A new approach to designing spatial filters for the detection of point sources is presented. The technique has been developed for the detection of missiles in infrared images. A missile is generally assumed to be a point source and the image of a point source is called the point spread function (PSF). Exploiting the fact that the PSF of many systems exhibits a circular symmetry, we propose a filter implemented in two steps. First, we apply a circularly symmetrical bandpass filter to the image. Unfortunately, this filter retains the edges of extended objects. To overcome that problem, we propose, in a second stage of processing, different techniques to attenuate those edges without affecting the point sources amplitude by noting that most of the edges appear in the magnitude of the Fourier transform of the image as a concentration of energy in spatial frequencies distributed along straight lines at different angles, whereas the Fourier transform of point sources is circularly symmetrical. We show some examples where the resulting filter attenuates significantly the extended targets while the amplitude of the point sources remains the same.
