| A study of combined compression and analysis (CCA) is presented, dealing specifically with theory and design issues associated with systems that both compress and analyze astronomical image data. Information theory, somewhat broadly interpreted, is presented as the appropriate theoretical context for a study of CCA systems. Some information-theoretic results are summarized that are relevant to CCA systems. A practical attributes-rate theory is proposed as a means of characterizing and comparing CCA systems. This theory is applied to eight different CCA systems designed to compress and extract point sources from infrared astronomical images. An image simulation involving a tractable source list model based on the Pareto distribution is presented. The tractable model is analyzed using information theory and mathematical analysis. Theoretical results include the derivation of equations for the differential entropy of the Pareto distribution and for the L1- and L2-norm image distortions in terms of extracted point source position and brightness errors.;A new data compression algorithm for encoding astronomical source lists is described. Two experiments are described, the first using simulated imagery based upon the tractable model, and the second using images from SPIRIT III, a spaceborne infrared sensor. The results are presented in terms of image distortion, detection performance, and estimation performance. The variation of performance with bit rate is shown for all performance measures. The variation of detection and estimation performance with point source flux is also shown. Algorithm SZ, the CCA system consisting of the source list compressor followed by a zerotree-wavelet residual encoding, was consistently a top performer in terms of all performance measures computed. While all of the compression algorithms showed an ability to reduce high-frequency quantum noise at certain bit rates, no evidence was found that such denoising brought about any improvement in point source detection or estimation performance. |
