The principles of communication theory were applied in the 1950's to optical imaging systems and to the analysis of images. Optical systems were analyzed in much the same way as linear systems (modulation transfer functions and channel capacities) and images were characterized in ways analogous to time signals (space-bandwidth products, spatial frequency content, etc.). Both coherently and incoherently illuminated optical systems can be treated using these concepts. Coherently illuminated systems are most useful for performing operation such as convolution, cross correlation, and spectral analysis because the Fourier transform of an optical signal physically exists and can, therefore, be measured or modified. The basic Fourier transform relationship for coherently illuminated systems is developed in this paper. It can be detected directly and used to estimate the distribution of spatial frequencies contained in the signal. Methods for constructing complex-valued spatial filters are described; these filters can be used to realize such operations as convolution or cross correlation, addition or subtraction, and differentiation or integration. Experimental results are given to illustrate the concepts and to susgest potential applications. To extend the range of applications, interface devices are needed to allow optical processing of two-dimensional raster-scanned time signals, wide bandwidth electrical signals, and incoherent optical signals. Interface devices are often needed to convert the output optical signal to an electrical signal for post-processing by a digital computer. For some applications, interface devices are needed to construct spatial filters in real time, so different operations can be performed on a given signal. The desired characteristics of these three interface devices and the current state of their development are briefly reviewed.