The issues raised by a number of projects were reported to this group.
Uniform Access to (Remote) Image Data
Standards for imaging and communication have been developed separately and in isolation from each other. Moreover, even the standards for imaging have been developed in an uncoordinated manner, each addressing the needs of a particular application area. Terminology has been developed in the different areas such that the same terms are used to mean different things.
The RACE II project AMICS (Advanced Multimedia and Image Communication Services) arose from the recognition of the need for a framework, that encompasses the characterization of images, identifies the operations on them, and relates the operations to the required standards.
The overall aim of the project was:
Four specific subgoals were identified that needed to be achieved to fulfil the overall aim:
- To establish an image communication platform for advanced multimedia services.
The project was completed in March 1994, having produced a design for the ICOA and software tools in support of it and demonstrated them as well as having a chip in fabrication.
- To design a framework for image communication, the Image Communication Open Architecture (ICOA) (related to standardization activities);
- To develop software tools for the ICOA;
- To consider hardware support for the ICOA;
- To demonstrate the ICOA.
The ICOA addresses the broad area of imaging and communication and fulfils the image communication requirements of a wide range of application areas including multimedia. Access to different file formats is encompassed by the ICOA. A means of describing the abstract storage order of image data was developed to allow a common means of addressing parts of an image independently of the file format.
An important requirement related to communication was that of providing uniform access to images whether they are stored locally or remotely. By providing such uniform access, communication services can be invoked as necessary, without the knowledge of the application, and other related services such as compression and conversion can be integrated. Such uniform access is provided through the ICOA Image Handling Interface(IHI). Within this context, access is deemed to cover not only access to existing images but also access to the means to create new images. Hence, the IHI provides for the storage and retrieval of image data, the creation and deletion of images, and the setting and inquiry of image attributes. Within the AMICS project, the IHI was realized by means of the ICOA Image Handler which took the form of an Open Distributed Processing (ODP) object.
The IHI and the Image Handler were demonstrated in the context of a multimedia remote teaching scenario. The concepts were also shown in providing access to subsets of very large remotely sensed images.
Delivering an Image Archive
The following issues have been raised by the Bristol ETS/JISC project to deliver an image archive (including a digital version of the Bristol Biomedical Video disk):
In addition to the above points, we feel that the following are also relevant in the context of multimedia :
- Standardisation of file formats across platforms, MAC, PC, PC windows, UNIX/X windows
- Image format conversion (We are developing batch processing tools for un-attended image format conversion)
- Image formats for X-ray images
- Reliability of compression for Medical images and any others with high resolution (high frequency) detail (We are currently funded to carry out an evaluation of different image capture cards for this purpose)
- Standardisation of images sizes (x,y resolution and colour depth)
- Use of standard/common palettes across platforms (including reserved colours for the window systems in use)
- Use of thumbnail images and recommended sizes thereoff
- The necessity to look at increasing display complexity and colour resolution
- The 50 colour limit on in-line images in Mosaic
- Use of batch processing programs for archive development
We have investigated some areas of the palette problems, and have tested 60x pixel thumbnails for net delivery. We are currently using lossless compression in GIF file format, at 320x~244 (preserving aspect ratio) in 256 colours.
We are archiving at the highest resolution currently available to us so that we can increase the resolution on all our delivery images over a period of time to keep up with advances in technology.
The Virtual Teaching Collection
The Virtual Teaching Collection (VTC) project is funded under the HEFCs' TLTP initiative. This is based at Cambridge University Museum of Archaeology and Anthropology and the Whipple Museum of the History of Science. The aim of the VTC project is to permit the delivery in digital form of rich visual presentations of objects normally distributed in separate museum collections, thus providing an important teaching a research resource.
The project has had to consider different image compression methods. Issues which have been addressed include whether the compression technique used should be lossy or lossless. Lossy can be fine if the end result is viewing by humans but not if there is machine processing of the image.
The factors affecting the choice include the retention of relevant information and the resulting filesize. The VTC group have considered the merits of JPEG and fractal compression. JPEG has a symmetric compression and decompression, fractal has asymmetric with decoding being quicker. There is a problem with fractal compression being proprietary and available on limited platforms. The project are considering using IFS compression which is a type of fractal compression but one which is unlikely to be made public. It does offer high quality fractal compression software.
Image Formats and Material Science
Kate Crennell discussed the use of teaching materials in the area of material science. This relates to a TLTP project which is being conducted by 18 universities in liaison with the Institute of Materials. It is a cross platform project. Kate presented her work which is based on the Acorn platform which she feels is a good platform for introductory materials as it is available in many schools. A prototype database of images is being worked on for schools. This is to be available on CD-ROM using Acorn computers which allow use of multimedia presentation. Common image types which need to be considered include:
We need to recommend to people which formats they should use for images as there are so may around. Kate noted that Acorn have an agreement with Kodak for use of Photo-CD.
- bitonal document
- static grey scale
- static three colour
- static multiband
- static volumetric
- timed sequences
Kate concluded the following:
- Although networks and the World Wide Web are useful places for finding information, she considers them to be virtually impossible tools for new users. Even experienced users need to know the time of day when the host server can give a decent response. The HE community needs a short tutorial, one A4 page, giving enough details to find the multimedia help they seek.
- Networks are places for archival storage of source materials for educational software, they will always be too expensive, or slow for use by small colleges of further education, whose needs would be better served by improved access to courses on CD-ROM.
- Mainframe computers and large workstations are not the most effective tools for teaching undergraduates. Microcomputers are much better, but there are problems in putting all our eggs in one basket of proprietary software, using PC clones and Microsoft products. Kate suggested that the report should recommend 3 micros, PC, Mac and Acorn. A comparative review of Acorn micros, PCs, and Macs shows that they are much better value for money (see reference below).
- Any list of archival image databases should include ones for particular subject areas, medicine, astronomy, art history, as well as more general ones, such as 'Peipa' (described below)
- Although Kodak have not release details of their 'Photo-CD' format, there is no denying that this will be the simplest way for many people to get digitised images, no need for fancy scanners or digitisers, just take 35mm film in to your local holiday snaps processor (see reference by Thompson below). Then select the resolution you want, and store it in your image archive. The HE community cannot afford to ignore the format 'Photo-CD'.
- The report should include a list of available methods of digitising images, hand held scanners, A4 flat scanners, capture from video recordings, etc. For the people who cannot get the images they need from an archive.
- Kate recently came across something called 'TWAIN' which appears to be the name given to a standard range of scanner drivers, making it much simpler for people to interface scanners if the software has a TWAIN interface.
- Image format converters are available for the PC as well as UNIX. Kate uses GWS, which she got from RAL PC Support. It's PD and should be included in any list of conversion utilities published in the report. Such utilities are often provided with scanner software, and it would be nice to know which formats are standard input/output for which popular PC packages, like 'Corel Draw' for example.
- Peipa. The Pilot European Image processing Archive IAPR Newsletter Volume 16 No.3 Page 5, Adrian F.Clark, (email: email@example.com) see next page
- The Risc PC vs Macs and PCs, Acorn User issue 142, May 1994 p 38, 39
- Back in the picture, Acorn User, Colin Thompson, May 1994 P52-53
The Pilot European Image Processing Archive
A common practical problem in pattern recognition, and particularly those areas that are concerned with processing image data, is that of reproducing other researchers' results: either the data are difficult to obtain, or one must implement their techniques in order to make a comparison with one's own. The vast majority of researchers are, of course, quite happy to provide copies of software and data upon request; but servicing large numbers of such requests quickly becomes tedious and time-consuming. It is thus attractive to contemplate the idea of an archive, to which people may up-load their software or data for dissemination to other researchers. Indeed, with a central repository for storing software and data, time is saved in both distributing one's own material and acquiring that of others. Most important, the easily availability of such software encourages comparisons with existing techniques on accepted datasets.
An archive of precisely this type is running on a pilot basis. It is known as PEIPA, the Pilot European Image Processing Archive, physically located at the University of Essex in the UK and connected to the outside world by a (shared) 2Mbit/sec link. The host machine is a Sun Sparcstation ELC, kindly provided by the British Machine Vision Association for this purpose, and uses disk space from a number of research grants and contracts. Additional disk space will soon be in place, funded by DG III of the European Commission.
The archive is operated by the author, and is associated with Technical Committee 5 (Benchmarking and Software) of the IAPR. It aims to provide a convenient mechanism for exchanging software relevant to all areas of pattern recognition and image processing, and to make available datasets for testing and evaluating algorithms --- the benchmarking of techniques rather than systems.
What's in the Archive
Every item in the archive is available free of charge for research use. The principal contents of the archive are software tools, mostly for processing image data (though see below).
These include popular packages such as:
It is a matter of policy that source code be normally available for software packages. The vast majority of software is written in C and runs on Unix-based workstations. Software for other platforms, such as PC and Macintosh, is welcomed.
- Khoros, a visualization package with signal- and image-processing facilities, originally from the University of New Mexico;
- LaboImage, from the Vision group at the University of Geneva;
- pbmplus and netpbm, for inter-converting between image formats;
- a collection of Hough transform programs;
- neural network and genetic algorithm software;
- image synthesis tools;
- implementations of codecs for the JPEG, MPEG, and H.261 standards.
The archive also provides support information, such as descriptions of popular image file formats, back-issues of the VisionList and Pixel mailing lists, and electronic literature surveys and bibliographies, the latter stored in BibTeX format.
The area of the archive that is growing most quickly is that devoted to datasets. The existing entries include laser rangefinder data, calibrated stereo and motion sequences. A large database of face images will be made available as soon as disk space limitations permit.
Areas of the archive that are comparatively new include a European distribution point for documents associated with the 'Image Understanding Environment', a DARPA-originated but now international programme for developing a common vision environment, technical reports, and summaries of the activities of research groups.
Accessing the Archive
You may access the archive electronically by anonymous FTP, or using the Internet gopher, or via the World-Wide Web (WWW).
FTP to host peipa.essex.ac.uk, log in as 'anonymous' and give your email address as the password. The archive is in directory 'ipa'.
Direct gopher to host peipa.essex.ac.uk and select the 'peipa' menu\item.
The URL of the main page concerning the archive is
This page also contains pointers to other information of interest on the Web.
The numeric address of peipa.essex.ac.uk is 22.214.171.124.
How You Can Help
An archive is only as good as its contents. The archive currently concentrates on the processing of image data, since that is the author's research field. It is hoped to expand the archive by incorporating software intended for other data types or for general pattern recognition, and to make available other types of data (e.g., speech). If you would like to contribute to the archive, please contact the author at the following address:
Dr Adrian F Clark
Dept. Electronic Systems Engineering
University of Essex
Tel: +44 206 872432
Fax: +44 206 872900
If you would be interested in coordinating a particular subject area of the archive, please let the author know.
Virtual Environments Visualisation