
294
7 Restoration
7.5 CONSTRAINED DECONVOLUTION
In the Wiener or the least squares restoration filter derived in the preceding
section, the following basic assumptions were made: that the undegraded
picture and the noise belonged to homogeneous random fields and that their
power spectra were known. In many situations one may not have a priori
knowledge to this extent.
If all that is known (aside from the nature of the degradation) is the variance
of the noise, then the method to be described in this section is applicable.
This method was first formulated by Phillips [78] and later refined by Twomey
[102,103] in the one-dimensional case. Th