Unevenly distributed scattered data makes it difficult to be visualized throughout the volume of interest. For this type of data, a two-step approach is usually employed. The first step involves the use of interpolation methods to model the sample data onto an intermediate uniform grid and the second step renders the intermediate grid to form the images of the visualization. A 3D grid with a reasonable resolution consists of hundreds of thousands grid nodes, which makes the modeling process time consuming. The aim of this project is to design and implement a GPU based system to speed up the modeling process.
Following the CUDA GPU programming model, an interpolation-based scattered data modeling system has been designed and implemented. Tests have been conducted with real world data. On an average, GPU performed 43 times faster than CPU for the test cases. The highest speed up is observed on an NVidia GTX 470 where the GPU performed 360 times faster than its CPU counterpart for a grid size of 128x128x128. Experiments have also been designed and performed to validate the system's stability.