Multi-coefficient Parallel Adaptive Wavelet Rendering
(2015) In LU-CS-EX 2015-05 EDA920 20151Department of Computer Science
- Abstract
- Adaptive Wavelet Rendering is a sampling method used for ray tracing in order to render photorealistic images. The concept of wavelets and the so-called discrete wavelet transform is used to create a multi-scale view of the image when sampling. This allows the method to identify image variance on different levels and therefore to differentiate and appropriately handle variance resulting from sharp edges or blurred regions, thus creating visually appealing images with minimal work even for complex scenes.
This thesis investigates the algorithm and specifically how it can be improved through multi-core concurrency. To this end an alternative version is proposed which works on multiple regions simultaneously. Parallelism is considered for... (More) - Adaptive Wavelet Rendering is a sampling method used for ray tracing in order to render photorealistic images. The concept of wavelets and the so-called discrete wavelet transform is used to create a multi-scale view of the image when sampling. This allows the method to identify image variance on different levels and therefore to differentiate and appropriately handle variance resulting from sharp edges or blurred regions, thus creating visually appealing images with minimal work even for complex scenes.
This thesis investigates the algorithm and specifically how it can be improved through multi-core concurrency. To this end an alternative version is proposed which works on multiple regions simultaneously. Parallelism is considered for both the original and the alternative version. Furthermore, they are compared both based on the qualita- tive difference between their results and their respective performance gains through concurrency.
It is shown that although the structure of the algorithm limits the potential for concur- rency, some improvements can be made, especially for the alternative multi-coefficient version with results maintaining high quality, thus making it better suited to todays highly parallel compute systems. Finally some future directions are considered based on the detailed analysis of how concurrency affects the major components of the algorithm. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/5267920
- author
- Somers, Robin LU
- supervisor
- organization
- course
- EDA920 20151
- year
- 2015
- type
- H3 - Professional qualifications (4 Years - )
- subject
- keywords
- Rendering, Ray tracing, Sampling, Wavelets, Adaptive
- publication/series
- LU-CS-EX 2015-05
- ISSN
- 1650-2884
- language
- English
- id
- 5267920
- date added to LUP
- 2015-04-08 08:22:23
- date last changed
- 2015-04-08 08:22:23
@misc{5267920, abstract = {{Adaptive Wavelet Rendering is a sampling method used for ray tracing in order to render photorealistic images. The concept of wavelets and the so-called discrete wavelet transform is used to create a multi-scale view of the image when sampling. This allows the method to identify image variance on different levels and therefore to differentiate and appropriately handle variance resulting from sharp edges or blurred regions, thus creating visually appealing images with minimal work even for complex scenes. This thesis investigates the algorithm and specifically how it can be improved through multi-core concurrency. To this end an alternative version is proposed which works on multiple regions simultaneously. Parallelism is considered for both the original and the alternative version. Furthermore, they are compared both based on the qualita- tive difference between their results and their respective performance gains through concurrency. It is shown that although the structure of the algorithm limits the potential for concur- rency, some improvements can be made, especially for the alternative multi-coefficient version with results maintaining high quality, thus making it better suited to todays highly parallel compute systems. Finally some future directions are considered based on the detailed analysis of how concurrency affects the major components of the algorithm.}}, author = {{Somers, Robin}}, issn = {{1650-2884}}, language = {{eng}}, note = {{Student Paper}}, series = {{LU-CS-EX 2015-05}}, title = {{Multi-coefficient Parallel Adaptive Wavelet Rendering}}, year = {{2015}}, }