Converging Computing Methodologies in Astronomy: Network Activities
- From vision models to image information retrieval.
Methods such as wavelets and multiresolution approaches, mathematical
morphology, and fuzzy methods have proven their worth in the framework of
accessing appropriate information from large image databases. Use of such
methods in this context is not a purely engineering task (i.e.\
them as stand-alone commands in an information retrieval framework).
Rather, the methods must be moulded together to allow semantically-driven
access to data. New methods, motivated by those mentioned, are required to
handle the huge quantities of data which stand ready to be analyzed.
- The data life-cycle -- methodological aspects.
The astronomical data life-cycle is highly digital: data capture is
increasingly on CCD electronic detectors, data are subject to image
processing and statistical treatment, and the final major stage in this
process involves (a) data archiving, and (b) publication. Not surprisingly,
the issues of electronic publishing and of digital libraries are viewed as
increasingly important. Data storage and access standards, together with
interface standards, must be coordinated. Data characteristics, including
dynamic aspects related to the life-cycle, strongly impact on methods
for analysis and treatment.
- From data integration to information integration.
- Particular data integration (data fusion) problems, such as integration
of data associated with different wavelength ranges, are of great
relevance in the context of large space- and ground-based observing
projects. In certain instances, this takes the form of co-addition in
image restoration. It also includes the enhancement of image restoration
and filtering approaches through building-in semantic information on the
cosmic objects of interest. Close, complementary use of multi-million
astronomical catalogs still has some distance to go in order to be tightly
coupled with analysis of images from large image databases.
Complementarity of such data rests on external information (semantics),
and determines the appropriate analysis methodologies. Large-scale
data integration prototypes have been pursued by NASA and ESA in recent
years (respectively, the Astrophysical Data System, ADS; and the European
Space Information System, ESIS), and these serve to point to 'issues
and concerns' which are solvable through improved methods.
- Classification of terabyte data collections, which includes neural
networks, decision/classification trees, and machine learning approaches.
Widely reused code and paradigms has not yet evolved out of extensive
astronomical work done in this area. Through coordination, this could be
- Long-term access to stored data, and the selection of the latter. What
should be, in the phrasing of M.J. Kurtz, 'the future of memory' [i.e.
humanity's memory, which includes image data and other data]? Ever
greater quantities of data are collected -- what should be (expensively)
stored, and how should this be done? Photographic plate directories
exist, and policies based on a range of theoretical issues are required
for scanning and maintenance. Major projects collect many terabytes of
data, and reuse is both necessary but also problematic.
- Beyond data, astronomy is all about information. Compression is a
byproduct of image models. Loss of information involved in compression
raises a range of questions. In its broadest sense, compression is
summarization, and therefore is part of the overall process of scientific
Last update: 2.12.94.
Comments on this screen to firstname.lastname@example.org