This paper shows how an effective noise suppression strategy can be in
corporated into image restoration algorithms. The widely-used Richards
on-Lucy method is used here. The residual in the fit of the restored i
mage, at each iteration, is analysed using a wavelet transform. In ord
er to suppress noise, significant wavelet coefficients, only, are reta
ined. The effectiveness of this procedure is illustrated with a range
of examples. Limitations of this approach, and pointers to work in pro
gress, are also outlined.