Analysis of coherence strategies for stereo occlusion culling

This paper explores the benefits that can be achieved for stereoscopic visualization when using occlusion culling strategies. Graphics processing units (GPUs) are improving their computational power and functionalities. On the other hand, models are also increasing their size and users require more demanding image quality. Occlusion culling provides significant frame rate speed-ups in densely occluded scenarios. This paper shows the limitations placed by compact and not densely occluded scenarios and the gains that can be achieved when rendering stereo images using occlusion culling. The experiments carried out test combinations of occlusion stereo coherence and occlusion frame coherence. The methods also take advantage of efficient depth sorting strategy and aggressive occlusion frame coherence, whose image artefacts have been found to be negligible. Results clearly point out that eye-independent frame coherence offers more benefits than mixing frame and stereo coherence. As in monoscopic occlusion culling, using simplified geometric models, instead of bounding boxes, when posting occlusion queries, improves significantly the number of objects discriminated as occluded. The algorithm presented can be easily implemented and provides a significant boost in performances. Copyright (C) 2007 John Wiley & Sons, Ltd.