|
Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.stsci.edu/stsci/meetings/irw/proceedings/reddingd.dir/section3_5.html
Дата изменения: Mon Apr 18 20:25:19 1994 Дата индексирования: Sun Dec 23 20:17:49 2007 Кодировка: Поисковые слова: п п п п п п п п п п п п п п п п |
Our work is continuing under two separate programs. We have a small amount of funding from ST ScI to provide a HST-specific point spread function generation code. This code will not perform prescription retrieval. It will be released (early in 1994) with the current best prescription data describing each of the WF/PC-1, FOC, WF/PC-2 and COSTAR/FOC cameras. It will be available as a stand-alone application, generating PSFs in an interactive session. These can be exported as FITS or binary format files for use in restoration. The code will be available in executable form for UNIX workstations. The code will also be available as a FORTRAN subroutine that can be called directly by a user code. For further information, contact Bob Hanisch at ST ScI (hanisch@stsci.edu).
We are also beginning a new task in collaboration with ST ScI, under the NASA Advanced Information Systems Research Program, to develop an integrated prescription-based image restoration application. The program will be written to make efficient use of a variety of new computer architectures, ranging from scalar workstations to one or more of the new massively parallel supercomputers. It will provide functions and features such as:
The application will exploit recent developments at JPL in parallel processing and distributed computing to provide a major increase in processing power, while retaining source-code portability, a user-friendly interface and the convenience of access through UNIX desk-top workstations. The user interface will provide a near-plain English command language and macro capability. Extensive graphical user interface features will be included.
The application will exploit an architecture-independent parallel programming environment that allows source code level portability for applications on various distributed memory architectures (Lee et al. 1990). This is achieved through an object-oriented data distribution function library, offering multiple distribution types. An application programmer can choose a proper distribution type according to the data access pattern of an application without considering the architectural constraints of a parallel system. We will exploit a message passing interface protocol (Dongarra et al. 1993) to achieve source-code level portability to various user environments so that a flexible mapping between the task and all available resources at the user site can be performed with one application program. Only a limited amount of machine-specific coding will be required to implement the application on a new architecture.
The code will be developed and released in phases, with a first (scalar) version available in about a year. When completed (in three years) the application will support distributed processing across heterogeneous computing environments, allowing efficient utilization of networked workstations and supercomputers.
In parallel with the code development, we will be performing extensive validation and verification of the various functions of the code. The performance limits of prescription retrieval with respect to noise, data diversity and other factors will be explored. Performance of prescription-based restoration will be compared to other methods. Working with images from the refurbished HST images, we will be pursuing ambitious objectives, such as recovery of the ``fuzz'' (underlying galaxies) of quasars, detecting faint companions of nearby stars, and extending the magnitude range over which stellar photometry can be done. Our ultimate objective is to make the very best quality restoration a matter of routine for all images from instruments such as HST.