US Patent 7623137 Successive-convolution-compositing technique for rendering translucent surfaces

One embodiment of the present invention provides a system that uses a successive-convolution technique to render a set of visible sprites which are visible through a translucent surface. During operation, the system first initializes a translucent light map to a background color. Next, the system iteratively processes each visible sprite, starting at a furthest visible sprite from the translucent surface, and iterating through successively nearer visible sprites until a nearest visible sprite is processed. While doing so, the system processes each visible sprite by first compositing the visible sprite into the translucent light map, and then incrementally blurring the resultant translucent light map.