HDF/IRIM Exposure Maps

The HDF/IRIM data was reduced with the DIMSUM software package developed by Peter Eisenhardt, Mark Dickinson, Adam Stanford, and John Ward. A modification of the DIMSUM package enabled the reductions to use a simple impletmentation of the "drizzling" process known as interlacing. First, translational offsets between each of the individual 256x256 IRIM exposures of the HDF were measured. Each pixel in the image was then "dropped" onto a larger pixel grid sampled at a resolution four times finer than that of the original detector, i.e. a 4x4 subsampled grid. The pixel was added to the single "sub-pixel" at the appropriate position in this finer grid. After many images were interlaced in this fashion, the finer pixel grid gets filled more or less uniformly thanks to the very large number of independantly dithered exposures which were obtained over the course of ten nights of observing.

The resulting interlaced images are accompanied by "exposure maps" which indicate how many seconds of exposure time were accumulated in each of the sub-pixels on the finer grid. The HDF/IRIM images themselves have been divided by these exposure maps in order to produce data uniformly normalized to counts per second. The exposure maps reflect both the small-scale (pixel to pixel) variations in the expected variance at each position in the image due to the different number of independant dither positions which populated each pixel, as well as the overall "roll-off" in exposure time toward the edges of the image mosaics due to the finite extent of the dither pattern.

The HDF/IRIM exposure maps are similar to the "weighting maps" produced by the true "drizzle" software (e.g. for the HDF/WFPC2 data), but there are some differences. First, as the name indicates, the exposure maps are simply maps of the exposure time per pixel, while the drizzled "weighting maps" represent the expected inverse variance at each pixel position. In creating the HDF/WFPC2 drizzled weighting maps, several sources of noise, including the background level and the detector readout and dark noise, were taken into account. For the HDF/IRIM exposure maps, only the total exposure time per pixel was used. If the background level were constant at all times, the HDF/IRIM exposure maps would essentially be equivalent to the drizzled weighting maps modulo a uniform, multiplicative scaling factor, because the IR data is always background limited due to the bright infrared sky.

As discussed in the section on the HDF/IRIM world coordinate system , the version 1 release of the IR data has been geometrically transformed into registration with the WFPC2 mosaics . The same geometric transformation was applied to the HDF/IRIM exposure maps. The parameter "fluxconserve" in the IRAF "geotran" task was set to "no" for the transformation of the exposure map images in order to preserve the actual effective exposure time at each pixel position.

Note that the mean level of the exposure map images (averaged over any scale larger than 4x4) is 1/16th the total exposure time which was actually obtained in a given IR bandpass.


Return to the HDF/IRIM home page.

Copyright © 1997 The Association of Universities for Research in Astronomy, Inc. All Rights Reserved.


Mark Dickinson: med@stsci.edu 10/24/96