A performance evaluation framework of multiresolution algorithms for real-time rendering.

摘要

Many multi-resolution algorithms have been proposed to solve real-time rendering problems. However, the performance of these algorithms have not been evaluated, and there is no comparison between them. It is not clear how well these algorithms perform and under what conditions their performance is optimal. Thus the advances in multi-resolution algorithms can not be widely applied in real-time rendering applications.; This thesis presents a performance evaluation framework for multi-resolution algorithms for real-time rendering. It consists of a set of automatic performance measures and a standard real-time rendering testbed. This framework provides a unified system environment to automatically measure a range of multi-resolution algorithms while they perform various real-time rendering benchmarks. Experimental results show that the measured performance data are meaningful and consistent, and can form the basis for performance comparison of the measured algorithms. Several multi-resolution algorithms have also been evaluated using a prototype of the framework. The performance results help the user to compare their performance. These encouraging results demonstrate the potential of the performance evaluation framework, and indicate that it can ultimately promote the application of multi-resolution algorithms in real-time rendering systems.