JPEG 2000 Image Compression JPEG 2000 is frame accurate, in that every single frame of the input

is contained in the compressed format. MPEG systems, on the

By Christine Bako [[email protected]] other hand, reduce the amount of data through temporal compression
(which does not encode each frame as a complete image), so
INTRODUCTION MPEG compression is not frame-accurate. For this reason, legal
The JPEG (Joint Photographic Experts Group) 2000 standard, issues restrict the use of MPEG compression in some security
finalized in 2001, defines a new image- coding scheme using applications. To get around this problem, security system and
state - of-the - art compression techniques based on wavelet equipment providers have had to develop their own compression
technology. Its architecture is useful for many diverse applications, schemes—or use the highly inefficient motion JPEG (M-JPEG)
including Internet image distribution, security systems, digital compression standard—in order to provide a compressed stream
photography, and medical imaging. that contains every single field of the original. They can now use
A lot of confusion exists as to what JPEG 2000 is and how it compares JPEG 2000 for new designs.
with other compression standards such as MPEG (Moving-Picture Internet Image Distribution
Experts Group) -2, MPEG-4, and the earlier JPEG. With brief Progressive coding, another feature of the JPEG 2000 standard,
comparisons to other compression standards, this article is primarily means that the bit stream can be coded in such a way as to contain
intended to highlight some of the often misunderstood and rarely less-detailed information at the beginning of the stream and more
mentioned potential-become-actual benefits of JPEG 2000. detailed information as the stream progresses. This makes it ideal
Applications for Internet/network applications—especially with large images
CCTV Security and low bandwidths—as the image can be seen instantly on the
When transmitting or storing picture information, compression decoding side, even with low-speed networks or image databases.
must be employed to maintain picture resolution while making The lower subbands are shown first, and more detail is added as
best use of limited channel bandwidth. Compression is defined as time progresses. The picture thus becomes sharper and more
lossless if full recovery of the original is available from the channel detailed over time, and the entire image does not have to be
without any loss of information; otherwise, it is lossy. Standards downloaded before it can be seen.
are required to ensure interoperability. JPEG 2000 is the only With the low-quality image instantly available, the user at the
standard compression scheme that provides for both lossless and receiving end can decide whether to view the picture in its fully
lossy compression. As such, it lends itself to applications that decoded version, or to pass it by and scan the next picture instead.
require high-quality images despite limitations on storage or Clients can view images at different resolutions or quality levels
transmission bandwidths. [compression rates] making them suitable for any transmission
An important feature of systems based on JPEG 2000 is the ability bandwidth, connection speed, or display device. In addition,
to extract a variety of resolutions, components, areas of interest, JPEG 2000 coding provides the option to zoom in or out on a
and compression ratios from a single JPEG 2000 code stream. This particular area of the image—or to display a particular region of
is not possible with any other compression standard because the the image at a different resolution or compression rate.
image size, bit rate, and quality must be specified on the encode High Definition
side and can not be determined or changed on the decode side. At extreme compression levels, JPEG 2000 video starts to blur,
For example, a closed- circuit TV (CCTV) security system can but is still quite viewable. MPEG or JPEG artifacts are much more
make use of this feature by sending a single JPEG 2000 code disturbing to the eye, with the picture visibly broken down into
stream over a low bandwidth network. High-resolution images small blocks at high compression ratios. The high image quality at
can be stored on a hard-disk drive (HDD), while several lower- medium-to-high bit rates and contents that contain a lot of motion,
resolution images are displayed on monitors. The operator on lack of block artifacts, and high efficiency make JPEG 2000 ideal
the receive side can decide what information to extract from for high-definition (HD) applications, such as digital cinema, HD
the single code stream sent. recording systems, and HD camera equipment.

MILITARY MEDICAL
• HD SATELLITE IMAGES • HIGH-QUALITY STORAGE
• HIGH-QUALITY TRANSMISSION • HIGH-QUALITY PROCESSING
• HIGH-QUALITY STORAGE
• REMOTE SENSING
INDUSTRIAL CONSUMER
• HIGH-QUALITY STORAGE • MOBILE PHONES
• REMOTE SENSING DIGITAL • PALM PILOTS
STILL
IMAGES

OFFICE HD/SD VIDEO

COPIER AUTOMATION PRODUCTION PROFESSIONAL
NETWORK • PROFESSIONAL CAMCORDER
SCANNERS • HDTV VIDEO EDITING
JPEG 2000 • DIGITAL CINEMA
SERVERS
CONSUMER
• HD CAMCORDER
• DIGITAL CINEMA

INTERNET CCTV
IMAGE DATABASES SECURITY MOTION DETECTION
STREAMING VIDEO NETWORK DISTRIBUTION
VIDEO SERVERS STORAGE

Figure 1. JPEG 2000 applications.

Analog Dialogue 38-09, September (2004) http://www.analog.com/analogdialogue 1

Many applications require exact bit-rate control, which only coefficients that describe a specific horizontal and vertical spatial
JPEG 2000 can provide. Exact bit-rate control is possible frequency range of the entire original image. This means that
because an entire frame or field is transformed at once; it is then lower-frequency, less-detailed information is contained in the
broken down into bit streams or code blocks that can be processed first transform level, while more- detailed, higher-frequency
independently with the techniques described below. In systems information is contained in higher transform levels. For simplicity,
using DCT, quantization is the only technique used, and this only two levels of transform are shown here. The first transform
makes exact bit-rate control difficult. In order to control bit rate level results in subbands LH1, HH1, HL1, and LL1. Only subband
in DCT systems, the information must be repeatedly re-processed LL1 is passed on for further filtering, generating the next transform
and re-quantized. The rate-control algorithm used in JPEG level and creating subbands LH2, HH2, HL2, and LL2.
2000 truncates each bit stream to meet a specific target bit rate, Equally sized code blocks, which are essentially bit streams of data,
adjusting the truncation and re-quantization of each code block’s are generated within each subband. This break-down is necessary
data as required. In addition to programming the target bit rate, for coefficient modeling and coding, and is done on a code-block-
the standard allows the user to specify a particular quality metric. by-code-block basis. In essence, the actual compression is achieved
In this case, the target bit rate will vary to meet the specified by truncating and/or re-quantizing the bit streams contained in each
quality factor, as long as the performance does not fall below a code block. These bit streams are then optimally truncated using
specific peak signal-to-noise ratio. The PSNR is an indication of a technique knows as post-compression-rate-control (PCRC).
picture quality comparable to perceived picture quality.
Code blocks can be accessed independently. Their bit streams are
JPEG 2000 Code Stream coded with three coding passes per bit plane. This process, called
A given input image or part of the image [tile] is sent to a set of context modeling, is used to assign information about the importance
wavelet filters, which transform the pixel information into of each individual coefficient bit. The code blocks can then be
wavelet coefficients, which are then grouped into several grouped according to their significance. On the decoding side it is
subbands [the use of wavelets in encoding was first explained in then possible to extract information according to its significance,
Analog Dialogue 30 -2 (1996)]. Each subband contains wavelet allowing the most significant information to be seen first.

ORIGINAL TILE T0: PRECINCT 0 OF HL1

LL2 HL2
HL1

WAVELET
TRANSFORM INTO LH2 HH2
SUBBANDS HL1, cb0
HH1, LH1, LL2, cb1
HL2, HH2, LH2

LH1 HH1

cb2 cb4
cb3 cb5

PRECINCT 0 OF LH1 PRECINCT 0 OF HH1

WAVELET COEFFICIENT DATA IS ARRANGED INTO

THE JPEG 2000 CODESTREAM

TILE0

RESOLUTION0 R1

LAYER0 L1 L0

Y Cb Cr Y Cb Cr Y Cb

PRECINCT0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2

PACKET 0 PACKET 1 PACKET 2 PACKET 3 PACKET 4

PACKET DATA FOR DATA FOR DATA FOR DATA FOR DATA FOR DATA FOR
HEADER FOR CODEBLOCK CODEBLOCK CODEBLOCK CODEBLOCK CODEBLOCK CODEBLOCK
PACKET 0 0, SUBBAND 1, SUBBAND 2, SUBBAND 3, SUBBAND 4, SUBBAND 5, SUBBAND
HL1 HL1 LH1 LH1 HH1 HH1

Figure 2. ENCODE—image over wavelet transform into subbands and resolutions.

2 Analog Dialogue 38-09, September (2004)

 TILE0

RESOLUTION0 R1

LAYER0 L1 L0

Y Cb Cr Y Cb Cr Y Cb

PRECINCT0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2 Pr0 Pr1 Pr2

DECODER 1: DECODER 2: DECODER 3:

HDD RECORDER HIGH QUALITY PRINTER CCTV DISPLAY

REQUESTS FULL RESOLUTION [R0] REQUESTS FULL RESOLUTION [R0] REQUESTS SMALL RESOLUTION [R1]
LOSSLESS [L0] LOSSLESS [L1] LOSSY [L0]
ALL COMPONENTS PRECINCT 0, ALL COMPONENTS Y COMPONENT ONLY
PACKET 0

Figure 3. DECODE—one JPEG 2000 stream is received by several decoders.

JPEG 2000 can contain a user-defined number of layers, which header information. Also, a table of contents can be stored on the
are defined by PCRC and context modeling. Each layer stands encode side, and allows a decoder to call up a certain resolution
for a particular compression rate, where the compression rate is on demand, without first having to decode or download the entire
achieved from the quantization-, rate-distortion-, and context JPEG 2000 code stream.
modeling processes. Layer 0, for example, contains bit streams-
from the lossy WT transform-that are heavily truncated, contain DCT versus WT
no coding passes, and thus provide the highest compression rate JPEG 2000 uses the wavelet transform (WT) to reduce the amount of
and the lowest quality. Layer 16 can then contain bit streams that information contained in a picture, while MPEG and JPEG systems
are less truncated and use a higher number of coding passes, thus use the discrete cosine transform (DCT). It is true that the WT requires
providing low compression and high quality. more processing power than the DCT, but MPEG systems require
more than just the DCT. The DCT, or any type of Fourier transform,
Tiles or images are further partitioned into precincts. Precincts expresses the signal in terms of frequency and amplitude—but only at
contain a number of code blocks, and are used to facilitate access a single instant in time. The WT transforms a signal into frequency
to a specific area within an image in order to process this area and amplitude over time, and is therefore more efficient. The figures
in a different way, or to decode only a specific area of an image. on the following page demonstrate this.
The JPEG 2000 bit stream is generated by arranging code blocks
or precincts into an array of packets with the lower subbands To obtain the same amount of information as with one WT pass,
coming first. the DCT must be used for every frequency; and each of these
frequencies must be transformed at each time instant for each
The JPEG 2000 stream starts with a main header containing 8  8 pixel block. In addition, MPEG systems use inter-frame
information such as: uncompressed image size, tile size, compression [motion estimation] in order to reduce the amount of
number of components, bit depth of components, coding data further for motion estimation. This requires storage of at least
style, transform levels, progression order, number of layers, two entire fields in external memory. The computation-intensive
code block size, wavelet filter type, quantization level, etc. The motion estimation process requires a very powerful processor.
entire image data, grouped in code blocks of LL, HL, LH, and Temporal compression can be used in JPEG 2000 systems, but it
HH subbands, follows the header. Data is not contained in the is not inherent in the JPEG 2000 standard.

Analog Dialogue 38-09, September (2004) 3

 4 1.0

3 0.8

0.6
2
0.4
1
0.2

0 0

–1 –0.2

–0.4
–2
–0.6
–3
–0.8

–4 –1.0
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200

Figure 4. Input signal 1 containing frequencies A, B, C, and D. Figure 5. Input signal 2 containing frequencies A, B, C, and D.

0.10
0.10

0.05
0.05

0
0

–0.05
–0.05

–0.10
–0.10

–0.15
–0.15 1200
500 1000
600
400 700 800 500
300 600 600 400
500 400 300
200 400 200
300 200 100
100 200
100 0
0

Figure 6. Wavelet transform of signal 1. Figure 7. Wavelet transform of signal 2.

FOURIER TRANSFORM OF THE COSINE SIGNAL FOURIER TRANSFORM OF THE COSINE SIGNAL
WITH 4 FREQUENCY COMPONENTS WITH 4 FREQUENCY COMPONENTS
500 500

450 450

400 400

350 350

300 300

250 250

200 200

150 150

100 100

50 50

0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
FREQUENCY (Hz) FREQUENCY (Hz)

Figure 8. Fourier transform of signal 1. Figure 9. Fourier transform of signal 2.

4 Analog Dialogue 38-09, September (2004)

 ADV202
WAVELET EC1 EC2 EC3
PIXEL I/F PIXEL I/F ENGINE

EXTERNAL
DMA CTRL
PIXEL
FIFO
HOST I/F INTERNAL BUS AND DMA ENGINE
CODE
FIFO
ATTRIBUTE
FIFO
EMBEDDED RISC MEMORY
ANCILLARY PROCESSOR SYSTEM
FIFO SYSTEM

Figure 10. ADV202 block diagram.

JPEG 2000’s Advantages Over Other Compression Standards provides a glueless interface to common video standards such as
All MPEG standards are complex and computation intensive. ITU.R.BT656, SMPTE274M, or SMPTE296M. It can create a
This translates into extensive processing latency and memory fully compliant JPEG 2000 code stream [.j2c, .jp2]. It can also
requirements in standard-definition (SD) applications. These provide raw code - block and attribute data, allowing the host
factors become even more of a problem when high-definition processor to have complete control over the generation - and
(HD) formats are considered, and JPEG 2000 becomes even compression processes.
more desirable. Another strength of JPEG 2000 is the standard Even though digital signal-processor (DSP) performance has
itself, which allows immense flexibility and control in many improved significantly, a DSP would have to perform 20 billion
different applications. There is also much versatility regarding instructions per second to match the performance of the ADV202
formats: JPEG 2000 supports anything from 8-bits per sample in a standard-definition encode application. Effectively serving
to an unlimited amount of bits per sample, whereas MPEG only as accelerators, the ADV202’s three dedicated on-chip entropy
supports 8-bit data. codecs are responsible for the high throughput rate.
JPEG 2000 continues to gain popularity, even though MPEG-2 is
the established standard for DVD and broadcast applications. JPEG CONCLUSION—THE OUTLOOK FOR JPEG 2000
2000 is also very popular in HD applications that require high-quality A major advantage of using a JPEG 2000 hardware solution is lower
storage or transmission of HD images over wireless or other links. latency than any other compression scheme, a factor which is of
particular importance in medical applications, for example.
The ADV202
Since the early 1990s, Analog Devices has invested heavily in Several major manufacturers of video or broadcast equipment
wavelet-compression R&D. We were the first to introduce a wavelet- have implemented JPEG 2000 into such future HD products as
compression hardware solution in 1996 with the ADV601. Now real-time encoding and decoding systems and video servers.
ADI’s newest wavelet codec, the ADV202, released in July 2004, is
The Digital Cinema Initiative (DCI) has recently announced
so far the only dedicated JPEG 2000 IC on the market. A complete
that it will use JPEG 2000 as the compression method in the
single-chip JPEG 2000 compression/decompression IC, the ADV202
delivery of digital motion pictures. The ADV202 has already
works with high-definition video, standard-definition video, and still
found its way into many designs in the CCTV/security market
images. It supports all features of the ISO/IEC15444-1 [JPEG 2000]
in video-over-network applications. For more information, please
image-compression standard [except Maxshift ROI]. Its patented
visit http://www.itscj.ipsj.or.jp/sc29/29w02901.pdf
SURF™ (spatial ultra-efficient recursive filtering) technology enables
low-power, low-cost wavelet-based compression. Containing a Because of its flexibility and image-compression quality, the
dedicated wavelet transform engine, three entropy codecs, a ADV202—operating under JPEG 2000—could find its way into
RISC processor, and on-board memory systems, the ADV202 virtually every design that uses image or video compression. b

Analog Dialogue 38-09, September (2004) 5