As the current graphics-accelerator hardware provides the capability of gen
erating 3D graphics images at interactive rates (30-60 frames/s), there is
a strong need for a good-quality shadowing algorithm that maps well to the
current graphics hardware architecture and is capable of producing shadowed
images at the same interactive rate. Our proposed shadowing method applies
the traditional, two-step Z-buffer approach, which takes full advantage of
the resources already provided in current graphics hardware. We employ a m
odified data structure on the light screen's Z-buffer, which dramatically i
ncreases the shadowing accuracy. This data structure, the ED-buffer, also s
implifies the shadow testing calculations considerably, needing only a coup
le of fixed-point multiplications and additions per shadow test. This is id
eal for hardware implementation. We also present a technique for optimizing
the use of the fixed memory resource in order to maximize the accuracy of
shadowing when the scene is projected from the lighting point of view on th
e six faces of a cube surrounding the light source.