TAKING IMAGE COMPRESSION INTO THE NEW MILLENNIUM

08 November 2006
In an effort to inform the Icograda community about international standards setting and new technology development, Icograda introduces the first of a series of articles originating from the International Organization for Standardization (ISO).
by Richard Clark and Jean Barda

JPEG 2000, a new standard to enrich imaging applications

The ISO JPEG committee, a meeting place for nationally appointed experts in imaging and communications technology, started to consider what this might be over five years ago. Based on their work on, and on submissions for a new lossless imaging standard (JPEG-LS), they identified a key weakness in the current standard as being its lack of architectural flexibility when integrated into imaging applications.

A range of possible additional features was identified, which were either impossible or impractical to achieve with the previous definition of the JPEG Standard. These have been under constant discussion by the experts since 1997 with the objective of producing a new JPEG standard to take image compression into the new millennium. Named JPEG 2000, the new standard greatly extends the set of capabilities offered by JPEG.

THE CORE OF A NEW JPEG 2000 STANDARD
The JPEG experts, meeting three times a year, managed to produce the core of a new JPEG 2000 standard at the end of year 2000. The structure of this new standard has been designed to be open to additional features and is at present structured as follows:

Part 1: Core coding system - the baseline implementation, intended to be implementable without payment of patent fees to the many patent holders who have contributed to its development;

Part 2: Extensions - adding further optional features which may imply patent license or royalty payments or which simply could not be agreed in time for inclusion in the baseline;

Part 3: Motion JPEG 2000 - allowing the same algorithms to be used on a time-based frame by frame basis, allowing moving and still sequences to be treated as closely aligned, whilst preserving such baseline capabilities as lossless image coding if needed;

Part 4: Compliance testing - providing test sequences which exercise the complexity of the standard by identifying correct coding behaviour;

Part 5: Reference software - to exemplify and demonstrate the standard's practicality, and provide implementers with a base for cross-coder testing;

Part 6: Compound image file format - for use in facsimile and document management applications where a document can be separately coded using optimum algorithms for separate sections and areas;

Part 8: JPSEC Secure JPEG 2000 - adding in a framework and capabilities for rights management, encryption, watermarking and similar technologies;

Part 9: JPIP interactivity tools, APIs and protocols - to allow JPEG 2000 images to be delivered efficiently over wide area networks to a variety of different terminals and applications;

Part 10: JP3D 3-D and floating point data - allowing 3D images to be coded efficiently;

Part 11: JPWL wireless - examining the particular constraints and needs that wireless based communications are likely to imply;

Part 12: ISO base media file format - shared with SC29's other main committee, MPEG, this is a common file format which is used by both JPEG 2000 and MPEG-4, and intended to form the basis for future standards such as MPEG-21.

A STANDARD OF MANY PARTS
Part 1 is now published and a number of leading companies are working on making it the standard for future large image databanks. One of the most interesting features is that it can accommodate lossless compression as well as lossy compression in the same file, the lossy one being part of the total file. Simply cutting short transmission allows an application to provide a reduced quality or lower resolution image directly from a lossless archive. This can be seen as a possible replacement for previous data structures where a lossy image was kept separately from the reference lossless content, possibly stored as TIFF, PNG or other reference format (an original JPEG image for instance / ISO 10918-1, Information technology - Digital compression and coding of continuous-tone still images: Requirements and guidelines). It is particularly relevant in application fields where lossless is needed for downstream processing and archiving, whilst lossy transmission is required for fast viewing (medical, satellite imaging, remote sensing applications for instance). A simple file format is also included (JP2 files).

Part 2 adds the JPX file format, extending considerably the amount and capabilities of metadata associated with the codestream carrying the image data. This has been designed to comply with the XML standard and takes advantage of the results of many independent initiatives on descriptive metadata, such as DIG35, Dublin Core, EXIF and others. In particular, special attention was given to IPR, Intellectual Property Rights, as defined and addressed by the WIPO Treaty. Part 2 also adds many optional features, some of which are likely to require patent licensing (the JPEG committee is optimistic that agreement has or can been obtained from the 20-plus holders of many hundreds of patents in this area, that royalty free and license fee free licensing can be achieved for implementations of Part 1 of the standard).

Part 3, Motion JPEG 2000, splits a video signal into a sequence of stills (just as in the cinema) and allows for slow motion, and forward/backward running, with adaptive quality for a given bandwidth. Part 3 is linked closely to MPEG-4 through the ISO Base Media File Format, a new Part 12.

Part 4 has been promoted to an International Standard and is about compliance testing, and guidelines for checking the conformity of a decoder to the standard. It includes many examples of images taken with different parameters and using different colour spaces an area where JPEG 2000 is intended to compete effectively with proprietary formats and an area which the original JPEG standard really ignored.

Part 5 aims at making available to the public some reference software, a process that was one of the best boosters for the current JPEG, thanks to the work of the Independent JPEG Group. Now published, it links to two separate published source implementations, written in Java and C.

Part 6 defines a syntax to describe images where different types of content are put into a compound file, to optimize the code and take advantage of other compression initiatives already on the market. Parts of the imaged document can use JPEG's JBIG sister committee, algorithms or even existing code such as facsimile or original JPEG images.

Part 7 was initially about hardware implementation but has now been withdrawn.

Part 8 is one of the four new work items initiated in 2001, addressing security in terms of content, considering methods to encrypt, scramble, watermark contents to help establish trust in the future e-commerce marketplace. Conditional access, registration, identification and traffic monitoring are considered in this work item.

Part 9 deals with interactivity, especially over the Internet, delivery of metadata and the use of protocols to perform functions such as displaying different levels of quality or resolution according to the user's needs and available bandwidth, segmenting the codestream thanks to its inherent scalability and indexing methods that the part defines.

Part 10 is addressing 3D and volumetric imaging, including floating point data, allowing for extension of the image concept in medical and volumetric representations.

Part 11 is about wireless image transmission, for instance when an image projector is connected wirelessly to a laptop, but primarily to help robust transmission of images in the noisy environment of cellphone links. Adding control codes to the codestream can help considerably the handling of images when some codes are erroneous or missing.

Part 12, recently introduced, is the common work item with MPEG where a compliant file format will be used for both MPEG-4 and Motion JPEG 2000 images, to provide a richer environment for users of such technology.

As of today, Parts 1 to 5 have reached the level of International Standard, Part 6 is in final draft status (FDIS), Parts 8 to 11 are at various stages in the Working Draft process and work on Part 12 has just started.

Quite obviously, JPEG 2000 is not going to replace the current JPEG images for hundreds of millions of images already encoded and disseminated, but it will take over slowly as users realize the additional functionality that is offered. In addition to many strong points for applications ranging from secure archives (lossless encoding) to Motion JPEG 2000 for digital cinema (no restrictions in dimensions of images and the possibility to transmit images in multiple formats using part of the codestream) and medical images. The attention to detail in metadata handling and its integration with the physical image allows for content protection and access control, thus respecting the WIPO (World Intellectual Property Organization) Treaty by introducing security aspects within the file.

When the current JPEG was originally defined in 1988, it took five years before it became a successful standard, now holding more than 80% of all visual material on the Internet. It is foreseen that JPEG 2000 will find application in some areas (such as digital archiving) even faster, less than three years after the final International Standard qualification, particularly since some transcoding between JPEG 2000 and the original JPEG is feasible. This is confirmed by visits to the JPEG web site, featuring an average 10,000 hits per day in recent months! And the new standard Parts 8 to 12 exist to maintain a high level of development and activity for the JPEG Experts.




About this article
This article was first published in ISO Bulletin, Volume 34, No 2, February 2003, pp.17-19 and is reprinted with permission. Visit the International Organization for Standardization (ISO)

Icograda/ISO liaison:
Icograda holds liaison status with the following ISO Technical Committees (TC) and Sub-committees (SC): TC 37 (Terminology and other language resources); TC 145 (Graphical symbols); TC 145/SC 1 (Public information symbols); TC 145/SC 2 (Safety identification, signs, shapes, symbols and colours); and TC 154 (Processes, data elements and documents in commerce, industry and administration).

About the authors
Richard Clark is JPEG Editor and WebMaster at Elysium Ltd, United Kingdom.

Jean Barda is Technical Director at NETIMAGE and PR Chairman, ISO/JPEG.