JP2019165290A - Encoder, decoder, and program - Google Patents

Abstract

To enable cutting of noise in a region in which a solid does not originally exist even if solid-representing solid data is irreversibly compressed and decompressed (decompressed).SOLUTION: An encoder 100 generates encoded data 20 in which data has been compressed by processing solid-processing voxel data with a three-dimensional DCT unit 104, a quantizing unit 106, and an encoding unit 108. The encoder 100 generates structure data 22 representing the presence or absence of a voxel in each lattice position in voxel data 10 by a structure data generating unit 110. A decoder, which has received the encoded data 20 and the structure data 22, decodes the encoded data 20, and eliminates a voxel in a lattice position in which it is indicated that the structure data 22 has no voxel.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an encoding device, a decoding device, and a program.

  Three-dimensional modeling apparatuses such as 3D printers (three-dimensional printing machines) are becoming widespread. As a data format for a 3D printer, for example, a format that describes a three-dimensional shape by polygon mesh representation, such as an STL (Standard Triangulated Language) format or a 3DS format, is widely used.

  In addition, the applicants have proposed a data format called “FAV” that describes a three-dimensional model formed by a 3D printer in a voxel expression (Non-Patent Document 1). The FAV format can express various characteristics other than the three-dimensional shape by giving the voxel various attributes such as color, material, and link strength with other voxels.

  As can be inferred from the enormous amount of bitmap representation data for a two-dimensional full-color image, voxel format data (called voxel data) that represents a 3D object has various attributes such as color. When it is given, the amount of data becomes very large. There is a need for an encoding method that can reduce the amount of voxel data.

  In the method disclosed in Patent Document 1, the original texture image stored in the storage medium is spatially converted by the discrete cosine transform unit to become spatial frequency image data. The polygon calculation unit calculates the size of the polygon and calculates the resolution ratio of the texture to be pasted. The spatial low-pass filter low-pass filters the spatial frequency image data at a cutoff frequency corresponding to the resolution ratio of the texture. The filtered spatial frequency image data is subjected to inverse spatial transformation and then reduced to a mipmap image having a predetermined size.

  A technique for reducing the amount of data by performing frequency conversion such as DCT (Discrete Cosine Transform) on volume data obtained by CT (Computed Tomography) and cutting high frequency components of the conversion result is known. (Non-Patent Documents 2 and 3).

Japanese Patent Laid-Open No. 7-230555

Tomoya Takahashi, Masahiko Fujii, “Next-generation 3D printing data format“ FAV ”that realizes the world's highest level of expressiveness” ”, [online], Fuji Xerox Technical Report, No. 26, 2017, [2018 January 26 search], Internet <URL: https://www.fujixerox.co.jp/company/technical/tr/2017/pdf/s_07.pdf> Morishita Hiroyuki, Ohno Yoshio, IPSJ Research Reports Graphics and CAD (CG) 1995 (63 (1995-CG-075)), 9-16, 1995-07-14 Junichi Haruyama, Nobuyuki Umezu, Proceedings of the Winter Conference of the Institute of Image Information and Television Engineers (2001), 90, 2001-12-06 Tomoya Takahashi, Masahiko Fujii, “Next-generation 3D printing data format“ FAV ”that realizes the world's highest level of expressiveness” ”, [online], Fuji Xerox Technical Report, No. 26, 2017, [2018 January 26 search], Internet <URL: https://www.fujixerox.co.jp/company/technical/tr/2017/pdf/s_07.pdf>

  Modeling data representing a solid to be modeled by a 3D printer or the like is sparse. That is, there are many lattice positions where a solid does not exist among the lattice positions in the coordinate system that defines the shape of the solid. When irreversible compression is simply performed on such modeling data, when the compressed data is decoded, mosquito noise occurs near the boundary between the region where the solid exists and the region where the solid does not exist, and there is no solid. A value indicating a solid attribute may appear at the lattice position of the region. If modeling is performed using such decoded data, a part of the solid is formed at a position where the solid should not exist.

  An object of the present invention is to provide a mechanism that can cut noise in a region where a solid does not originally exist even when the solid data representing the solid is irreversibly compressed and decompressed (decoded).

  The invention according to claim 1 extracts structure data indicating the presence or absence of voxels at each lattice position from the three-dimensional data representing a solid as a collection of voxels each having attribute values arranged at lattice positions in space. An extracting means that performs irreversible compression on the attribute values at the respective lattice positions of the three-dimensional data, and generates compressed data, and outputs the structure data and the compressed data in association with each other Means for encoding.

  The invention according to claim 2 is a structure encoding means for encoding the structure data, wherein the structure code corresponding to the structure data of the target structure element composed of one or more adjacent lattice positions Further comprising: a structure encoding means for determining according to context information indicating the presence or absence of voxels at one or more predetermined lattice positions around the structure element, wherein the output means replaces the structure data with the structure encoding The encoding apparatus according to claim 1, wherein encoded structure data including the structure code determined by the means is output in association with the compressed data.

  The invention according to claim 3 has a structure code table for encoding the structure data for each value of the context information, and the structure encoding means has a structure corresponding to the structure data of the structural element of interest. The encoding device according to claim 2, wherein a code is determined using the structural code table corresponding to the value of the context information for the target structural element.

  According to a fourth aspect of the present invention, the structure encoding means uses a set of a plurality of adjacent lattice positions as the target structural element, and indicates a combination of the presence / absence of voxels at each lattice position included in the target structural element 4. The encoding device according to claim 2, wherein data is encoded using the context information indicating presence / absence of voxels at a plurality of predetermined lattice positions three-dimensionally around the target structural element. .

  The invention according to claim 5 is an adding means for adding a voxel having an attribute value obtained from an attribute value of one or more of the voxels in the vicinity of the lattice position to a lattice position where there is no voxel in contact with the voxel; And the compression means performs the lossy compression on the stereoscopic data to which the voxel is added by the addition means.

  The invention according to claim 6 is a structure that indicates whether or not there is a voxel at each lattice position from three-dimensional data that represents a computer by a collection of voxels each having an attribute value arranged at a lattice position in space. Extraction means for extracting data, irreversible compression on the attribute values at the respective grid positions of the stereoscopic data, compression means for generating compressed data, and outputting the structure data and the compressed data in association with each other This is a program for functioning as output means.

  The invention according to claim 7 corresponds to the irreversible compression for the compressed data, the means for inputting the structure data and the compressed data corresponding to each other output from the encoding device according to claim 1 For the lattice position indicated as having no voxel in the structure data, the attribute value of the lattice position in the decompressed data, Means for changing to a value indicating that there is no voxel.

  According to an eighth aspect of the present invention, there is provided a computer, the means for inputting the corresponding structure data and the compressed data output from the encoding device according to the first aspect, and the lossy compression for the compressed data. Means for generating decompression data for each lattice position by performing decompression processing corresponding to the above, for the lattice position indicated as having no voxels in the structure data, the attribute value of the lattice position in the decompression data is set. , A program for causing a function to change to a value indicating that there is no voxel.

  According to the first, sixth, seventh, or eighth aspect of the present invention, even in a case where stereoscopic data representing a solid is irreversibly compressed and decompressed (decoded), noise in a region where the solid does not originally exist can be cut.

  According to the second and third aspects of the present invention, data compression can be performed on structural data associated with compressed data.

  According to the invention which concerns on Claim 4, encoding efficiency can be improved rather than the system which encodes structure data for every lattice position.

  According to the fifth aspect of the present invention, it is possible to prevent or reduce the lossy compression noise caused by the surface of the solid indicated by the three-dimensional data from appearing inside the solid in the decoding result of the compressed data.

It is a figure which illustrates the structure of the encoding apparatus of embodiment. It is a figure for demonstrating the process of a three-dimensional DCT part. It is a figure which illustrates the internal structure of a three-dimensional DCT part. It is a figure which illustrates the structure of the decoding apparatus of embodiment. It is a figure which illustrates the structure of the encoding apparatus of a modification. It is a figure which shows the example of the lattice position group which comprises the structural element which consists of a several lattice position, and context information corresponding to it. It is a figure for demonstrating the structure of the encoding of a structural element. It is a figure which illustrates the structure of the encoding apparatus of the further modification.

  An embodiment of an encoding apparatus according to the present invention will be described with reference to FIG.

  An encoding apparatus 100 illustrated in FIG. 1 includes a block cutout unit 102, a three-dimensional DCT unit 104, a quantization unit 106, an encoding unit 108, and a structure data generation unit 110.

  Voxel data 10 is input to this device. The voxel data 10 is data that defines, for example, a solid modeled by a three-dimensional modeling apparatus by a collection of voxels. A voxel is an element that is a unit of a solid, and a three-dimensional space that encloses the shape of the solid is an eyelet-like shape that can be divided into a lattice pattern by straight lines with equal intervals parallel to the coordinate axes of x, y, and z. Housed in individual small spaces. This small space is hereinafter referred to as a lattice position or a cell. The small space has a cubic shape, and in a typical example, the voxel is a cube-shaped element that occupies the entire small space, but the shape of the voxel is not limited to this. Each voxel has one or more attributes such as color, material, and strength of relationship with adjacent voxels (for example, representing the strength of the joint). Among the lattice positions in the space, there are a lattice position where there are voxels constituting a solid and a lattice position where there is no such voxel. The voxel data 10 includes, for each lattice position, information indicating the presence / absence of a voxel at the lattice position, and information indicating one or more attribute values of the voxel for the lattice position where the voxel exists. . The FAV format shown in Non-Patent Document 1 is an example of a data format that can express such voxel data 10.

  The block cutout unit 102 cuts out a block from the voxel data 10. In the example shown in (0) of FIG. 2, the block 300 is a block made up of a predetermined number (e.g., 8) voxels in each of the vertical, horizontal, and depth directions.

  The three-dimensional DCT unit 104 performs a three-dimensional discrete cosine transform (DCT) on the block cut out by the block cutout unit 102.

  In one example, the three-dimensional DCT unit 104 performs the process shown in FIG. That is, first, the three-dimensional DCT unit 104 executes the one-dimensional DCT 1041 along a predetermined (ie, predetermined) direction, for example, the x direction, with respect to the block 300 as shown in (1) of FIG. That is, in this case, DCT is applied to the attribute values (for example, color values) of voxels arranged in the X direction in each column in the same y coordinate in a layer where the z coordinate is constant. Do. Such processing is executed for each layer having a different z coordinate. Note that there are lattice positions without voxels in the column. For such lattice positions, for example, a predetermined attribute value (for example, 0), an attribute value of an adjacent voxel, or a surrounding voxel group (for example, the relevant voxel) One-dimensional DCT is performed after padding (padding) with the average value of attribute values of the voxel group belonging to the block including the lattice position. The processing result of the one-dimensional DCT for a certain column is overwritten on the value of each voxel in that column. Next, the three-dimensional DCT unit 104 performs a 90-degree rotation 1043 process around a certain coordinate axis (for example, the z-axis) on the block that is the processing result of the one-dimensional DCT 1041. Next, as shown in (2) of FIG. 2, the three-dimensional DCT unit 104 executes the one-dimensional DCT 1045 on the rotated block in the same direction as before. Since the block has been rotated 90 degrees, the one-dimensional DCT 1045 is a DCT in the coordinate axis direction different from the one-dimensional DCT 1041 described above. Next, the three-dimensional DCT unit 104 performs 90-degree rotation 1047 with a coordinate axis different from the previous 90-degree rotation 1043 as the rotation axis, on the processing result block of the one-dimensional DCT 1045. Then, as shown in (3) of FIG. 2, the three-dimensional DCT unit 104 executes the one-dimensional DCT 1049 on the rotated block in the same direction as before. The one-dimensional DCT 1049 is a DCT in the direction of coordinate axes different from the one-dimensional DCT 1041 and the one-dimensional DCT 1045. Through the above processing, three-dimensional DCT is performed on the block.

  Returning to the description of FIG. The quantization unit 106 performs a known quantization process on the data of the 3D DCT result of each block output from the 3D DCT unit 104. As a result, the amount of data is reduced, but the processing of the encoding device 100 is an irreversible compression process involving quantization, so even if the encoded data of the compression result is decoded, the data is exactly the same as the original voxel data. Dont return. In the case of voxel data, as in the case of JPEG compression of an image, when encoded data obtained by such compression is decoded, noise such as mosquito noise occurs.

  The encoding unit 108 generates encoded data (compressed data) 20 for the voxel data 10 by performing a known encoding process such as entropy encoding on the data resulting from the quantization performed by the quantization unit 106. To do.

  The structure data generation unit 110 generates structure data 22 indicating the presence or absence of voxels at each lattice position (cell) of the block cut out by the block cutout unit 102. The structure data is data in which values indicating the presence / absence of voxels at each grid position (which can be expressed by 1 bit per grid position) are arranged according to a predetermined arrangement order of the grid positions. When the voxel data 10 is in the FAV format, the voxel data 10 includes data named “voxel_map” indicating the presence / absence of the voxel at each lattice position. “Voxel_map” is extracted, and the structure data 22 may be generated by extracting information of each cell in the block from this “voxel_map”.

  The encoded data 20 generated by the encoding unit 108 and the structural data 22 generated by the structural data generation unit 110 are output in a state of being associated with each other.

  Next, an example of a decoding device 200 that decodes the encoded data 20 will be described with reference to FIG.

  The decoding apparatus 200 includes a decoding unit 202, an inverse quantization unit 204, a three-dimensional inverse DCT unit 206, and a mask processing unit 208. The decoding apparatus 200 receives encoded data 20 and structure data 22 corresponding to the encoded data 20. The decoding device 200 decodes (decompresses) the compressed encoded data 20.

  The decoding unit 202 decodes the input encoded data 20 by a decoding process corresponding to the encoding process of the encoding unit 108. The inverse quantization unit 204 inversely quantizes the decoding result of the decoding unit 202. The three-dimensional inverse DCT unit 206 performs inverse DCT processing on the data of the inverse quantization result. As a result, the frequency space information returns to the real space attribute value.

  Since the voxel data restored in this way includes noise due to quantization, an attribute value may appear at a lattice position where there is no voxel (and therefore no attribute value). In the case of voxel data for modeling purposes, if an attribute value appears at a lattice position that originally does not have a voxel, the lattice position is formed, and a shape different from the original shape is formed. In order to prevent such a situation, the decoding apparatus 200 includes a mask processing unit 208.

  The mask processing unit 208 refers to the structure data 22 corresponding to the encoded data 20 and in the original voxel data 10 among the attribute values of the decoding result of each lattice position (that is, the output of the three-dimensional inverse DCT unit 206). The value of the lattice position where there is no voxel is masked (that is, changed to a value indicating that there is no voxel). Since the voxel data output from the mask processing unit 208 has been subjected to lossy compression, the attribute value of each voxel may not be exactly the same as the original voxel data 10, but the three-dimensional shape is the original voxel data. It will be the same as 10.

  Next, a modified example of the present embodiment will be described with reference to FIGS.

  In the above embodiment, the structure data 22 is raw data. On the other hand, the encoding apparatus 100 of this modification includes a structure encoding unit 112 that compresses and encodes the structure data 22 generated by the structure data generation unit 110. The structure encoding unit 112 performs context-based compression encoding on the structure data 22. In FIG. 5, the structure data generation unit 110 generates the structure data from the blocks generated by the block cutout unit 102, but instead, the structure data is generated from the voxel data 10 input to the block cutout unit 102. May be generated. This also applies to the apparatus configuration shown in FIG.

  For this compression encoding, the structure encoding unit 112 generates context information from the voxel data 10 and compresses the structure data based on the context information. The context information includes the voxel at one or more predetermined grid positions that are three-dimensionally surrounding the grid position of interest (the voxel may not be present at that grid position, but is referred to as “target voxel” for convenience). It is a pair of presence or absence.

  For example, in the example illustrated in FIG. 6, four voxels adjacent to each other in a 2 × 2 matrix in the target layer of the voxel data 10 are set as the target voxel, and four reference lattices adjacent to the target voxel group in the target layer The context information is a combination of the presence / absence of voxels in eight reference voxels including a position (referred to as “reference voxel” for convenience) and four reference voxels immediately below the four target voxels. Here, in the same layer (that is, a layer consisting of a group of lattice positions having the same z coordinate), encoding (and decoding) is advanced by zigzag scanning in which the x direction is the main scanning direction in the layer and the y direction is the sub scanning direction. At the same time, when the encoding (decoding) of one layer is finished, the process proceeds to the encoding (and decoding) of the next layer in the z direction. The layer that is currently being encoded (or decoded) is the target layer, and the set of lattice positions that are the target of the current encoding (or decoding) within the target layer (that is, the illustrated 4) A set of two voxels of interest). In one example, the reference voxel in the target layer is closer to the target structural element in the main scanning direction and the sub-scanning direction (that is, the value of the reference voxel is known at the time of encoding or decoding). Adjacent grid positions. Further, the layer immediately below is a layer immediately before the target layer along the direction of progress of encoding when the z direction in which encoding or decoding proceeds is viewed as upward.

  In this example, the structural data is encoded in units of structural elements, with a set of four adjacent voxels of interest as one structural element as illustrated. That is, in the coding of the structure data in the attention layer, the code is determined for every 2 × 2 4 voxels. For example, the order shown in the figure is defined for the four target voxels in the structural element. The structure data for a certain structural element is a value obtained by arranging values indicating the presence or absence of voxels at the lattice position of each target voxel included in the structural element (for example, expressed by a 1-bit value representing 0 or 1) according to the order. It is. Similarly, for reference voxels, for example, the order shown in the figure is defined. The context information is a value in which values indicating the presence or absence of voxels at the lattice positions of the respective reference voxels are arranged according to the order.

  An example of the structure data encoding process performed by the structure encoding unit 112 will be described with reference to FIG. In this example, the attention voxel and the reference voxel illustrated in FIG. 6 are used.

  In this example, the structure encoding unit 112 has a code table corresponding to each value of context information. In the example of FIG. 6, since the context information is an 8-bit value indicating the presence or absence of voxels at the positions of eight reference voxels (lattice positions), the code table is prepared to the power of 2 (that is, 256) (shown in the figure). Code table # 0 to # 255). In each code table, for each value of the structure data of the target structure element to be encoded (indicated as “target structure data” in the figure), a code word corresponding to the value is registered. In the example of FIG. 6, the target structural element is four target voxels, and the target structure data is a 4-bit value indicating whether or not there is a voxel at the position of the four target voxels. The code table for each context information is obtained by scanning the voxel data 10 to be encoded once, obtaining the appearance frequency of the combination of the structural data of the structural element in the voxel data 10 and the context information, and information on the appearance frequency. May be generated as being optimized for the voxel data. Also, a set of code tables for each context information obtained from a large number of past voxel data examples may be registered in the structure encoding unit 112 and used for encoding.

  The structure encoding unit 112 reads the value of the target structure element (target structure data) from the structure data generated by the structure data generation unit 110, and generates context information corresponding to the target structure element. Then, the structure encoding unit 112 outputs a code word corresponding to the target structure data in the code table corresponding to the context information as a structure code that is an encoding result of the target structure data. By this processing, the structure data is converted into encoded structure data 22a.

  The decoding apparatus 200 corresponding to the encoding apparatus 100 of this modification may decode the structure data from the encoded structure data 22a and supply the structure data to the mask processing unit 208. In this decoding, context information is obtained from decoding results (presence / absence of voxels and color values) of lattice positions that have already been obtained, and information indicated by a decoding target code is read out from a code table corresponding to the context information. That is, when decoding a certain structural code, the decoding apparatus 200 identifies a code table (see FIG. 6) corresponding to the context information for the structural element corresponding to the structural code, out of the structural code code table, The value of the target structure data corresponding to the structure code is read from the code table, and this is used as the decoding result. In this modification, since the structural information is encoded in units of structural elements including a plurality of voxels, the encoding efficiency is improved as compared with the case of encoding in units of voxels.

  Next, another modification will be described with reference to FIG.

  The encoding apparatus 100 according to this modification includes a voxel adding unit 101. The voxel adding unit 101 has attribute values of one or more voxels in the vicinity of the lattice position X with respect to a lattice position adjacent to the voxel among the lattice positions where no voxel of the voxel data 10 is present (for convenience, referred to as “lattice position X”). Add voxels with attribute values determined according to.

  Here, the “neighborhood” of the grid position X is, for example, a grid position adjacent to the grid position (cell) X (for example, 6 neighborhoods that touch the cell in a plane, 18 neighborhoods that include a border on the side, or further points. It is also possible to set the value in the vicinity of 26 including the one touching by the above. Further, the vicinity of the lattice position adjacent to the lattice position X (that is, the lattice position at a distance of two cells from the lattice position X) may be set as “neighboring”. In general, a group of lattice positions (cells) within a predetermined number of cells from the lattice position X may be set as “neighborhood”. The lattice position X where there is no voxel and the group of voxels located in the vicinity thereof may be obtained from the structure data generated from the voxel data 10 by the structure data generation unit 110 (indicating whether there is a voxel at each lattice position).

  The voxel adding unit 101 obtains the attribute value of the voxel to be added to the lattice position X by a predetermined calculation method from the attribute value of the voxel located near the lattice position X where there is no voxel. This calculation method may be, for example, a method of obtaining an average of attribute values of neighboring voxels, or a method of performing weighted averaging of attribute values of neighboring voxels according to the distance from the lattice position X. Also good. Of course, other calculation methods may be used.

  The voxel adding unit 101 adds a voxel to each of the lattice positions adjacent to the voxel among the lattice positions where there is no voxel, so that the surface of the solid represented by the original voxel data 10 (that is, the space without the solid and the solid) ) Is covered with one layer of additional voxel group. The voxel addition unit 101 may repeat such a voxel addition process N times so as to cover the surface of the solid represented by the original voxel data 10 with an additional voxel layer having a thickness corresponding to N voxels. The process performed by the voxel adding unit 101 may be regarded as a process of expanding the solid represented by the original voxel data 10. By the processing of the voxel adding unit 101,

  The block cutout unit 102, the three-dimensional DCT unit 104, the quantization unit 106, and the encoding unit 108 perform the same processing as described above on the voxel data to which voxel addition is performed by the voxel addition unit 101.

  As described above, when lossy compression including processing such as quantization is performed, strong noise such as mosquito noise that does not exist in the original data appears in the decoded data. Such noise appears particularly in the vicinity of an edge whose attribute value changes abruptly. However, particularly in the case of voxel data 10 representing a solid to be modeled, strong noise appears in the vicinity of the boundary between the solid and the outside. When such a noise attribute value appears at a lattice position outside the solid, the lattice position is recognized as a voxel constituting the solid, but this is determined by the processing of the mask processing unit 208 using the structure data. Deleted. However, noise also appears inside the solid near the boundary. Such strong noise can cause the attribute value of the voxel inside the solid to change greatly from its original value, but such a change cannot be handled by the mask processing unit 208.

  On the other hand, since the encoding apparatus 100 of the modified example of FIG. 8 covers the surface of the solid indicated by the original voxel data 10 with an additional voxel having an attribute value based on the attribute value of the voxel near the surface. The boundary between the outside and the outside moves outside the original solid. For this reason, the voxel addition unit 101 adds a voxel layer having a sufficient thickness to the three-dimensional surface of the voxel data 10, so that when the lossless compression is applied to the data after the addition of the voxel, the vicinity of the three-dimensional surface indicated by the decoded data Is prevented or reduced from appearing inside the original solid.

  Since the added voxel is deleted by the mask processing unit 208, it does not affect the modeling.

  The encoding apparatus 100 and the decoding apparatus 200 exemplified above can be configured as hardware logic circuits in one example. As another example, the encoding device 100 and the decoding device 200 may be realized, for example, by causing a built-in computer to execute a program that represents the function of each functional module in each device. Here, the computer includes, for example, a processor such as a CPU, a memory (primary storage) such as a random access memory (RAM) and a read only memory (ROM), an HDD controller that controls an HDD (hard disk drive), and the like as hardware. A network interface that performs control for connection to a network such as an I / O (input / output) interface or a local area network has a circuit configuration connected via a bus, for example. Also, portable non-volatile recording of various standards such as a disk drive and a flash memory for reading and / or writing to a portable disk recording medium such as a CD or a DVD via the I / O interface, for example. A memory reader / writer for reading from and / or writing to a medium may be connected. A program in which the processing contents of each functional module exemplified above are described is stored in a fixed storage device such as a hard disk drive via a recording medium such as a CD or DVD, or via a communication means such as a network, and stored in a computer. Installed. The program stored in the fixed storage device is read into the RAM and executed by a processor such as a CPU, whereby the functional module group exemplified above is realized. The encoding device 100 and the decoding device 200 may be configured by a combination of software and hardware.

DESCRIPTION OF SYMBOLS 100 Coding apparatus, 101 Voxel addition part, 102 Block cutout part, 104 Three-dimensional DCT part, 106 Quantization part, 108 Coding part, 110 Structure data generation part, 112 Structure coding part, 200 Decoding apparatus, 202 Decoding part 204 Inverse quantization unit 206 Three-dimensional inverse DCT unit 208 Mask processing unit

Claims (8)

Extraction means for extracting structure data indicating the presence or absence of voxels at each lattice position from the three-dimensional data represented by a collection of voxels each having an attribute value arranged at lattice positions in space;
Compression means for performing irreversible compression on the attribute values at the respective grid positions of the three-dimensional data and generating compressed data;
Output means for associating and outputting the structure data and the compressed data;
An encoding device including: Structure encoding means for encoding the structure data, wherein a structure code corresponding to the structure data of the target structural element consisting of one or more adjacent lattice positions is set to a predetermined one around the target structural element. Further comprising a structure encoding means for determining according to context information indicating the presence or absence of voxels at the above lattice positions,
The encoding apparatus according to claim 1, wherein the output unit outputs encoded structure data including a structure code determined by the structure encoding unit in association with the compressed data, instead of the structure data. A structure code table for encoding the structure data for each value of the context information;
The structure encoding unit determines a structure code corresponding to the structure data of the target structure element using the structure code table corresponding to the value of the context information for the target structure element. The encoding device described.   The structure encoding means uses a set of a plurality of lattice positions adjacent to each other as the target structural element, and receives structure data indicating a combination of the presence or absence of voxels at each lattice position included in the target structural element. The encoding device according to claim 2 or 3, wherein encoding is performed using the context information indicating the presence or absence of voxels at a plurality of predetermined lattice positions that are three-dimensionally surrounding. An additional means for adding a voxel having an attribute value obtained from an attribute value of one or more voxels in the vicinity of the lattice position to a lattice position where the voxel is not in contact with the voxel;
The encoding device, wherein the compression unit performs the lossy compression on the stereoscopic data to which the voxel is added by the addition unit. Computer
Extraction means for extracting structure data indicating the presence or absence of voxels at each lattice position from the three-dimensional data represented by a collection of voxels each having an attribute value arranged at a lattice position in space,
Compression means for performing irreversible compression on the attribute values at the respective grid positions of the three-dimensional data and generating compressed data;
Output means for outputting the structure data and the compressed data in association with each other;
Program to function as. Means for inputting the structure data and the compressed data corresponding to each other output by the encoding device according to claim 1;
Means for generating decompressed data for each lattice position by performing decompression processing corresponding to the lossy compression on the compressed data;
For the lattice position indicated as having no voxel in the structure data, means for changing the attribute value of the lattice position in the decompressed data to a value indicating that there is no voxel;
A decoding device. Computer
Means for inputting the structure data and the compressed data corresponding to each other output by the encoding device according to claim 1;
Means for generating decompressed data for each lattice position by performing decompression processing corresponding to the lossy compression on the compressed data;
For the lattice position indicated as having no voxel in the structure data, means for changing the attribute value of the lattice position in the decompressed data to a value indicating that there is no voxel;
Program to function as.