Inevitably high definition colour optical images consume considerable disc capacity so that it becomes a matter of importance to define exactly what resolution standard is needed. This avoids unnecessary expenditure on monitors, lenses, and storage capacity for resolution which is not really required. The choice of loss-less compression as well as consideration of lossy compression in aspects of the overall application are matters which can be determined only in detailed discussion on archival usage and customer need. Images stored on such WORMS are not simply duplicating those held on the original carrier they have major advantages over the originals. Most significant is the total safeguarding of colour which is in a state of continuous deterioration in any analogue and/or original chemical form. Significant too is the consummate ease with which copies can be cloned in seconds against the considerable skill and time necessary to produce copies from original photographs negatives. Handling too is immensely facilitated which is particularly significant in high speed search for exactly the image which fits a particular need. The human eye is adept at high speed image identification: no conventional album or slide store can compare with disc image scanning if judged on speed, cost and convenience.
or
The state of the art in retrieval is particularly relevant to HKGIS needs. Though image retrieval by image element input is already in the experimental stage its practical use in an application such as that here under consideration is not likely for at least 3 or 4 years there is still a considerable amount of development needed. What is, however, available is retrieval through neural databases. Second to copying on to disc in cost of labour is probably indexing or cataloguing to ensure effective and rapid retrieval. Whilst the existing indexing and finding aids would be brought into the new system new entry material would, preferably, be subject to different processing. Probably this would take the form of a brief - though fully adequate in length - written description of the content of each photograph. This would be fed into a neural database and would thereafter provide retrieval of all such entered material on whatever basis were needed. Neural databases can be extremely convenient, fast and, through their ability to learn continuously what the user really has in his mind, ever more accurate. They do
do not therefore supersede retrieval by number, date, place and event etc, but add to these factors limitless possibilities of retrieval by association, image treatment, detailed subject content, juxta position, relationship to moving image coverage etc, etc.
Neural databases are now being seriously investigated by the great newsagencies and undoubtedly represent the shape of thing to come. In terms of cost they are extraordinarily attractive and certainly in the case of the HKGIS juxta position of photographic and cuttings archives allow separate
linked retrieval of material instantly as needed for press, television, publishing etc. The fact that many of the cuttings are in Chinese poses no problem: the system could be set up to cope indifferently with English or Chinese either separately or linked. Finally neural databases score over software for free text search and other comparable systems in their distinctive