alien@essex.ac.uk ("Adrian F Clark") (05/23/91)
The following is an article I wrote for the next edition of the IAPR
Newsletter on current standardisation work in the image processing
area.
An International Standard for Image Processing
and Interchange
Adrian F Clark
University of Essex, UK
alien@uk.ac.essex
How many readers have been given a tapeful of images for some
important work, only to find that no information has been supplied
about the tape format? In such cases, the effort involved in ensuring
that the tape has been read correctly and converted into one's own
file format is usually at least as great as the effort involved in
processing it. Equally irritating, how many times have you received a
program from someone, only to find it uses an image processing library
which you do not have? You might be surprised -- and hopefuly pleased
-- to learn that many of these problems could disappear in the near
future due to work on an international standard for Image Processing
and Interchange (IPI).
Initial work has been proceeding for over two years now, but the
project really started in earnest at a recent International Standards
Organisation (ISO) meeting in Norwich, UK. The standardisation effort
is to take place under the aegis of ISO SC24 (`Computer Graphics and
Computer Imaging') and will be the first standard in this area to
cover both an Application Programmer's Interface (API) and an Image
Interchange Format (IIF). Both the API and the IIF will be based on a
Common Imaging Architecture (CIA). The time is not yet opportune to
discuss the content of the standard in detail, but an outline of some
aspects is appropriate. (However, the reader should be aware that
even these may change as the standard develops.)
Part I, the CIA, will describe a general model for the processing of
imagery (an `architectural' model), including aspects such as the
effects of regions in both source and destination images. The CIA
will also define models for image data-types and several non-image
data-types such as histograms. It will also indicate the
`specialisations' necessary to tie down the other parts of the
standard. For example, an image will be presented as an N-dimensional
object, but the current API will only support images of up to five
dimensions (three spatial, time and spectral).
Part II, the API, is based on the Programmer's Imaging Kernel (PIK)
being developed in the USA. This API will support most common
operations on images, of both the image-to-image (e.g., FFT) and the
image-to-non-image (e.g., histogram) varieties. Some facilities for
manipulating non-image data-types will also be included. Both image
and non-image data-types may be based on several basic representations
-- integer, floating-point, complex, etc. The API is designed to
support operations on pixel data, and explicitly excludes `image
understanding' operations, which typically involve non-pixel
representations.
There are likely to be several levels of conformance to the standard,
and the API in particular, but the mechanism for achieving this has
not yet been discussed. It is worth noting that a number of PIK and
PIK-like image processing packages are now available; these give a
reasonable idea of the functionality that the API will offer, though
the packages themselves will probably require significant modification
to conform to the standard.
The final part of the standard will describe the IIF. There are many
widely-used image formats already in existance, some of which are de
facto standards in particular applications areas (e.g., FITS in the
astronomical community) or are already standardised (there is an ANSI
image tape format standard for electronic pre-press). Of course,
recent standards from JPEG and MPEG, among others, involve some
compression, and the IIF is likely to incorporate their
recommendations.
At the abovementioned Norwich meeting, editors for the various parts
of the standard were appointed:
Adrian F. Clark, University of Essex, UK
William K. Pratt, Sun Microsystems, USA
Rainer Hofmann, Fraunhofer Group, Germany
and a timetable for the development of the standard was laid out.
Most of the standard will be written, distributed for comment, and
edited in 1991; trial implementations and further editiing will take
place during 1992-3. The final International Standard is timetabled
for early 1994. (Be warned, however, that such timetables almost
invariably slip somewhat.)
What will the impact of the standard be? Its effect will obviusly be
greater in some areas of interest to the IAPR than others: for
example, greater functionality is likely to be available to support
image restoration and reconstruction than computational vision, simply
because the former includes a larger proportion of pixel manipulation.
However, be assured of one thing: the standard will affect you,
particularly as add-on hardware which interfaces to the programmer via
the API becomes available. Hence, it is important to have as wide a
consideration of the drafts of the standard as possible. If you would
like to be involved, contact your national standards body or, if they
are unable to help you, the author of this article.
--
Dr Adrian F. Clark JANET: alien@uk.ac.essex
INTERNET: alien%uk.ac.essex@nsfnet-relay.ac.uk FAX: (+44) 206-872900
BITNET: alien%uk.ac.essex@ac.uk PHONE: (+44) 206-872432 (direct)
Dept ESE, University of Essex, Wivenhoe Park, Colchester, Essex, C04 3SQ, UK.