Publication: Three-dimensional volume visualization
Loading...
Date
2012-06-01
Authors
Lim, Han Ying
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Rapid growing size of volumetric data resulted from advancement ofdata-acquisition technology had placed mounting pressure in the volumevisualization field to develop a high-speed volume renderer. Over the years, tremendous approaches and optimization algorithms were developed to caterfor the demands. However, much of the algorithms proposed strived hard tobalance between the performance optimization and quality optimization. Most ofthe existing optimization algorithms reduced the amount of volumetric data to be processed, before or during the rendering, to reduce the computation cost.These algorithms gain enhancement in computational speed at the expense of losing fidelity of the image. Therefore, there is strong urge for an optimizationalgorithm which enhanced the performance without sacrificing the quality. Theaim of this project is to enhance the computational speed of the 3-D volume visualization technique, and to give interactive visualization of a rendered volume from all viewing angles. The methodology proposed to achieve the goal
is to implement parallel processing onto volume visualization algorithm using OpenMP. Unlike the other existing optimization algorithms, this method doesnot discard any of the volume data to achieve the reduction in computation time. The parallelized algorithm achieves speed up through distributing the computation load on simultaneously executing threads. The performance of parallel processing was investigated by comparing to the performance of serial executed volume visualization algorithm. The project also seeks for the maximum capability of speed up achieved by parallelized algorithm through manipulating the number of threads. Another limitation of the existing algorithmsis that, these algorithms highly dependent on the characteristic and structure of the volume data to achieve optimum performance. The performance varies across different set of volume data. To justify this, experiments had been carried out to compare the performance of early ray termination and OpenMP algorithm. This project also aimed to introduce interactivity into 3-D volume
visualization via volume rotation. In this project, volume rotation algorithm wasincorporated with the volume rendering techniques to enable the interactive view of the 3-D volume visualization. Forward mapping and reverse mapping implementation methods were investigated, and the result were evaluated based on the visual inspection. The experimental results revealed that the OpenMP parallel algorithm has superior performance over serial execution of volume visualization algorithm. Also, this project proved that parallel OpenMP algorithm’s performance does not degrade when the volume data set is highly transparent. This experimental finding reflects that the volume data.
characteristic have no effect on the performance of the OpenMP algorithms, as
the operation is independent on the volume data structure and characteristic.