One- and Two-Dimensional Processing in Line Scanning Systems

Abstract

The performance of line scanning systems is analyzed in this paper. The results apply to systems such as infrared scanners, TV systems, and radar systems. The main objective of the analysis is to compare the performance obtainable when optimum two-dimensional processing of the data is used and the performance obtainable when only one-dimensional (along the lines) processing is used. Optimization under the constraint that only processing along the lines is available is a new and interesting problem in optimum filter theory, and the results are of practical importance since in most systems this constraint is imposed. Since it is possible to modify the systems to accommodate arbitrary two-dimensional processing at the expense of some system complexity, it is important to determine the corresponding performance improvement. The emphasis is on the general theory, and many special cases deserve quantitative treatment. In this paper quantitative comparisons are made for two cases; specifically, top hat and cone-shaped frequency responses of the aperture are considered. These characteristics occur in coherent and incoherent scanners in which the “resolution” is limited by the optical aperture of the scanner. Three performance criteria are used: least-squares error, maximum-peak signal to rms noise ratio, and minimum system resolution. Table I shows the improvement of arbitrary two-dimensional filtering over one-dimensional filtering along the lines.