Implementation of an incoherent optical image restoration method: limitations related to optical subtraction

Incoherent optical information processing is known to be far less sensitive to noise than coherent processing. However, it is restricted to real non-negative quantities. In cases like approximate deconvolution for image restoration, where positive and negative quantities are involved, the use of incoherent illumination is possible only at the expense of some additional sophistication needed to incorporate the negative parts; in particular, the restoration can be achieved by combining two incoherent convolutions and one subtraction. This paper describes an experiment of this type, with special emphasis on the practical implementation and on the difficulties related to the subtraction. Alignment and normalization of the two images to be subtracted are critical issues. If the subtraction is performed optically, the dynamic range of this operation is an additional problem.