Maria A. Nieto-Santisteban (STScI), Joel D. Offenberg (Raytheon/STX and STScI), Dale Fixsen (Raytheon/STX and STScI), Robert J. Hanisch (STScI), H.S. (Peter) Stockman (STScI), Scott Stallcup (STScI)
Reference URL: http://ngst.gsfc.nasa.gov/cgi-bin/iptsprodpage?Id=14
The Next Generation Space Telescope (NGST) will have a huge detector array (at least 8k x 8k pixels, with typically 64 non-destructive read-outs per image) and will be placed in a far orbit from Earth (such as the L2 Lagrange point). The distant orbit limits the data downlink bandwidth and requires some amount of on-board image processing and data compression.
STScI is participating in a study supported by NASA's Remote Exploration and Experimentation (REE) project aimed at determining the feasibility of using a high performance parallel computer on board the NGST. Such a computer would be composed of inexpensive, off-the-shelf processors. In order to avoid the expense of using radiation-hardened components, the processors would be configured for redundancy and applications software will need to be designed to be fault tolerant. As part of STScI's study, we are evaluating how such a computer could be used to do real-time cosmic ray detection and removal so that we can combine the 64 read-outs into one image with minimal loss of information. We are also evaluating various methods for data compression so that the resulting image can be transmitted to the ground efficiently.
This paper gives a summary of the performance (both in terms of compression ratios and computational time required) of several lossless compression methods on different types of images. We also show results of prescaling the image prior to compression using a square-root function. This imposes a slightly lossy compression, but the scaling can be adjusted so as to retain the desired number of noise bits.