New for: D4
The first is due to the huge memory consumption which significantly slows
down the hardware-accelerated rendering process if the size of available
texture memory is limited. In these cases it is useful to compress the
light field, i.e. to use block encoding or vector quantization. For vector
quantization a hardware-accelerated alogrithm is presented which directly
renders the compressed light field without the need of previous
decompression.
A second shortcoming is that light fields can only
represent static scenes. If something in the scene is changed the whole
light field has to be recomputed. In order to overcome this geometry light
fields are introduced which decouple the goemetry of the represented
object and the illumination from the surrounding environment in such a way
that both may be changed independently.