Correcting Inter-Waveband Positional Errors in Sloan Images

The telescope of the Sloan Digital Sky Survey (SDSS [1]) was designed from scratch to minimize "down-time": the telescope does not stop to image a portion of the sky or change filters: instead, images in all 5 wavebands are captured continuously as the telescope moves along a Great Circle on the sky, with all five cameras--one for each waveband--capturing the same area of sky in sequence about 1 minute apart as that region drifts across the camera's field. This is called "Drift Scan Mode", and although it maximizes the light captured per unit time, it means that image pixels extracted from the SDSS database do not correspond to physical pixels on the camera, but were instead constructed by a software pipeline that "tracks" the sky as it drifts across camera pixels. Unfortunately, this pipeline makes no attempt to ensure that image pixels across wavebands correspond to the same patch of sky. This confounds attempts at cross-waveband comparisons of the same object, since there is no 1-to-1 correspondence
between inter-waveband pixels, and location on the sky.

To solve this problem and reconstruct images for all wavebands that are better suited to precise inter-waveband astrometry, we use the positions of stars visible in any two wavebands to estimate the inter-waveband positional shift. We then use the FITS file to transfer flux between adjacent pixels to construct new images in which the "same" pixel corresponds to the same patch of sky across all wavebands. In this abstract we describe the process used to achieve this and quantify the errors in flux and position introduced by our method. The attached image shows an example of an original SDSS image containing a galaxy, the same image shifted half a pixel (which seems to produce worst-case errors), and the residual after a second shift back to the original position. Note that the largest errors occur near stars, whereas the background pixels, and the galaxy, are virtually invisible in the residual image.

Theme – Data Processing Pipelines and Science-Ready Data