Aliasing in computer generated images produces artifacts which degrade training effectiveness. A rigorous implementation of low-pass filtering required to prevent aliasing requires computation too extensive to be incorporated into real-time computer image generation (CIG) systems. As a result, current CIG systems employ poor approximations to proper filtering, and aliasing still occurs.

This paper discusses the theory of image filtering and demonstrate a new real-time antialiazing technique developed from the theory. The new technique represents a much closer approximation to the rigorous solution and therefore produces images of much higher quality than current real-time techniques. At the same time it requires less computation.