Thanks to recent algorithm improvements and new trends in computer
architecture researchers were able to show that ray tracing can be a
viable alternative to commonly used Z-buffer rendering algorithms,
especially when one tries to achieve visual realism. Until now, ray
tracing applications were restricted to static scenery, since
interactive performance could only be achieved by the use of a
precomputed acceleration structure (e.g. kd-tree). The large building
time for these hierarchies implied that the scalability of the
algorithm was lost. We try to address this problem by investigating
another construction strategy known as "lazy build". By efficiently
finding and separating irrelevant parts of the scene, we will be able
to improve rendering performance and make the algorithm scalable.