We present a new algorithm, called Multiresolution Regularized Expectation
Maximization (MREM), for the reconstruction of gamma-ray intensity maps fro
m COMPTEL data. The algorithm is based on the iterative Richardson-Lucy sch
eme to which we added a wavelet thresholding step in order to eliminate ima
ge-noise in the reconstruction. The wavelet thresholding explicitly account
s for spatial correlations in the data, and adapts the angular resolution l
ocally, depending on the significance of the signal in the data.
We compare the performance of MREM to that of the maximum entropy and the R
ichardson-Lucy algorithms by means of Monte-Carlo simulations of COMPTEL 1.
809 MeV gamma-ray line observations. The simulations demonstrate that the m
aximum entropy and Richardson-Lucy algorithms provide virtually identical r
econstructions which are heavily disturbed by image noise. MREM largely sup
presses this noise in the reconstructions, showing only the significant str
uctures that are present in the data.
Application of MREM to COMPTEL 1.8 MeV gamma-ray line data results in a 1.8
09 MeV sky map that is much smoother than the maximum entropy or Richardson
-Lucy reconstructions presented previously. The essential features of this
map are (1) an asymmetric galactic ridge emission reaching from l approxima
te to 45 degrees to l approximate to 240 degrees, (2) a bright localised em
ission feature in the Cygnus region around (l, b) approximate to (80 degree
s, 0 degrees), (3) two emission spots at l = 317 degrees and l = 332 degree
s situated in the galactic plane, and (4) an extended emission region aroun
d (l, b) approximate to (160 degrees, 0 degrees). Comparison of the MREM ma
p to the simulated reconstructions demonstrates that the 1.809 MeV emission
is confined to the galactic plane.