The RADIANCE lighting simulation and rendering system
SIGGRAPH '94 Proceedings of the 21st annual conference on Computer graphics and interactive techniques
Detail to attention: exploiting visual tasks for selective rendering
EGRW '03 Proceedings of the 14th Eurographics workshop on Rendering
Visual attention for efficient high-fidelity graphics
Proceedings of the 21st spring conference on Computer graphics
Perceptual rendering of participating media
ACM Transactions on Applied Perception (TAP)
A physically-based client-server rendering solution for mobile devices
Proceedings of the 6th international conference on Mobile and ubiquitous multimedia
| Hi-index | 0.00 |
High-fidelity rendering is very computationally expensive making it difficult to achieve interactive rates except for simple scenes. Recent selective rendering techniques [Yee et al. 2001; Cater et al. 2003], which alter the number of rays cast per pixel, have been explored to achieve this goal. These approaches have shown that rendering times can be significantly reduced without perceptual degradation. In traditional ray-traced images aliasing is removed by supersampling the image plane. In this sketch we identify the threshold at which decreasing the number or rays shot per pixel would result in no perceptual degradation. We conduct psychophysical experiments using four realistic environments and one test environment as a comparison. This test scene was designed to exhibit high spatial frequencies and thus was a worst case for aliasing. We determine the computational bound by varying the number of rays shot per pixel in both still images and animations. The lighting simulation system Radiance [Ward 1994] is adapted for use in these experiments. The results can be used in the design of more effective perceptual selective rendering algorithms; using the computatinal bound as an indication of where to threshold the rendering process [Sundstedt et al. 2005]. Selective rendering will alter this threshold based on perceptual importance of pixels within the image. This will reduce computation time while maintaining a perceptually high quality result for realistic scenes.