1994A&A...284.1037B

In this paper we present a new method for the construction of high-resolution images from the IRAS survey data base. The method, Pyramid Maximum Entropy (Pme), is a special application of multi-channel maximum entropy image construction techniques, using pyramid images in which each of the (K+1) image channels covers the same map area, but contains 2^kx2^k (k=0,1,...,K) pixels. The final result is a 2^K x 2^K-pixel image. Pme is a two-dimensional application of the MemSys5 software package. As Pme explicitly includes spatial correlations in the images, it is capable of reconstructing emission at virtually any relevant scale length. MemSys5 provides the capability to compute an error estimate of the MaxEnt result(s). Pme uses this feature to calculate an error map: a map of error estimates of the corresponding image pixel values. This is a powerful tool for determining the reliability of the results. It allows us to quantify point source fluxes as well as flux densities in diffuse areas, e.g., the disks of galaxies. We present images in all four survey wavelength bands (12, 25, 60, and 100 µm) of three nearby galaxies with different morphologies, to illustrate the quality of the Pme results. A summary of the principles of Bayesian image (re-)construction, on which Pme is based, and the mathematical foundation for incorporating spatial correlations in images are included in two appendices.