Accelerated walkthroughs of complex scenes based on visibility culling and image-based rendering

A new algorithm that combines visibility precomputing and image-based rendering to accelerate walkthroughs of complex scenes is presented in the paper, In the preprocessing phase, our algorithm samples the environment in a 3D lattice structure, tile potentially visible surface list of square-frustum at each cell in the lattice are pre-computed and stored in a cube at the center of the cell. Thr intermediate views of the whole scent: from any viewpoint inside the cell are generated by rendering only the potentially visible surfaces lying within the current view-frustum. Environment map: synthesized by the image of distant visible surfaces in tile PVSLs, are employed to speedup the algorithm further. Experiment results show that only a small fraction of the surfaces in a complex scene are involved in the rendering phase by our algorithm, so it is not only very effective for densely occluded environments, but also suitable for mostly unoccluded outdoor scenes.