ISO/IEC 14496-16:2011/Amd 4:2017
(Amendment)Information technology — Coding of audio-visual objects — Part 16: Animation Framework eXtension (AFX) — Amendment 4: Pattern-based 3D mesh coding (PB3DMC)
Information technology — Coding of audio-visual objects — Part 16: Animation Framework eXtension (AFX) — Amendment 4: Pattern-based 3D mesh coding (PB3DMC)
Technologies de l'information — Codage des objets audiovisuels — Partie 16: Extension du cadre d'animation (AFX) — Amendement 4: Codage de maille 3D fondé sur un modèle
General Information
Relations
Standards Content (Sample)
INTERNATIONAL ISO/IEC
STANDARD 14496-16
Fourth edition
2011-11-01
AMENDMENT 4
2017-10
Information technology — Coding of
audio-visual objects —
Part 16:
Animation Framework eXtension (AFX)
AMENDMENT 4: Pattern-based 3D mesh
coding (PB3DMC)
Technologies de l'information — Codage des objets audiovisuels —
Partie 16: Extension du cadre d'animation (AFX)
AMENDEMENT 4: Codage de maille 3D fondé sur un modèle
Reference number
ISO/IEC 14496-16:2011/Amd.4:2017(E)
©
ISO/IEC 2017
---------------------- Page: 1 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
COPYRIGHT PROTECTED DOCUMENT
© ISO/IEC 2017, Published in Switzerland
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized otherwise in any form
or by any means, electronic or mechanical, including photocopying, or posting on the internet or an intranet, without prior
written permission. Permission can be requested from either ISO at the address below or ISO’s member body in the country of
the requester.
ISO copyright office
Ch. de Blandonnet 8 • CP 401
CH-1214 Vernier, Geneva, Switzerland
Tel. +41 22 749 01 11
Fax +41 22 749 09 47
copyright@iso.org
www.iso.org
ii © ISO/IEC 2017 – All rights reserved
---------------------- Page: 2 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
Foreword
ISO (the International Organization for Standardization) and IEC (the International Electrotechnical
Commission) form the specialized system for worldwide standardization. National bodies that are
members of ISO or IEC participate in the development of International Standards through technical
committees established by the respective organization to deal with particular fields of technical
activity. ISO and IEC technical committees collaborate in fields of mutual interest. Other international
organizations, governmental and non-governmental, in liaison with ISO and IEC, also take part in the
work. In the field of information technology, ISO and IEC have established a joint technical committee,
ISO/IEC JTC 1.
The procedures used to develop this document and those intended for its further maintenance are
described in the ISO/IEC Directives, Part 1. In particular, the different approval criteria needed for
the different types of document should be noted. This document was drafted in accordance with the
editorial rules of the ISO/IEC Directives, Part 2 (see www.iso.org/directives).
Attention is drawn to the possibility that some of the elements of this document may be the subject
of patent rights. ISO and IEC shall not be held responsible for identifying any or all such patent
rights. Details of any patent rights identified during the development of the document will be in the
Introduction and/or on the ISO list of patent declarations received (see www.iso.org/patents).
Any trade name used in this document is information given for the convenience of users and does not
constitute an endorsement.
For an explanation on the voluntary nature of standards, the meaning of ISO specific terms and
expressions related to conformity assessment, as well as information about ISO's adherence to the
World Trade Organization (WTO) principles in the Technical Barriers to Trade (TBT) see the following
URL: www.iso.org/iso/foreword.html.
This document was prepared by Joint Technical Committee ISO/IEC JTC 1, Information technology,
Subcommittee SC 29, Coding of audio, picture, multimedia and hypermedia information.
A list of all parts in the ISO/IEC 14496 series can be found on the ISO website.
© ISO/IEC 2017 – All rights reserved iii
---------------------- Page: 3 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
Information technology — Coding of audio-visual
objects —
Part 16:
Animation Framework eXtension (AFX)
AMENDMENT 4: Pattern-based 3D mesh coding (PB3DMC)
5.2 Geometry tools
Add the following subclauses:
5.2.7 Pattern-based 3D mesh coding (PB3DMC)
5.2.7.1 Overview
In practical applications, many 3D models consist of a large number of connected components. And these
multi-connected 3D models usually contain lots of repetitive structures in various transformations, as
shown in Figure Amd4.1. In order to increase their efficiency, compression methods for this kind of 3D
models should be able to extract the redundancy existing in the repetitive structures.
Figure Amd4.1 — 3D models with a large number of connected components and repetitive
structures
This document presents an efficient compression algorithm for multi-connected 3D models by taking
advantage of discovering repetitive structures in the input models. It allows discovering of structures
repeating in various positions, orientations and scaling factors, where the 3D model is then organized
into “pattern-instance” representation. A pattern is the representative geometry of the corresponding
repetitive structure. The connected components belonging to a repetitive structure are called
instances of the corresponding pattern and represented by the pattern ID and their transformation, i.e.
the combination of reflection, translation, rotation and possible uniform scaling, with regards to the
pattern. The instance transformation consists of four parts: reflection part, translation part, rotation
part and possible scaling part.
The repetitive structure discovery-based compression algorithm proposed in this subclause, brings
significant compression gain compared to the static 3D model compression algorithms provided by
SC3DMC when 3D models present repetitive features. This document defines the compressed bitstream
syntax and semantics for this repetitive structure discovery-based compression approach.
© ISO/IEC 2017 – All rights reserved 1
---------------------- Page: 4 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
5.2.7.2 General compression approach
5.2.7.2.1 Identification of repetitive and symmetric structures in a 3D model
This subclause defines several terms that are used all across this document and presents the different
steps that are used in order to identify repetitive structures.
There are two types of repetitive structures in 3D models:
— Unconnected repetitive structure: One unconnected repetitive structure consists of all the
connected components which are invariant in various positions, orientations and scaling factors.
One unconnected repetitive structure includes one pattern, which corresponds to one connected
component in this structure, and all the other connected components are its instances.
— Connected repetitive structure: One connected repetitive structure consists of all the surface
patches which are invariant in various positions, orientations and scaling factors. In other words, it
is a repetitive structure that may be found within one connected component. Connected repetitive
structure is sometimes also called symmetric structure. This document uses both naming.
Similar to unconnected repetitive structure, one symmetric structure includes one pattern, which
corresponds to one surface patch in this structure, and all the other surface patches are instances
of the pattern.
PB3DMC discovery of the two types of repetitive structures is done in four steps as described below,
considering that the input 3D model is as shown in Figure Amd4.2.
Figure Amd4.2 — Input 3D model
STEP 1: Identification of unconnected repetitive structures and unique part.
The input 3D model is divided into two parts, the unconnected repetitive structures and the unique part
which includes all those components that are not included in any unconnected repetitive structures. In
Figure Amd4.2, clouds, leaves and houses on the right are all unconnected repetitive structures.
STEP 2: Choose patterns and instances within unconnected repetitive structures.
Following STEP 1, for each unconnected repetitive structures, one pattern shall be chosen among all
repetitions. Other repetitions then become instances of their related pattern. The input 3D model is
then divided into three parts as follows:
— A: Patterns of all unconnected repetitive structures (see Figure Amd4.3);
— B: Instances of all unconnected repetitive structures (see Figure Amd4.4);
— C: Unique part which is not included in any unconnected repetitive structures (see Figure Amd4.5).
2 © ISO/IEC 2017 – All rights reserved
---------------------- Page: 5 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
Figure Amd4.3 — A: Patterns of all unconnected repetitive structures
Figure Amd4.4 — B: Instances of all unconnected repetitive structures
Figure Amd4.5 — C: Unique part which is not included in any unconnected repetitive structures
© ISO/IEC 2017 – All rights reserved 3
---------------------- Page: 6 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
As shown in Figure Amd4.6, part A can be further divided into two types:
— D: Patterns of unconnected repetitive structures which do not include any symmetric structures;
— E: Patterns of unconnected repetitive structures which include symmetric structures.
Figure Amd4.6 — Part A further divided into two types
4 © ISO/IEC 2017 – All rights reserved
---------------------- Page: 7 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
As shown in Figure Amd4.7, part C can be further divided into two types:
— F: Unique part which is not included in any unconnected repetitive structures and which does not
include unconnected repetitive structures and symmetric structures;
— G: Unique part which is not included in any unconnected repetitive structures and which includes
symmetric structures.
Figure Amd4.7 — Part C can be further divided into two types
STEP 3: Identification of symmetric structures.
Since instances will be coded with reference to their related patterns, symmetric structures
are identified either within pattern or unique parts in the 3D model. Consequently, as shown in
Figure Amd4.8, the input in STEP 3 is the input model, except instances of all unconnected repetitive
structures discovered in STEP 2.
NOTE The input in STEP 3 is the input model, except instances of unconnected repetitive structure.
Figure Amd4.8 — Input of STEP 3
© ISO/IEC 2017 – All rights reserved 5
---------------------- Page: 8 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
STEP 4: Choose patterns and instances within symmetric structures.
Similar to STEP 2 on repetitive structures, STEP 4 consists choosing patterns and instances within
symmetric structures. As shown in Figure Amd4.9, the following four types of data are defined:
— H: Patterns of all symmetric structures;
— I: Instances of all symmetric structures;
— J: Unique parts of those unique components which do not belong to any unconnected repetitive
structures but have symmetric structures;
— K: Unique parts on those unconnected-repetitive-structure patterns including symmetric
structures.
Figure Amd4.9 — Result of STEP 4
5.2.7.2.2 Two-instance reconstruction modes
While the bitstream needs to embed all the instance data, it should also be efficient and should address
several applications where sometimes either bitstream size or decoding efficiency or error resilience
matters the most.
Therefore, two options are proposed in reconstructing the data of one instance, i.e. its pattern ID (ID
being the actual position of the patterns in the compressed bitstream of patterns: 1 for first pattern, 2
for second pattern, etc.), its reflection transformation part (F), its translation transformation part (T),
its rotation transformation part (R) and its scaling transformation part (S), from the bitstream. Both of
them have their own pros and cons.
Option (A) elementary instance data mode (ID, F, T, R, S, ID, F, T, R, S…): Using this mode, the pattern
ID, reflection transformation part, translation transformation part, rotation transformation part and
scaling transformation part of one instance are packed together in the bitstream.
6 © ISO/IEC 2017 – All rights reserved
---------------------- Page: 9 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
Pros:
— It is error resilient. The decoder can recover from losing the transformation of some instances.
— Online decoding, which means that the instances can be decoded one by one during actual reading
of the compressed bitstream. There is no need to wait for the reading of the whole compressed
bitstream to be finished.
— Higher codec speed.
— The codec needs no buffer.
Cons:
— Relatively larger compressed 3D model size.
Option (B) grouped instance data mode (ID, ID, F, F, T, T, R, R, S, S): Using this mode, the pattern
ID, reflection transformation part, translation transformation part, rotation transformation part and
scaling transformation part of one instance are packed together in the bitstream.
Pros:
— Relatively smaller compressed 3D model size.
Cons:
— The decoder is no longer error resilient.
— Off-line decoding, which means the decoder can only start decoding after reading the whole
compressed bitstream.
— Lower codec speed.
— Buffer is necessary.
The bitstream definition includes both of the above two options. The encoder can choose the one which
fits its application better.
Since instances may have larger reconstruction error than its related patterns (error being defined
as the distance between the original component and the component restored from the pattern and
instance transformation), some data fields of the bitstream are defined to denote the compressed
instance reconstruction error to guarantee the decoded 3D model quality. Whether or not to record the
decoding error of an instance is based on the quality requirement.
5.2.7.2.3 Reconstruction of instance transformation
The instance transformation is reconstructed from four parts: reflection part, rotation part, translation
part and possible scaling part.
— The reflection part is represented by a 1-bit flag.
— The rotation part is reconstructed from the three Euler angles (alpha, beta, gamma).
— The translation part is represented by a vector (x, y, z) (translation vector).
— The scaling part is represented by the uniform scaling factor S of the instance.
© ISO/IEC 2017 – All rights reserved 7
---------------------- Page: 10 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
NOTE The number under each sub-figure is the corresponding occupancy codes.
a) 2D example of space subdivision for quad-tree construction
NOTE 1 The red line illustrates the breadth first traversal of the binary tree.
NOTE 2 The occupancy codes of the tree nodes, sorted according to the breadth first traversal, construct the
occupancy code sequence that describes the tree structure.
b) Example of breadth first traversal of the binary tree
NOTE For a predefined traversal order “top left–top right–bottom right–bottom left”, the input bitstream is
decoded.
c) 2D example of quad-tree reconstruction process
Figure Amd4.10 — Reconstruction of instance transformation
8 © ISO/IEC 2017 – All rights reserved
---------------------- Page: 11 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
While using grouped instance transformation mode, all instance translation vectors are compressed by
octree (OT) decomposition-based compression algorithm, which recursively subdivides the bounding
box of all instance translation vectors in an octree data structure, as illustrated by the 2D example
in Figure Amd4.10 a). Each octree node subdivision is represented by the 8-bit long occupancy code,
which uses a 1-bit flag to signify whether a child node is nonempty. An occupancy code sequence
describing the octree is generated by breadth first traversing the octree, as illustrated by the 2D
example in Figure Amd4.10 b). To decode these instance translation vectors, the octree is reconstructed
by breadth first traversing the octree from top to bottom, as shown in Figure Amd4.10 c). First, the top
layer is obtained by decoding the first long occupancy code. If any nodes at Layer i have more than one
“1” in its occupancy code, e.g. “01100000”, the codec decodes necessary number of symbols to append
as the children of such nodes. This process continues until all the leaf nodes have only one “1” or are
the terminal code. The occupancy code sequence is decoded by dividing it into several intervals and
decoded them with different probability models. Since instances may have extremely close translation
vectors, which are called as duplicate translation vectors, some data fields of the bitstream are defined
to denote these duplicate translation vectors.
5.2.7.2.4 Reconstruction of instance attributes
In practical applications, besides geometry, 3D models usually have various attributes such as normal,
colour and texture coordinates. Requiring instances to have the same attributes of patterns will limit the
number of repetitive structures that can be discovered and decrease the compression ratio of PB3DMC.
Thus, only the geometry is checked during repetitive structure discovery and the instances may have
attributes different from the corresponding pattern’s attributes. There are two reconstruction modes
for instance attributes.
— Share attribute mode: The instance shares the pattern attribute data and does not need data fields
to represent its attributes.
— Specific attribute mode: The instance has its own attributes and need separate data fields to
represent its attributes in the bitstream.
When the elementary instance data mode is used, one data field is defined to denote how to reconstruct
the attributes of an instance from the bitstream. The attribute data of one instance (A) follows the
other data of the instance, i.e. (ID, F, T, R, S, A, ID, F, T, R, S, A…). When the grouped instance data mode is
used, all instances should either share the pattern attribute data or have their own attribute data. The
instance data part of the bitstream is like (ID, ID, F, F, T, T, R, R, S, S, A, A).
5.2.7.2.5 Reconstruction of textured image(s)
It is expected that the repetitive structures in 3D models share textured image(s)/portion(s), as
well as geometry, as shown in Figure Amd4.11. Given the geometric matching relationship between
patterns and instances, the texture redundancy can be removed. Thus, the compression ratio can be
further improved by removing the textured image redundancy with the help of geometric repetition
information.
© ISO/IEC 2017 – All rights reserved 9
---------------------- Page: 12 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
a) Input 3D model b) Original textured c) Re-organized textured image
with texture image containing after removing redundancy
many repeated parts
Figure Amd4.11 — Reconstruction of a textured image
Let I and P denote the instance and the corresponding pattern, respectively. Let I and P denote the
d d
reconstructed I and P, respectively. Two options for reconstructing the textured image(s)/portion(s) of
I are proposed.
d
a) Regular textured image reconstruction mode
I uses the reconstructed textured image (portion) indicated by the reconstructed texture coordinates
d
of itself or P . There are three possible cases that could invoke this mode in the encoder side.
d
— I and P use the same texture coordinates.
— The textured image contents of I and P are extremely similar, although they use different texture
coordinates. In this case, the redundant portion of the textured image will be removed before
compression and I will reuse the texture coordinates of P .
d d
— I and P use the different texture coordinates and their textured image contents are different. In this
case, I will have its own reconstructed texture coordinates.
d
b) Compensated textured image reconstruction mode
The textured image(s)/portion(s) of I is reconstructed from the textured image(s)/portion(s) indicated
d
by the reconstructed texture coordinates of I and P . This mode will be invoked at the encoder
d d
side if the texture content of I and P are similar but have non-ignorable difference. In this case, the
corresponding textured image(s)/portion(s) indicated by the texture coordinates of I is updated to the
difference from the texture content of P. The textured image(s)/portion(s) of I is reconstructed by the
d
following formula:
Ix,,yI= xy −Ix,,yD+128 eTEX xy =TEXx,,yT+ EX x yy −128
() () () () () ()
II PI IP
dd d
where
is the reconstructed textured image of I ;
d
DeTEXx ,y
()
I
d
are the textured images indicated by the reconstructed texture coordi-
TEXx , y and
()
I
d
nates of I and P .
d d
TEXx(),y
P
d
10 © ISO/IEC 2017 – All rights reserved
---------------------- Page: 13 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
Using the method described above, both the texture coordinates and textured images need to be
modified. The details of bitstream definition related to texture coordinate decoding is described in
5.2.7.4.11. The details of textured image decoding are out of the scope of this document.
5.2.7.2.6 Reconstruction of the unique part
Unique part refers to those components of the original model which are not repetitive. The components
belonging to unique part are called as unique components. The unique components are reconstructed
by translating the corresponding decoded components to the reconstructed positions of the original
components.
Although the centre of those components need to be compressed and outputted to the bitstream later,
the rate-distortion performance will be optimized as the bounding box of those components could be
much smaller (in most cases) and less quantization bits are required for the same coding error.
Especially, if there is no or not enough repetitive structures to guarantee the bitrates saving using the
pattern-instance representation, all the components of the input 3D model will be regarded as unique
part. Compared with using the raw representation, there will still be bitrates saving.
5.2.7.2.7 Reconstruction of symmetric structures
As described in 5.2.7.2.1, there might be repetitive structures among various components and/or within
one component. The former is called unconnected repetitive structures because any instance of this kind
of repetitive structures does not share boundaries with the other parts of the 3D model. The latter is
called connected repetitive structures or symmetric structures because any instance of the symmetric
structures shares boundaries with the other parts of the 3D model.
The boundary of an instance could be defined as consisting of vertices, triangles or other elements
of the 3D model. Because of the unavoidable vertex position coding error introduced by most 3D
model compression algorithms, using the same method to represent and compress symmetric
structures might cause cracks on the boundaries of the decoded symmetric instances, as shown in
Figures Amd4.12 a) and b). In order to avoid these cracks, stitching information should be recorded in
the compressed bitstream for stitching the decoded symmetric instances and their adjacent parts in the
3D model. One example of the stitching information, as shown in Figure Amd4.12 c), is the difference
between the vertex positions recovered from different instances which correspond to the same vertex
on the original 3D model.
© ISO/IEC 2017 – All rights reserved 11
---------------------- Page: 14 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
a) Common boundary b) Cracks between c) Stitching information
in the input model adjacent instances could be the difference be-
in the decoded model tween vertices (black lines)
NOTE The crack on the common boundary of adjacent symmetric instances caused reconstruction during
decoding and one example of the stitching information for removing the crack.
Key
1 symmetric instance, I
1
2 symmetric instance, I
2
3 decoded symmetric instance, I
1
4 decoded symmetric instance, I
2
a
Common boundaries of I and I .
1 2
b
The crack caused by recovering I and I from different decoded patterns and transformations.
1 2
Figure Amd4.12 — Reconstruction of symmetric structures
For one unconnected repetitive structure whose pattern consists of symmetric structures, there
are two options to represent its instances. Let I and P denote the instance and the corresponding
pattern, respectively. Let P denote the pattern of the symmetric structure belonging to P. For the sake
S
of simplicity, suppose P has no unique part when represented by the instances of P , then I could be
S
represented as:
— Option A: one instance of P, which is represented by instances of P ;
S
— Option B: instances of P .
S
As instances of symmetric structures might decrease the quality of the decoded 3D model and need
extra stitching information, option A is chosen to limit the number of symmetric instances.
Thus, for the purpose of benefiting compression, discovering repetitive structures on the surface of a
3D model can be divided into two steps:
— first, discover the unconnected repetitive structures;
— then, discover the symmetric structures on the surfaces of the patterns of the unconnected repetitive
structures and the unique part.
As a result of the two-step repetitive structure discovery, the entire original 3D model is represented
by the following:
a) patterns, which consist of the patterns of unconnected repetitive structures not including any
symmetric structures (D), the patterns of symmetric structures (H), the unique parts on those
unconnected-repetitive-structure patterns including symmetric structures (K), and the unique
parts of those unique components which do not belong to any unconnected repetitive structures
but have symmetric structures (J). All the objects here are indexed according to their actual
position in the bitstream;
12 © ISO/IEC 2017 – All rights reserved
---------------------- Page: 15 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
b) instances of symmetric structure patterns (I) (only one pattern per symmetric structure), which
are represented by the pattern IDs (actual position in the bitstream) and transformations from the
symmetric patterns to instances;
c) stitching information that is used to stitch the decoded symmetric pattern instances and their
adjacent parts on the decoded 3D model, which could be other decoded symmetric pattern
instances or the two types of unique parts defined in a);
d) instances of unconnected-repetitive-structure patterns (B);
e) unique components (G), which do not belong to any unconnected repetitive structures and do not
include any symmetric structures;
The decoder reconstructs the original 3D model as follows.
— Reconstruct the unique components, which do not belong to any unconnected repetitive structures
and do not include any symmetric structures.
Reconstruct all patterns defined in a). All the patterns defined in a) are indexed according to their
positions in the bitstream.
— Reconstruct all symmetric instances.
— Reconstruct all unconnected-repetitive-structure patterns and unique components which include
symmetric structures, using the recovered patterns, symmetric instances and stitching information.
Each unconnected-repetitive-structure pattern or unique component is indexed using the minimum
index of its patterns defined in a).
— Reconstruct all unconnected-repetitive-structure instances.
5.2.7.2.8 Error compensation
When reconstructing an instance by applying transformation matrix on the corresponding pattern, it
may result in a larger reconstruction error than when directly decoding the corresponding component.
Thus, an error compensation mode may be used for compensating the reconstruction error of the
vertices of those instances which suffer from large vertex reconstruction error. The reconstruction error
of a vertex is defined as the distance between its original and reconstructed positions. As the amount
of vertex reconstruction error to be encoded may vary drastically, different levels of quantization are
used for different vertex reconstruction error according to its scale.
The information related with error compensation is recorded in three layers, i.e. the bitstream, instance
and vertex layer, as shown in Figure Amd4.13.
At the bitstream header layer, an error compensation flag informs whether or not error compensation
data is present in the bitstream. When this flag is set to 1 (true), a lookup table containing all three
numbers of quantization bits that can be used for error compensation follows.
At the instance layer, each instance has one flag to indicate whether or not error compensation data is
present for at least one of its vertices. If no error compensation data is present in the whole bitstream
(as indicated by the error compensation flag in the header layer), all these instance error compensation
flags are omitted.
At the vertex layer, if the error compensation mode of its parent instance is activated, a quantization
bits (QB) lookup table ID is present in order to indicate to the decoder the number of quantization bits
used when encoding.
NOTE The choice of a 3-value QB lookup table rather than directly recording the number of quantization
bits for each vertex allows saving 3 bits per vertex (2 bits for QB lookup table entry vs 5 bits for number of
quantization bits). Since there are many vertex information entries in the bitstream, the introduction of such a
lookup table helps saving significant bitrate.
© ISO/IEC 2017 – All rights reserved 13
---------------------- Page: 16 ----------------------
ISO/IEC 14496-16:2011/Amd.4:2017(E)
All vertices of an instance in which at least one vertex has error compensation data contain QB lookup
table ID. The QB lookup table allows using three different levels of quantization since the first ID (0) of
the lookup table is reserved for vertices that do not have compensated reconstruction error.
NOTE The information
...
Questions, Comments and Discussion
Ask us and Technical Secretary will try to provide an answer. You can facilitate discussion about the standard in here.