Energy-Efficient Photon Mapping
Mobile devices such as cell phones, personal digital assistants (PDAs), and laptops continue to increase in memory and processor speed at a rapid pace. In recent years it has become common for users to check their email, browse the internet, or play music and movies while traveling. The performanc...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Others |
| Published: |
Digital WPI
2007
|
| Subjects: | |
| Online Access: | https://digitalcommons.wpi.edu/etd-theses/798 https://digitalcommons.wpi.edu/cgi/viewcontent.cgi?article=1797&context=etd-theses |
| Summary: | Mobile devices such as cell phones, personal digital assistants (PDAs), and laptops continue to increase in memory and processor speed at a rapid pace. In recent years it has become common for users to check their email, browse the internet, or play music and movies while traveling. The performance gains are also making mobile graphics renderers more viable applications. However, the underlying battery technology that powers mobile devices has only tripled in capacity in the past 15 years whereas processor speeds have seen a 100-fold increase in the same period. Photon mapping, an extension of ray-tracing, is a robust global illumination algorithm used to produce photorealistic images. Photon mapping, like ray-tracing, can render high-quality specular highlights, transparent and reflective materials, and soft shadows. Complex effects such as caustics, participating media, and subsurface scattering can be rendered more efficiently using photon mapping. This work profiles the energy use of a photon-mapping based renderer to first establish what aspects require the most energy. Second, the effect several photon mapping settings have on image quality is measured. Reasonable tradeoffs between energy savings and moderately diminished image quality can then be recommended, making photon mapping more viable on mobile devices. Our results show that image quality is affected the least as settings corresponding to final gather computations are adjusted. This implies that a user can trade a modest decrease in image quality for significant gains in energy efficiency. Suggestions are made for using energy more efficiently when rendering caustics. Results also show that, although overall energy use is higher with larger image resolutions, per-pixel energy costs are cheaper. |
|---|