| Abstract [eng] |
Occlusion culling is a method to determine which scene objects are not visible and discarding them. Because these objects are not visible they do not contribute to the final image and only waste computing resources. There are several methods to determine occluded objects, however all these methods use the same idea. Difference between these methods is in implementation – whether computation is performed on CPU or GPU, whether hierarchical trees are used, whether we compute and use results right now or use results from the previous frame, etc. This thesis proposes a new hybrid occlusion culling method, which combines both, CPU and GPU, and uses some information from the previous frame. Proposed method has been implemented and performance results were compared with some other methods. Obtained results are presented along with conclusion. |
