点云点的码字序列与第(i-1)个点云点的码字序列相同，则表示第i个点云点可以与第(i-1)个点云点划分至一个点云组，可以将第i个点云点添加至第(i-1)个点云点所在的初始点云组。具体的，若第(i-1)个点云点不存在初始点云组，则可以基于第i个点云点及第(i-1)个点云点创建一个初始点云组；若第(i-1)个点云点存在初始点云组，则将第i个点云点添加至第(i-1)个点云点所在的初始点云组。例如，i为4，若第4个点云点的码字序列与第3个点云点的码字序列相同，则将第4个点云点添加至第3个点云点所在的初始点云组。若该第3个点云点不存在初始点云组，则可以基于第3个点云点与第4个点云点创建一个初始点云组；若该第3个点云点存在初始点云组，则可以将第4个点云点添加至第3个点云点所在的初始点云组。
步骤S506，i＝＝k？。
在本申请实施例中，检测i是否为k，即检测k个点云点是否分组完成，若i为k，则执行步骤 S513；若i不为k，则执行步骤S507。
步骤S507，i++。
在本申请实施例中，对i的值加一，即处理下一个点云点。
步骤S508，生成当前点云组，i++。
在本申请实施例中，可以获取第(i-1)个点云点所在的初始点云组，将该第(i-1)个点云点所在的初始点云组确定为一个点云组。在一个实施例中，若第(i-1)个点云点不存在初始点云组，则将第(i-1)个点云点组成一个点云组。进一步地，可以检测i是否为k，若i为k，则执行步骤S513；若i不为k，则对i进行更新，即i++，执行步骤S509。
步骤S509，组统计值e是否为分组单位阈值。
在本申请实施例中，检测组统计值e是否为分组单位阈值，若e为分组单位阈值，则执行步骤S511；若e不为分组单位阈值，则执行步骤S510。其中，e的初始值为0。
步骤S510，e++。
在本申请实施例中，对e的值加一，返回执行步骤S503。
步骤S511，重置组统计值e。
在本申请实施例中，重置组统计值e，即，将组统计值e重置为初始值，执行步骤S512。
步骤S512，当前点云组的M1个邻接点云组中包含的点云点的数量均值大于第一点数阈值时，减小分组移动位数；当前点云组的M1个邻接点云组中包含的点云点的数量均值小于第二点数阈值时，增大分组移动位数。
在本申请实施例中，更新分组移动位数，即更新L的值，返回执行步骤S503。
步骤S513，生成当前点云组，结束分组。
在本申请实施例中，基于第i个点云点所在的初始点云组，生成当前点云组；在一个实施例中，若第i个点云点不存在初始点云组，则基于第i个点云点生成当前点云组。此时，得到由k个点云点组成的M个点云组，结束分组。
进一步地，请参见图6，图6是本申请实施例提供的一种解码过程中的点云预测处理的方法流程图。如图6所示，该点云预测处理过程包括如下步骤：
步骤S601，获取目标点云点所在的目标点云组的备选点集合。
在本申请实施例中，备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数。具体实现过程可以参见图3中的步骤S301所示的具体描述。
步骤S602，从备选点集合中获取目标点云点关联的预测参考点。
在本申请实施例中，具体实现过程可以参见图3中的步骤S302所示的具体描述。
步骤S603，基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。
在本申请实施例中，具体实现过程可以参见图3中的步骤S303所示的具体描述。
基于上述过程，可以得到k个点云点中任意一个点云点的属性预测值。k为正整数，目标点云点是指k个点云点中的任意一个点云点。
步骤S604，获取目标点云点所对应的码流，对目标点云点所对应的码流进行解码处理，得到目标点云点的目标属性残差。
在本申请实施例中，对目标点云点所对应的码流进行解码处理，得到目标点云点的目标变换系数。对目标变换系数进行反变换处理，得到目标点云点的目标属性残差。
上述实施例中，通过直接对目标点云点的目标变换系数进行反变换处理，得到目标点云点的目标属性残差，可以提升目标属性残差的获取效率。
在一个实施例中，当基于组数限制阈值进行编码时，该目标点云点所对应的码流是指该包括目标点云点所在的目标点云组在内的多个点云组的码流；或者，当该码流是逐组进行编码时，该目标点云点所对应的码流是指目标点云组的码流等。具体的，计算机设备可以对目标点云点所对应的码流进行解码处理，得到该码流所对应的变换系数序列，对该码流对应的变换系数序列进行反排序，得到该码 流所对应的点云组中各个点云点的变换系数，包括目标点云点的目标变换系数。其中，该反排序是基于编码过程中对变换系数的排序方式进行还原的一种排序方式。
步骤S605，基于目标属性预测值与目标属性残差，确定目标点云点的目标属性重建值。
在本申请实施例中，可以将目标属性预测值与目标属性残差进行属性融合，得到目标点云点的目标属性重建值。该属性融合可以是属性相加等。
其中，上述步骤S601至步骤S603，与步骤S604的执行顺序并不做限制，即，也可以先执行步骤S604，再执行步骤S601至步骤S603等。
在本申请实施例中，获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；从备选点集合中获取目标点云点关联的预测参考点，基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。在此基础上，可以基于目标属性预测值进行编码或解码，即，实现对点云的编解码处理，其中，备选点集合是从点云组中所获取的，该点云组是基于点云点的坐标码字对点云点进行分组得到的，而从获取到的备选点集合中获取目标点云点关联的预测参考点，可以考虑到各个点云组之间的空间关联性，进而，可以基于各个点云组之间的空间关联性进行相应的属性预测，从而提高点云预测的准确性。并且在此基础上进行编解码，可以提高编解码的性能及效率。
其中，上述编码过程与解码过程可以在同一个计算机设备中实现，也可以在不同的计算机设备中实现。例如，在不同的计算机设备中实现时，可以参见图7，图7是本申请实施例提供的一种数据交互架构图。如图7所示，计算机设备701可以对k个点云点进行编码，得到编码得到的码流，其中，该k个点云点所对应的码流的数量可以为一个或至少两个。计算机设备701可以将编码得到的码流发送至计算机设备702，计算机设备702可以对获取到的码流进行解码，从而可以得到k个点云点，具体可以得到k个点云点分别对应的属性重建值。在一个实施例中，计算机设备701可以从计算机设备701中获取k个点云点，也可以从计算机设备702中获取k个点云点，或者，可以从其他的关联设备中获取k个点云点，在此不做限制。
进一步地，请参见图8，图8是本申请实施例提供的一种点云预测处理装置示意图。该点云预测处理装置可以是运行于计算机设备中的一个计算机可读指令(包括程序代码等)，例如该点云预测处理装置可以为一个应用软件；该装置可以用于执行本申请实施例提供的方法中的相应步骤。如图8所示，该点云预测处理装置800可以用于图3所对应实施例中的计算机设备，具体的，该装置可以包括：备选集合获取模块11、参考点获取模块12及属性预测模块13。
备选集合获取模块11，用于获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
参考点获取模块12，用于从备选点集合中获取目标点云点关联的预测参考点；
属性预测模块13，用于基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。
其中，M个点云组分别对应的分组移动位数均为默认分组移动位数；或者，
当与目标点云组相邻的M 1个邻接点云组中分别包含的点云点的数量均值大于第一点数阈值时，目标点云组所对应的分组移动位数，小于M 1个邻接点云组分别对应的分组移动位数；当与目标点云组相邻的M 1个邻接点云组中分别包含的点云点的数量均值小于第二点数阈值时，目标点云组所对应的分组移动位数，大于M 1个邻接点云组分别对应的分组移动位数；M 1为小于M的正整数。
其中，备选点集合为在M个点云组中位于目标点云组之前，且与目标点云组相邻的点云组；备选点集合所包括的点云点的总数小于或等于第三点数阈值。
其中，备选点集合为在M个点云组中位于目标点云组之前的点云组；每个备选点集合所包括的点云点的数量大于或等于组中点数阈值。
其中，备选点集合为在M个点云组中位于目标点云组之前的N个点云组；N为正整数，N为默认邻 接组阈值。
其中，一个点云组对应一个点云组序，点云组序是指对应的点云组中所包括的点云点的坐标码字，在经过对应的点云组的分组移动位数移动后得到的；
当目标点云组所对应的目标分组移动位数为坐标维度位数的倍数时，备选点集合所对应的备选点云组序在经过坐标维度位数的第一倍数进行移动后得到的备选移动序列，与目标点云组所对应的目标点云组序在经过坐标维度位数的第一倍数进行移动后得到的目标移动序列相同；坐标维度位数是指每个点云组所包括的点云点的坐标码字所对应的维度数量；
当目标点云组所对应的目标分组移动位数不为坐标维度位数的倍数时，备选点集合所对应的备选点云组序在经过补充维度位数进行移动后得到的备选移动序列，与目标点云组所对应的目标点云组序在经过补充维度位数进行移动后得到的目标移动序列相同；补充维度位数是指目标分组移动位数与坐标维度的余数，与坐标维度之间的位数差值；或者，补充维度位数是指位数差值，与坐标维度位数的第二倍数之和。
其中，备选点集合的数量为至少两个；该参考点获取模块12，包括：
备选获取单元121，用于在至少两个备选点集合中，分别选择d个点云点，将至少两个备选点集合分别对应的d个点云点，确定为目标点云点关联的预测参考点；d为正整数。
其中，备选点集合的数量为至少两个；该参考点获取模块12，包括：
候选获取单元122，用于在至少两个备选点集合中，分别选择d个点云点，作为候选点云点；
距离选择单元123，用于获取候选点云点与目标点云点之间的第一点间距离，基于第一点间距离对候选点云点进行排序，从排序后的候选点云点中获取目标点云点关联的预测参考点。
其中，备选点集合的数量为至少两个；该参考点获取模块12，包括：
备选排序单元124，用于获取至少两个备选点集合所包括的点云点与目标点云点之间的第二点间距离，基于第二点间距离对至少两个备选点集合所包括的点云点进行排序，从排序后的至少两个备选点集合所包括的点云点中，获取目标点云点关联的预测参考点。
其中，备选点集合的数量为至少两个；该参考点获取模块12，包括：
优先级获取单元125，用于获取至少两个备选点集合分别对应的集合优先级，基于集合优先级对至少两个备选集合进行排序，在排序后的至少两个备选集合中，获取目标点云点关联的P个预测参考点；P为正整数。
其中，该属性预测模块13，包括：
坐标获取单元131，用于获取预测参考点的预测参考坐标，获取目标点云点的目标坐标；
权重获取单元132，用于基于预测参考坐标与目标坐标之间的坐标距离，确定预测参考点的参考权重；
属性预测单元133，用于获取预测参考点的参考属性重建值，基于参考属性重建值与参考权重进行加权处理，得到目标点云点的目标属性预测值。
其中，该装置800还包括：
残差获取模块14，用于获取目标点云点的目标属性实际值，基于目标属性实际值与目标点云点的目标属性预测值之间的差值，得到目标点云点的目标属性残差；
量化处理模块15，用于对目标属性残差进行量化转换，得到目标点云点的目标变换系数。
其中，该装置800还包括：
第一编码模块16，用于获取目标点云组所包括的点云点的变换系数，对目标点云组所包括的点云点的变换系数进行编码，得到目标点云组所对应的目标码流；或者，
第二编码模块17，用于获取组数限制阈值，基于组数限制阈值获取包含目标点云组的g个待编码点云组，获取g个待编码点云组所包括的点云点的变换系数，对g个待编码点云组所包括的点云点的变换系数进行编码，得到g个待编码点云组所对应的组码流；g为正整数，g小于或等于组数限制阈值。
进一步地，请参见图9，图9是本申请实施例提供的一种点云预测处理装置示意图。该点云预测处理装置可以是运行于计算机设备中的一个计算机可读指令(包括程序代码等)，例如该点云预测处理装 置可以为一个应用软件；该装置可以用于执行本申请实施例提供的方法中的相应步骤。如图9所示，该点云预测处理装置900可以用于图6所对应实施例中的计算机设备，具体的，该装置可以包括：备选集合获取模块21、参考点获取模块22、属性预测模块23、码流获取模块24、码流解码模块25及属性重建模块26。
备选集合获取模块21，用于获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
参考点获取模块22，用于从备选点集合中获取目标点云点关联的预测参考点；
属性预测模块23，用于基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值；
码流获取模块24，用于获取目标点云点所对应的码流；
码流解码模块25，用于对目标点云点所对应的码流进行解码处理，得到目标点云点的目标属性残差；
属性重建模块26，用于基于目标属性预测值与目标属性残差，确定目标点云点的目标属性重建值。
其中，该码流解码模块25，包括：
初始解码单元251，用于对目标点云点所对应的码流进行解码处理，得到目标点云点的目标变换系数；
系数变换单元252，用于对目标变换系数进行反变换处理，得到目标点云点的目标属性残差。
本申请实施例提供了一种点云预测处理装置，该装置可以获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；从备选点集合中获取目标点云点关联的预测参考点，基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。在此基础上，可以基于目标属性预测值进行编码或解码，即，实现对点云的编解码处理，其中，备选点集合是从点云组中所获取的，该点云组是基于点云点的坐标码字对点云点进行分组得到的，而从获取到的备选点集合中获取目标点云点关联的预测参考点，可以考虑到各个点云组之间的空间关联性，进而，可以基于各个点云组之间的空间关联性进行相应的属性预测，从而提高点云预测的准确性。并且在此基础上进行编解码，可以提高编解码的性能及效率。
参见图10，图10是本申请实施例提供的一种计算机设备的结构示意图。如图10所示，本申请实施例中的计算机设备可以包括：一个或多个处理器1001、存储器1002和输入输出接口1003。该处理器1001、存储器1002和输入输出接口1003通过总线1004连接。存储器1002用于存储计算机可读指令，该计算机可读指令包括程序指令，输入输出接口1003用于接收数据及输出数据，如用于数据交互；处理器1001用于执行存储器1002存储的程序指令。
其中，该处理器1001在进行编码时，可以执行如下操作：
获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
从备选点集合中获取目标点云点关联的预测参考点，基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。
该处理器1001在进行解码时，可以执行如下操作：
获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
从备选点集合中获取目标点云点关联的预测参考点，基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值；
获取目标点云点所对应的码流，对目标点云点所对应的码流进行解码处理，得到目标点云点的目 标属性残差，基于目标属性预测值与目标属性残差，确定目标点云点的目标属性重建值。
在一些可行的实施方式中，该处理器1001可以是中央处理单元(central processing unit，CPU)，该处理器还可以是其他通用处理器、数字信号处理器(digital signal processor，DSP)、专用集成电路(application specific integrated circuit，ASIC)、现成可编程门阵列(field-programmable gate array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
该存储器1002可以包括只读存储器和随机存取存储器，并向处理器1001和输入输出接口1003提供指令和数据。存储器1002的一部分还可以包括非易失性随机存取存储器。例如，存储器1002还可以存储设备类型的信息。
具体实现中，该计算机设备可通过其内置的各个功能模块执行如该图3或图6中各个步骤所提供的实现方式，具体可参见该图3或图6中各个步骤所提供的实现方式，在此不再赘述。
本申请实施例通过提供一种计算机设备，包括：处理器、输入输出接口、存储器，通过处理器获取存储器中的计算机可读指令，执行该图3中所示方法的各个步骤，进行点云预测处理操作。本申请实施例实现了获取目标点云点所在的目标点云组的备选点集合；备选点集合属于M个点云组，M个点云组包括目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；从备选点集合中获取目标点云点关联的预测参考点，基于预测参考点对目标点云点进行预测处理，得到目标点云点的目标属性预测值。在此基础上，可以基于目标属性预测值进行编码或解码，即，实现对点云的编解码处理，其中，备选点集合是从点云组中所获取的，该点云组是基于点云点的坐标码字对点云点进行分组得到的，而从获取到的备选点集合中获取目标点云点关联的预测参考点，可以考虑到各个点云组之间的空间关联性，进而，可以基于各个点云组之间的空间关联性进行相应的属性预测，从而提高点云预测的准确性。并且在此基础上进行编解码，可以提高编解码的性能及效率。
本申请实施例还提供一种计算机可读存储介质，该计算机可读存储介质存储有计算机可读指令，该计算机可读指令适于由该处理器加载并执行图3或图6中各个步骤所提供的点云预测处理方法，具体可参见该图3或图6中各个步骤所提供的实现方式，在此不再赘述。另外，对采用相同方法的有益效果描述，也不再进行赘述。对于本申请所涉及的计算机可读存储介质实施例中未披露的技术细节，请参照本申请方法实施例的描述。作为示例，计算机可读指令可被部署为在一个计算机设备上执行，或者在位于一个地点的多个计算机设备上执行，又或者，在分布在多个地点且通过通信网络互连的多个计算机设备上执行。
该计算机可读存储介质可以是前述任一实施例提供的点云预测处理装置或者该计算机设备的内部存储单元，例如计算机设备的硬盘或内存。该计算机可读存储介质也可以是该计算机设备的外部存储设备，例如该计算机设备上配备的插接式硬盘，智能存储卡(smart media card，SMC)，安全数字(secure digital，SD)卡，闪存卡(flash card)等。进一步地，该计算机可读存储介质还可以既包括该计算机设备的内部存储单元也包括外部存储设备。该计算机可读存储介质用于存储该计算机可读指令以及该计算机设备所需的其他程序和数据。该计算机可读存储介质还可以用于暂时地存储已经输出或者将要输出的数据。
本申请实施例还提供了一种计算机可读指令产品或计算机可读指令，该计算机可读指令产品或计算机可读指令包括计算机可读指令，该计算机可读指令存储在计算机可读存储介质中。计算机设备的处理器从计算机可读存储介质读取该计算机可读指令，处理器执行该计算机可读指令，使得该计算机设备执行图3或图6中的各种可选方式中所提供的方法，可以基于目标属性预测值进行编码或解码，即，实现对点云的编解码处理，其中，备选点集合是从点云组中所获取的，该点云组是基于点云点的坐标码字对点云点进行分组得到的，而从获取到的备选点集合中获取目标点云点关联的预测参考点，可以考虑到各个点云组之间的空间关联性，进而，可以基于各个点云组之间的空间关联性进行相应的属性预测，从而提高点云预测的准确性。并且在此基础上进行编解码，可以提高编解码的性能及效率。以上所揭露的仅为本申请较佳实施例而已，当然不能以此来限定本申请之权利范围，因此依本申请权 利要求所作的等同变化，仍属本申请所涵盖的范围。

Claims (23)

  1. 一种点云预测处理方法，其特征在于，由计算机设备执行；所述方法包括：
    获取目标点云点所在的目标点云组的备选点集合；所述备选点集合属于M个点云组，所述M个点云组包括所述目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
    从所述备选点集合中获取所述目标点云点关联的预测参考点；
    基于所述预测参考点对所述目标点云点进行预测处理，得到所述目标点云点的目标属性预测值。
  2. 如权利要求1所述的方法，其特征在于，所述M个点云组分别对应的分组移动位数均为默认分组移动位数。
  3. 如权利要求1所述的方法，其特征在于，所述M个点云组分别对应的分组移动位数满足以下条件：
    当与所述目标点云组相邻的M 1个邻接点云组中分别包含的点云点的数量均值大于第一点数阈值时，所述目标点云组所对应的所述分组移动位数，小于所述M 1个邻接点云组分别对应的分组移动位数；当与所述目标点云组相邻的M 1个邻接点云组中分别包含的点云点的数量均值小于第二点数阈值时，所述目标点云组所对应的所述分组移动位数，大于所述M 1个邻接点云组分别对应的分组移动位数；M 1为小于M的正整数；所述第一点数阈值大于所述第二点数阈值。
  4. 如权利要求1所述的方法，其特征在于，所述备选点集合包括在所述M个点云组中位于所述目标点云组之前，且与目标点云组相邻的至少一个点云组；所述备选点集合所包括的点云点的总数小于或等于第三点数阈值。
  5. 如权利要求1所述的方法，其特征在于，所述备选点集合包括在所述M个点云组中位于所述目标点云组之前的点云组；每个所述位于所述目标点云组之前的点云组所包括的点云点的数量大于或等于组中点数阈值；所述组中点数阈值，是针对所述M个点云组中各个点云组统一设置的阈值。
  6. 如权利要求1所述的方法，其特征在于，所述备选点集合包括在所述M个点云组中位于所述目标点云组之前的N个点云组；N为正整数，N为默认邻接组阈值；所述默认邻接组阈值，是所述目标点云组之前、且与所述目标点云组相邻的点云组的数量阈值。
  7. 如权利要求1所述的方法，其特征在于，一个点云组对应一个点云组序，所述点云组序是指对应的点云组中所包括的点云点的坐标码字，在经过所述对应的点云组的分组移动位数移动后得到的；
    当所述目标点云组所对应的目标分组移动位数为坐标维度位数的倍数时，所述备选点集合所对应的备选点云组序在经过所述坐标维度位数的第一倍数进行移动后得到的备选移动序列，与所述目标点云组所对应的目标点云组序在经过所述坐标维度位数的第一倍数进行移动后得到的目标移动序列相同；所述坐标维度位数是指所述每个点云组所包括的点云点的坐标码字所对应的维度数量。
  8. 如权利要求7所述的方法，其特征在于，当所述目标点云组所对应的目标分组移动位数不为所述坐标维度位数的倍数时，所述备选点集合所对应的备选点云组序在经过补充维度位数进行移动后得到的备选移动序列，与所述目标点云组所对应的目标点云组序在经过所述补充维度位数进行移动后得到的目标移动序列相同；所述补充维度位数是指余数与所述坐标维度位数之间的位数差值，所述余数是指所述目标分组移动位数与所述坐标维度位数的余数；或者，所述补充维度位数是指所述位数差值与所述坐标维度位数的第二倍数之和。
  9. 如权利要求1所述的方法，其特征在于，所述备选点集合的数量为至少两个；所述从所述备选点集合中获取所述目标点云点关联的预测参考点，包括：
    在至少两个备选点集合中，分别选择d个点云点，将所述至少两个备选点集合分别对应的d个点云点，确定为所述目标点云点关联的预测参考点；d为正整数。
  10. 如权利要求1所述的方法，其特征在于，所述备选点集合的数量为至少两个；所述从所述备选点集合中获取所述目标点云点关联的预测参考点，包括：
    在至少两个备选点集合中，分别选择d个点云点，作为候选点云点，获取所述候选点云点与所述目标点云点之间的第一点间距离，基于所述第一点间距离对所述候选点云点进行排序，从排序后的候 选点云点中获取所述目标点云点关联的预测参考点。
  11. 如权利要求1所述的方法，其特征在于，所述备选点集合的数量为至少两个；所述从所述备选点集合中获取所述目标点云点关联的预测参考点，包括：
    获取至少两个备选点集合所包括的点云点与所述目标点云点之间的第二点间距离，基于所述第二点间距离对所述至少两个备选点集合所包括的点云点进行排序，从排序后的所述至少两个备选点集合所包括的点云点中，获取所述目标点云点关联的预测参考点。
  12. 如权利要求1所述的方法，其特征在于，所述备选点集合的数量为至少两个；所述从所述备选点集合中获取所述目标点云点关联的预测参考点，包括：
    获取至少两个备选点集合分别对应的集合优先级，基于所述集合优先级对所述至少两个备选集合进行排序，在排序后的至少两个备选集合中，获取所述目标点云点关联的P个预测参考点；P为正整数。
  13. 如权利要求1所述的方法，其特征在于，所述基于所述预测参考点对所述目标点云点进行预测处理，得到所述目标点云点的目标属性预测值，包括：
    获取所述预测参考点的预测参考坐标，获取所述目标点云点的目标坐标；
    基于所述预测参考坐标与所述目标坐标之间的坐标距离，确定所述预测参考点的参考权重；
    获取所述预测参考点的参考属性重建值，基于所述参考属性重建值与所述参考权重进行加权处理，得到所述目标点云点的目标属性预测值。
  14. 如权利要求1所述的方法，其特征在于，所述方法还包括：
    获取所述目标点云点的目标属性实际值，基于所述目标属性实际值与所述目标点云点的目标属性预测值之间的差值，得到所述目标点云点的目标属性残差；
    对所述目标属性残差进行量化转换，得到所述目标点云点的目标变换系数；所述目标变换系数用于对所述目标点云点进行编码。
  15. 如权利要求13所述的方法，其特征在于，所述方法还包括：
    获取所述目标点云组所包括的点云点的变换系数，对所述目标点云组所包括的点云点的变换系数进行编码，得到所述目标点云组所对应的目标码流。
  16. 如权利要求13所述的方法，其特征在于，所述方法还包括：
    获取组数限制阈值，基于所述组数限制阈值获取包含所述目标点云组的g个待编码点云组，获取所述g个待编码点云组所包括的点云点的变换系数，对所述g个待编码点云组所包括的点云点的变换系数进行编码，得到所述g个待编码点云组所对应的组码流；g为正整数，g小于或等于所述组数限制阈值。
  17. 一种点云预测处理方法，其特征在于，由计算机设备执行；所述方法包括：
    获取目标点云点所在的目标点云组的备选点集合；所述备选点集合属于M个点云组，所述M个点云组包括所述目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
    从所述备选点集合中获取所述目标点云点关联的预测参考点；
    基于所述预测参考点对所述目标点云点进行预测处理，得到所述目标点云点的目标属性预测值；
    获取所述目标点云点所对应的码流，对所述目标点云点所对应的码流进行解码处理，得到所述目标点云点的目标属性残差，基于所述目标属性预测值与所述目标属性残差，确定所述目标点云点的目标属性重建值。
  18. 如权利要求17所述的方法，其特征在于，所述对所述目标点云点所对应的码流进行解码处理，得到所述目标点云点的目标属性残差，包括：
    对所述目标点云点所对应的码流进行解码处理，得到所述目标点云点的目标变换系数；
    对所述目标变换系数进行反变换处理，得到所述目标点云点的目标属性残差。
  19. 一种点云预测处理装置，其特征在于，所述装置包括：
    备选集合获取模块，用于获取目标点云点所在的目标点云组的备选点集合；所述备选点集合属于 M个点云组，所述M个点云组包括所述目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
    参考点获取模块，用于从所述备选点集合中获取所述目标点云点关联的预测参考点；
    属性预测模块，用于基于所述预测参考点对所述目标点云点进行预测处理，得到所述目标点云点的目标属性预测值。
  20. 一种点云预测处理装置，其特征在于，所述装置包括：
    备选集合获取模块，用于获取目标点云点所在的目标点云组的备选点集合；所述备选点集合属于M个点云组，所述M个点云组包括所述目标点云组；每个点云组所包括的点云点的坐标码字，在经过所在点云组对应的分组移动位数移动后是相同的；M为正整数；
    参考点获取模块，用于从所述备选点集合中获取所述目标点云点关联的预测参考点；
    属性预测模块，用于基于所述预测参考点对所述目标点云点进行预测处理，得到所述目标点云点的目标属性预测值；
    码流获取模块，用于获取所述目标点云点所对应的码流；
    码流解码模块，用于对所述目标点云点所对应的码流进行解码处理，得到所述目标点云点的目标属性残差；
    属性重建模块，用于基于所述目标属性预测值与所述目标属性残差，确定所述目标点云点的目标属性重建值。
  21. 一种计算机设备，其特征在于，包括一个或多个处理器、存储器、输入输出接口；
    所述处理器分别与所述存储器和所述输入输出接口相连，其中，所述输入输出接口用于接收数据及输出数据，所述存储器用于存储计算机可读指令，所述处理器用于调用所述计算机可读指令，以使得所述计算机设备执行权利要求1-18任一项所述的方法。
  22. 一个或多个计算机可读存储介质，其特征在于，所述计算机可读存储介质存储有计算机可读指令，所述计算机可读指令适于由处理器加载并执行，以使得具有所述处理器的计算机设备执行权利要求1-18任一项所述的方法。
  23. 一种计算机程序产品，包括计算机可读指令，其特征在于，所述计算机可读指令被一个或多个处理器执行时实现权利要求1-18任一项所述的方法。
PCT/CN2022/135899 2022-03-11 2022-12-01 点云预测处理方法、装置、计算机、存储介质 WO2023169007A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/236,279 US20230394711A1 (en) 2022-03-11 2023-08-21 Point cloud prediction processing methods and apparatuses, computer, and storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210243498.9 2022-03-11
CN202210243498.9A CN116781908A (zh) 2022-03-11 2022-03-11 点云预测处理方法、装置、计算机、存储介质

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/236,279 Continuation US20230394711A1 (en) 2022-03-11 2023-08-21 Point cloud prediction processing methods and apparatuses, computer, and storage medium

Publications (1)

Publication Number Publication Date
WO2023169007A1 true WO2023169007A1 (zh) 2023-09-14

Family

ID=87937131

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/135899 WO2023169007A1 (zh) 2022-03-11 2022-12-01 点云预测处理方法、装置、计算机、存储介质

Country Status (3)

Country Link
US (1) US20230394711A1 (zh)
CN (1) CN116781908A (zh)
WO (1) WO2023169007A1 (zh)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110418135A (zh) * 2019-08-05 2019-11-05 北京大学深圳研究生院 一种基于邻居的权重优化的点云帧内预测方法及设备
CN112218079A (zh) * 2020-08-24 2021-01-12 北京大学深圳研究生院 一种基于空间顺序的点云分层方法、点云预测方法及设备
CN113596466A (zh) * 2020-04-30 2021-11-02 Oppo广东移动通信有限公司 一种最近邻搜索方法、编码器、解码器及存储介质

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110418135A (zh) * 2019-08-05 2019-11-05 北京大学深圳研究生院 一种基于邻居的权重优化的点云帧内预测方法及设备
CN113596466A (zh) * 2020-04-30 2021-11-02 Oppo广东移动通信有限公司 一种最近邻搜索方法、编码器、解码器及存储介质
CN112218079A (zh) * 2020-08-24 2021-01-12 北京大学深圳研究生院 一种基于空间顺序的点云分层方法、点云预测方法及设备

Also Published As

Publication number Publication date
US20230394711A1 (en) 2023-12-07
CN116781908A (zh) 2023-09-19

Similar Documents

Publication Publication Date Title
US9847913B2 (en) System and method for gathering information
CN104040542B (zh) 用于在易失性存储器内保持关系型数据的列向量的技术
CN104737165B (zh) 用于内存数据库查询处理的最优数据表示和辅助结构
Gurung et al. SQuad: Compact representation for triangle meshes
CN108628898B (zh) 数据入库的方法、装置和设备
WO2022121648A1 (zh) 点云数据编码方法、解码方法、设备、介质及程序产品
CN113452655A (zh) 一种分布式训练方法、梯度通信方法、装置以及计算设备
WO2022121650A1 (zh) 点云属性的预测方法、编码器、解码器及存储介质
WO2021218509A1 (zh) 一种最近邻搜索方法、编码器、解码器及存储介质
WO2023169007A1 (zh) 点云预测处理方法、装置、计算机、存储介质
JP2021192227A (ja) 画像処理方法と装置、電子機器、コンピュータ記憶媒体及びコンピュータプログラム
CN112332854A (zh) 霍夫曼编码的硬件实现方法、装置及存储介质
US20230086264A1 (en) Decoding method, encoding method, decoder, and encoder based on point cloud attribute prediction
TW202406344A (zh) 一種點雲幾何資料增強、編解碼方法、裝置、碼流、編解碼器、系統和儲存媒介
CN116760661A (zh) 数据存储方法、装置、计算机设备、存储介质和程序产品
JP6903773B2 (ja) 大きなデータをより小さな表現に変換する、およびより小さな表現を当初の大きなデータに戻して再変換するためのシステムおよび方法
Borysenko et al. The Fibonacci numeral system for computer vision
CN111459980A (zh) 监控数据的存储和查询方法及装置
US20210157485A1 (en) Pattern-based cache block compression
WO2023169019A1 (zh) 点云编解码方法、装置、计算机、存储介质
CN112598139A (zh) 类别编码方法、装置、设备、存储介质及程序产品
CN114640357B (zh) 数据编码方法、设备及存储介质
Netalkar et al. Distributed compression and decompression for big image data: LZW and Huffman coding
US20240087176A1 (en) Point cloud decoding method and apparatus, point cloud encoding method and apparatus, computer device, computer-readable storage medium
CN116708480B (zh) 一种基于Datax框架的数据同步方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22930630

Country of ref document: EP

Kind code of ref document: A1