PIXON-BASED DECONVOLUTION

We describe a new technique, pixon-based deconvolution, for the recons truction of images and spectra from low signal-to-noise data. In ''tra ditional'' techniques, such as Maximum Likelihood (ML) and Maximum Ent ropy (ME), the model parameters (e.g., pixel size) are considered ''nu isance parameters'', and generally held fixed in the course of deconvo lution. Pixon techniques allow the model parameters to change accordin g to the information contained in the data. The pixon model admits onl y that level of detail which is statistically justified by the data, t hus greatly reducing the production of spurious sources and signal cor related residuals, common problems in ML and ME reconstructions. As a result, gains in sensitivity and spatial/energy resolution may be real ized. Sample applications to data from OSSE and COMPTEL will be presen ted.