ISO/IEC 21122-2:2019

INTERNATIONAL ISO/IEC
STANDARD 21122-2
First edition
2019-07
Information technology — JPEG XS
low-latency lightweight image coding
system —
Part 2:
Profiles and buffer models
Reference number
ISO/IEC 21122-2:2019(E)
©
ISO/IEC 2019

---------------------- Page: 1 ----------------------
ISO/IEC 21122-2:2019(E)

COPYRIGHT PROTECTED DOCUMENT
© ISO/IEC 2019
All rights reserved. Unless otherwise specified, or required in the context of its implementation, 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
CP 401 • Ch. de Blandonnet 8
CH-1214 Vernier, Geneva
Phone: +41 22 749 01 11
Fax: +41 22 749 09 47
Email: copyright@iso.org
Website: www.iso.org
Published in Switzerland
ii © ISO/IEC 2019 – All rights reserved

---------------------- Page: 2 ----------------------
ISO/IEC 21122-2:2019(E)

Contents Page
Foreword .iv
Introduction .v
1 Scope . 1
2 Normative references . 1
3 Terms, definitions, symbols and abbreviated terms . 1
3.1 Terms and definitions . 1
3.2 Conformance language . 4
3.3 Operators . 4
3.3.1 Arithmetic operators . 4
3.3.2 Logical operators . 4
3.3.3 Relational operators . 4
3.3.4 Precedence order of operators . 4
3.3.5 Mathematical functions . 5
4 Specifications . 5
4.1 Symbols . 5
4.2 Abbreviated terms . 7
4.3 General provisions . 7
5 Buffer model . 8
5.1 General system block diagram . 8
5.2 Influencing variables on the required buffer sizes . 9
5.3 Role of the buffer model . 9
Annex A (normative) Profiles, levels and sublevels .10
Annex B (normative) Packet-based JPEG XS decoder model .22
Annex C (normative) Packet-based constant bit rate buffer model .28
Annex D (informative) Encoder model, latency bounds and codestream conformance
properties for the packet-based constant bit rate buffer model .35
Annex E (informative) JPEG XS latency analysis .40
Bibliography .47
© ISO/IEC 2019 – All rights reserved iii

---------------------- Page: 3 ----------------------
ISO/IEC 21122-2:2019(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.
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) or the IEC
list of patent declarations received (see http: //patents .iec .ch).
Any trade name used in this document is information given for the convenience of users and does not
constitute an endorsement.
For an explanation of 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 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 21122 series can be found on the ISO website.
Any feedback or questions on this document should be directed to the user’s national standards body. A
complete listing of these bodies can be found at www .iso .org/members .html.
iv © ISO/IEC 2019 – All rights reserved

---------------------- Page: 4 ----------------------
ISO/IEC 21122-2:2019(E)

Introduction
ISO/IEC 21122-1 (JPEG XS) specifies a single syntax designed to serve a wide range of applications, bit
rates, resolutions, qualities, and services. Its main target applications are video transport over video
links and IP networks, real-time video storage, video memory buffer, omni-directional video capture
system, head-mounted displays for virtual or augmented reality and sensor compression for the
automotive industry. These applications have different requirements in terms of complexity, latency
and compression efficiency. Even within a given application field, different requirements are usually
identified depending on the targeted use case.
Considering the impracticality of implementing the full syntax of ISO/IEC 21122-1, and in order to
meet the requirements of the different target applications while safeguarding as much as possible the
interoperability enabled by the common syntax defined in ISO/IEC 21122-1, a limited number of subsets
of this syntax are stipulated by means of “profiles”, “levels”, and “sublevels”.
The coding tools specified in ISO/IEC 21122-1 allow encoder and decoder implementations to limit the
end-to-end latency to a fraction of the frame size. To ensure this property, this document specifies a
buffer model, consisting of a decoder model and a transmission channel model.
© ISO/IEC 2019 – All rights reserved v

---------------------- Page: 5 ----------------------
INTERNATIONAL STANDARD ISO/IEC 21122-2:2019(E)
Information technology — JPEG XS low-latency lightweight
image coding system —
Part 2:
Profiles and buffer models
1 Scope
This document defines a limited number of subsets of the syntax specified in ISO/IEC 21122-1 and
a buffer model to ensure interoperability between implementations in the presence of a latency
constraint.
2 Normative references
The following documents are referred to in the text in such a way that some or all of their content
constitutes requirements of this document. For dated references, only the edition cited applies. For
undated references, the latest edition of the referenced document (including any amendments) applies.
ISO/IEC 21122-1, JPEG XS low-latency lightweight image coding system — Part 1: Core coding system
3 Terms, definitions, symbols and abbreviated terms
3.1 Terms and definitions
For the purposes of this document, the terms and definitions given in ISO/IEC 21122-1 and the
following apply.
ISO and IEC maintain terminological databases for use in standardization at the following addresses:
— ISO Online browsing platform: available at https: //www .iso .org/obp
— IEC Electropedia: available at http: //www .electropedia .org/
3.1.1
blanking codestream fragment
placeholder codestream fragment (3.1.8) representing blanking periods
3.1.2
horizontal blanking period
timespan expressed in units of the grid point sampling rate between the last pixel (3.1.22) of an image
line ― not being the last line of an image ― and the first pixel of the next image line
3.1.3
vertical blanking period
timespan in units of the grid point sampling rate between the last line of an image [including the
horizontal blanking periods (3.1.2)] and the first line of the next image
3.1.4
buffer model
combination of a decoder model (3.1.12) and a channel model (3.1.6) whose behaviour can be defined by
a set of parameters
© ISO/IEC 2019 – All rights reserved 1

---------------------- Page: 6 ----------------------
ISO/IEC 21122-2:2019(E)

3.1.5
buffer model instance
specific configuration of a buffer model (3.1.4) specified by the assignment of well-defined values to the
buffer model parameters
3.1.6
channel model
model describing the temporal behaviour of the transmission channel (3.1.29) connecting an encoder
and a decoder
3.1.7
coded codestream fragment
continuous sequence of bits in the codestream containing exactly one packet body and a well-defined
number of packet headers, markers and marker segments
3.1.8
codestream fragment
either coded codestream fragment (3.1.7), or blanking codestream fragment (3.1.1)
3.1.9
code group
group of quantization indices in sign-magnitude representation representing a quantized coefficient
group (3.1.10)
3.1.10
coefficient group
number of horizontally adjacent wavelet coefficients from the same band and precinct
3.1.11
cycle
single clock period of an encoder or decoder clocked implementation
3.1.12
decoder model
combination of a decoder unit (3.1.14) and a decoder smoothing buffer (3.1.13)
3.1.13
decoder smoothing buffer
memory buffer that is used to level out changes in the number of bits read by a decoder unit (3.1.14) per
time unit
3.1.14
decoder unit
module reading a variable number of bits per time unit to generate decoded output pixels (3.1.22) with
a fixed rate
3.1.15
decomposition level
set of wavelet coefficients resulting from a particular level (3.1.21) of recursive application of a wavelet
transform
3.1.16
encoder model
combination of an encoder unit (3.1.18) and an encoder smoothing buffer (3.1.17)
3.1.17
encoder smoothing buffer
memory buffer that is used to level out changes in the number of bits generated by an encoder unit
(3.1.18) per time unit
2 © ISO/IEC 2019 – All rights reserved

---------------------- Page: 7 ----------------------
ISO/IEC 21122-2:2019(E)

3.1.18
encoder unit
module transforming a sequence of input pixels (3.1.22) with constant rate into a conforming
codestream, producing a bit sequence with variable number of bits generated per time unit
3.1.19
fill level
number of bits stored in the encoder or decoder smoothing buffer (3.1.13)
3.1.20
nominal bits per pixel value
mean number of bits allocated per encoded pixel (3.1.22) which is used to derive the sublevel (3.1.28)
constraints by assuming an image with well-defined dimensions and frame rate derived from the level
(3.1.21)
3.1.21
level
defined set of constraints on the amount of decoded sampling grid points (3.1.25) to be processed by an
encoder or decoder, both in the spatial and time dimensions
Note 1 to entry: The same set of levels is defined for all profiles. Individual implementations may, within the
specified constraints, support a different level for each supported profile.
3.1.22
pixel
position in the sample grid (3.1.24) that is populated by a sample value of at least one component
3.1.23
profile
specified subset of the codestream syntax together with admissible parameter values
3.1.24
sample grid
abstract coordinate system on which image sample values are positioned
3.1.25
sampling grid point
position on the sample grid (3.1.24), specified by integer horizontal and vertical offset relative to the
origin of the sample grid
3.1.26
smoothing buffer unit
level (3.1.21) and sublevel (3.1.28) dependent number of bits by which the smoothing buffer size of the
decoder model (3.1.12) is specified
3.1.27
start of transmission
SoT
time at which the transmission channel (3.1.29) starts transmission relative to the start of encoding of
the first codestream fragment (3.1.8) of a codestream
3.1.28
sublevel
defined set of constraints on the amount of codestream bits to be processed by an encoder or decoder,
per unit of time, per column, and per image
Note 1 to entry: The same set of sublevels is defined for all profiles. Individual implementations may, within the
specified constraints, support a different sublevel for each supported profile.
© ISO/IEC 2019 – All rights reserved 3

---------------------- Page: 8 ----------------------
ISO/IEC 21122-2:2019(E)

3.1.29
transmission channel
facility transferring bits from a source entity to a target entity
3.1.30
transmission channel capacity
maximum number of bits per time unit that a transmission channel (3.1.29) can transfer from a source
entity to a target entityConventions
3.2 Conformance language
ISO/IEC’s use of verbal forms is detailed at:
https: //www .iso .org/foreword -supplementary -information .html
The keyword "reserved" indicates a provision that is not specified at this time, shall not be used, and
may be specified in the future. The keyword "forbidden" indicates "reserved" and in addition indicates
that the provision will never be specified in the future.
3.3 Operators
NOTE Many of the operators used in document are similar to those used in the C programming language.
3.3.1 Arithmetic operators
+ addition
− subtraction (as a binary operator) or negation (as a unary prefix operator)
× multiplication
/ division without truncation or rounding
3.3.2 Logical operators
|| logical OR
&& logical AND
! logical NOT
3.3.3 Relational operators
> greater than
≥ greater than or equal to
< less than
≤ less than or equal to
== equal to
!= not equal to
3.3.4 Precedence order of operators
Operators are listed in descending order of precedence. If several operators appear in the same line,
they have equal precedence. When several operators of equal precedence appear at the same level in an
4 © ISO/IEC 2019 – All rights reserved

---------------------- Page: 9 ----------------------
ISO/IEC 21122-2:2019(E)

expression, evaluation proceeds according to the associativity of the operator either from right to left
or from left to right.
Operators Type of operation Associativity
() expression left to right
[] indexing of arrays left to right
– unary negation
×, / multiplication, division left to right
+, − addition and subtraction left to right
< , >, ≤, ≥ relational left to right
& bitwise AND left to right
| bitwise OR left to right
3.3.5 Mathematical functions
ceil of x: returns the smallest integer that is greater than or equal to x
x
 
 
floor of x: returns the largest integer that is less than or equal to x
x
 
 
|x| absolute value of x, |x| equals –x for x < 0, otherwise x
sign(x) sign of x, 0 if x is 0, +1 if x is positive, −1 if x is negative
10t ≥
ξ t
()
step function ξ t =
() 
0otherwise

max (x ) maximum of a sequence of numbers [x ] enumerated by the index i
i i i
4 Specifications
4.1 Symbols
A = [a ,a ,…,a ] sequence of elements a – a
1 2 n 1 n
concatenation of two sequences A and B
AB||
C(i) codestream i
D number of clock cycles between the first bit written into the decoding smoothing
c2d
buffer and the decoding start of the first codestream fragment of a stream of code-
stream fragments
F (C(i)) first codestream fragment of codestream C(i)
first
F (C(i)) last codestream fragment of codestream C(i)
last
H height of the image in sampling grid points
f
© ISO/IEC 2019 – All rights reserved 5

---------------------- Page: 10 ----------------------
ISO/IEC 21122-2:2019(E)

H maximum picture height in sampling grid points
max
L maximum number of sampling grid points per image
max
l (t) fill level of the encoding smoothing buffer in bits at the end of cycle t
enc
l (t) fill level of the decoding smoothing buffer in bits at the end of cycle t
dec
l capacity in bits of the encoding smoothing buffer
enc,max
l capacity in bits of the decoding smoothing buffer
dec,max
 number of bits that can be read from the decoding smoothing buffer in cycle t
l t
()
dec
l (t) sum of encoder and decoder smoothing buffer fill level in bits at cycle t
sum
all integer numbers being strictly larger than zero

all integer numbers being greater than or equal to zero

0
N size of the horizontal blanking line in sampling grid point clock periods
b,x
N size of the vertical blanking period in sampling grid lines
b,y
N nominal number of bits allocated per pixel for compression
bpp
N number of components in an image
c
N (f) number of coefficient groups within codestream fragment f
cg
N number of coefficient groups associated to a codestream fragment representing a
cg,hz
horizontal blanking period
N number of coefficient groups associated to a codestream fragment representing a
cg,vt
vertical blanking period
N (i) number of codestream fragments within a codestream i
f
N number of coefficients in a code group
g
N number of vertical decomposition levels
L,y
N number of precincts per sampling grid line
p,x
N number of precincts per sampling grid column
p,y
N number of decoder smoothing buffer units for a given profile
sbu
set of rational numbers

r (t) number of bits read and removed from the decoder smoothing buffer in clock cycle t
dec
R transmission channel capacity, expressed in bits per cycle (having a duration of T)
trans
R (l , l ) maximum admissible encoded throughput in bits per second for a given level
t,max m s
R max grid point sample rate (in samples per second) at decoder output
s,max
S (f) number of bits forming the codestream fragment f
bits
6 © ISO/IEC 2019 – All rights reserved

---------------------- Page: 11 ----------------------
ISO/IEC 21122-2:2019(E)

S targeted maximum number of bytes of an encoded codestream
c,max
S (l , l ) size of the smoothing buffer unit in bytes for level l and sublevel l
sbu m s m s
S (p) smoothing buffer increment in bits for a profile p
sbo
S (l , l ) maximum size of an encoded codestream in bytes of level l and sublevel l
sl,max m s m s
s [i] subsampling factor of component i in horizontal direction
x
s [i] subsampling factor of component i in vertical direction
y
T clock period defining the frequency by which code groups are processed by an encoder
enc
T clock period defining the frequency by which code groups are processed by a decoder
dec
t (f) timestamp in cycles at which the codestream fragment f is written to the encoder
enc,write
smoothing buffer
t (f) timestamp in cycles at which decoder starts decoding codestream fragment f
dec,start
t (f) timestamp in cycles at which codestream fragment f is removed from the decoder
dec,read
smoothing buffer
Tbmd buffer model type
W [i] width of component i in samples
c
W maximum column width in sampling grid points for a given profile
c,max
w (t) number of bits written into the decoder smoothing buffer in clock cycle t
dec
W width of the image in sampling grid points
f
W maximum picture width in sampling grid points
max
4.2 Abbreviated terms
bpp bits per pixel
DWT discrete wavelet transform
IDWT inverse discrete wavelet transform
RCT reversible colour transform
IRCT inverse reversible colour transform
4.3 General provisions
For a concrete application, only a subset of the codestream syntax specified in ISO/IEC 21122-1 is needed.
Profiles as specified in 3.1.23 define corresponding interoperability points for those applications. In
addition to profiles, levels (as specified in 3.1.21) and sublevels (specified in 3.1.28) limit the maximum
throughput in the encoded (codestream) and decoded (pixel, spatial) domain. This allows creating cost-
efficient implementations serving the needs of the desired applications.
Profiles, levels and sublevels shall be as specified in Annex A.
Keeping the end-to-end latency of an encoding-decoding chain under a given threshold is one of the main
goals pursued by the methods defined in ISO/IEC 21122-1. To reach this goal, the definition of buffer
models is necessary, consisting of a decoder model and a transmission channel model. The interaction
© ISO/IEC 2019 – All rights reserved 7

---------------------- Page: 12 ----------------------
ISO/IEC 21122-2:2019(E)

of a hypothetical reference decoder including its decoding smoothing buffer with a constant bitrate
channel feeding this buffer shall be as specified in Annexes B and C. The size of this decoding smoothing
buffer is specified in Annex A. Codestreams shall be formed such that this decoding smoothing buffer
never overflows or underflows.
Buffer models are further discussed in Annex D. The buffer model provides encoders with the necessary
information to generate codestreams that can be decoded by an arbitrary decoder implementation
ensuring system interoperability.
In addition to the size of the decoder smoothing buffer, end-to-end latency also depends on the latency
inherent to each processing step of the encoding-decoding chain whose methods are described
in ISO/IEC 21122-1. To help implementers estimate the latency of their device, Annex E gives useful
information on the minimum latency that can be achieved by the different methods described in
ISO/IEC 21122-1.
5 Buffer model
5.1 General system block diagram
The JPEG XS coding system addresses applications where coded images are transferred from a source
to a target, as shown in Figure 1. To this end, the encoder is compressing a continuous stream of input
pixels into a sequence of bits. These bits are forwarded by means of a transmission channel to the
decoder that decompresses the bits to produce a continuous stream of output pixels.
Key
1 encoder clock
2 pixel data
3 encoder unit
4 encoder smoothing buffer
5 transmission channel
6 decoder smoothing buffer
7 decoder unit
8 decoder clock
9 variable bit rate
Figure 1 — General system block diagram
The time instances at which the encoder has to process each pixel are determined by an encoding clock.
Similarly, the time instances at which the decoder has to produce each output pixel are determined by a
decoding clock. Both clocks are generated by the system.
NOTE In implementations, these clocks can be the same or differ in both frequency and phase. The presented
model is independent of whether clocks are synchronized or not.
8 © ISO/IEC 2019 – All rights reserved

---------------------- Page: 13 ----------------------
ISO/IEC 21122-2:2019(E)

In accordance with ISO/IEC 21122-1, the pixels of an image are translated into coefficient groups
represented as code groups in the codestream. The number of bits necessary to code these code groups
may vary from group to group. As a consequence, the encoder writes encoded bits at a variable rate into
the encoder smoothing buffer. Similarly, the decoder reads the codestream at a variable rate from the
decoder smoothing buffer.
In case the maximum bit rate of the transmission channel is below the peak bit rate generated by the
encoder, an encoder smoothing buffer is necessary to decouple generation of bits by the encoder from
transmission of bits over the transmission channel. Similarly, a decoder smoothing buffer needs to be
provided that decouples the arrival of bits at the rate afforded by the transmission channel and the
consumption of bits by the decoder per clock.
Correct operation requires that the decoder buffer never overflows. This is because the decoder is
not able to pause the arrival of bits from the transmission channel. Moreover, a buffer underflow in
the decoder buffer needs to be avoided. This is because the decoder is required to output pixels in
accordance with the timing of its output interface. Hence it needs to be ensured that the bits to be read
from the decoding buffer to produce the next pixel in accordance with the decoding clock are available
in this decoding buffer.
5.2 Influencing variables on the required buffer sizes
Avoiding any buffer overflow or underflow, as discussed in subclause 5.1, requires properly sizing the
decoder smoothing buffer. Moreover, the time at which decoding starts is delayed relative to the starting
time of encoding and the start of transmission needs to be carefully set. Those values are influenced by
many system parameters, for example:
— The maximum transmission channel bit rate.
— The granularity at which the encoder writes the encoded data and the decoder reads the encoded data.
— The rate control strategy applied by the encoder.
These dependencies cause that encoders and decoders are only interoperable in well-defined conditions.
Defining these conditions is the purpose of the buffer model defined in Annex B and Annex C
5.3 Role of the buffer model
The core coding system defined in ISO/IEC 21122-1 can be implemented on a large variety of platforms
using many different implementation strategies. Thus, interoperability cannot be achieved by precisely
specifying the temporal behaviour of a conforming decoding implementation. Instead, the buffer model
defines a simplified decoder model. Interoperability is then achieved by mandating that a conforming
decoder shall decode all bit streams being decodable by the simplified decoder model. Similarly, a
conforming encoder shall not create bit streams that cannot be decoded by the simplified decoder model.
To this end, Annex B defines a generic JPEG XS decoder model that precisely defines the temporal
behaviour of the decoder model assuming a processing granularity of codestream packets. While such
a model already defines some fundamental properties of the decodable codestreams, it is still not
sufficient to ensure interoperability. The reason is that otherwise codestreams could be constructed
that would only be decodable by the decoder model if the transmission channel could transport bits
arbitrarily fast. In practice, this is obviously not the case. Consequently, interoperability also requires
defining a channel model over which an encoder sends the codestreams to the decoder.
Annex C defines such a cha
...