We examine cosmic ray rejection methodology on data collected from InSb and Si:As detectors. The application of an Up-the-Ramp sampling technique with cosmic ray identification and mitigation is the focus of this study. This technique is valuable for space-based observatories which are exposed to high-radiation environments. We validate the Up-the-Ramp approach on radiation-test data sets with InSb and Si:As detectors which were generated for SIRTF. The Up-the-Ramp sampling method studied in this paper is over 99.9% effective at removing cosmic rays and preserves the structure and photometric quality of the image to well within the measurement error.

We present an algorithm to optimally process uniformly sampled array image data obtained with a nondestructive readout. The algorithm discards full wells, removes cosmic ray (particle) hits and other glitches, and makes a nearly optimum estimate of the signal on each pixel. The algorithm also compresses the data. The computer requirements are modest, and the results are robust. The results are shown and compared to results of Fowler sampled and processed data. Non-ideal detector performance may require some additional code, but this is not expected to cost much processing time. Known types of detector faults are addressed.

We present a data simulation package designed to create a series of simulated data samples for a detector with non-destructive sampling capability. The original intent of this software was to provide a method for generating simulated images for the Next Generation Space Telescope, but it is general enough to apply to almost any non-destructive detector or instrument. MultiDataSim can be used to generate "practice" sampling strategies for an instrument or field-of-view, and thus can be used to identify optimal observation strategies.

The NASA Yardstick design for the Integrated Science Instrument Module (ISIM) utilizes 64 1k x 1k InSb imagers. Since multiple readouts per image are required to remove cosmic ray events and to reduce the readout noise, the implied downlink rates can approach 0.7 Tbytes per day. These would exceed the downlink capabilities of RF links from L2 or more distant orbits. Some form of on-board data analysis and compression is mandatory: a factor of 100 compression for an L2 orbit, factors of 1000-10000 for a 1 AU drift-away or 1 x 3 AU elliptical orbit. Supported by funding from JPL to study the uses of low-cost, on- board parallel processors, we are studying algorithms and architectures to perform cosmic-ray rejection and lossless compression of the NIR images, while optimizing the resulting signal- to-noise.