This paper will describe a training system for the cost-effective, real time simulation of periscope visuals using Raster Graphic, Computer Image Generation Techniques. The design is optimised to present high definition target images against a realistic background, with emphasis on sufficient detail and realism to allow periscope observer training in target detection, observation, and classification. A channelized architecture is employed in which target data bases are separately processed to form individual target images. Dynamic background images are generated by a background channel. Unlike conventional approaches, targets and background do not form part of an overall data base; outputs from the channels are mixed together on a priority basis in real time. Target detail is thus maintained independently of overall scene complexity. Smooth edges and motion are sustained by incorporating sub-pixel area antialiasing throughout.