On Temporal Coherence used for Hidden Surface Removal based on Ray Casting: Techniques, Results and Limitations

surface removal can be based on software ray casting algorithm. In the
context of this particular visibility computation we focus on utilization of
temporal coherence for online and offline walkthroughs for static scenes. We
discuss a few techniques in context of reprojection for ray casting, that in
particular combination give a conservative visibility algorithm. Further, we
show that the properties of well modeled scene can influence the accuracy
obtained by approximate algorithms. In order to get the exact hidden surface
removal algorithm utilizing temporal coherence, we apply the properties from
epipolar geometry, splat kernel 3x3 pixels for reprojection from a previous
image of an animation, a mere restriction of camera motion, and subsampling
1/9 of pixels by an ordinary ray shooting algorithm. In our algorithm
reprojection is only means for shooting rays towards appropriate objects in
order to get exact visibility results. We have found out for tested
walkthroughs that more than 80\% of pixels can be computed by a single
ray-object intersection. The efficiency of our reprojection-based technique
in temporal domain is sensitive to the visual complexity of the rendered
images in animation. Theoretically, the time complexity of our exact
algorithm is at most by 88\%. The presented algorithm is conceptually
incomparable to rasterization approaches for polygonal models since it can
process arbitrary shapes of objects using ray-object intersection and its
time complexity is linear with the number of pixels in the image. Finally,
we show by using the complexity measure based on an ideal ray shooting
algorithm that further significant algorithmic improvement of hidden surface
removal based on ray casting is rather unlikely to happen in future.