关键词: GPU, 非结构化网格, 光照, 梯度估计, 体绘制

Abstract:

Gradient estimation is necessary for illumination effects which play an important role in volume rendering. Compared with the structured grids, it is difficult to estimate the vertex gradient for unstructured grids due to the complicated topology. Futhermore,realtime rendering can be hardly achieved due to the computation complexity of the resampled point gradient and the difficultity of implemention on GPUs. As a result, most of the unstructuredgrid volumes have not been lit. This paper presents a method to estimate the vertex gradient with high precision for unstructured grids. It employs the volumeweighted extrapolation and the inversedistance extrapolation to compute a vertex gradient using a group of cell gradients estimated by the Green theorem. Meanwhile, to compute the gradient at a resampled point, we present an efficient method based on the cell divergence which makes the cost much lower than the recent method of fulllinear interpolation。The realtime performance of our algorithm even for the relative large data sets can be achieved by its GPU implementaion with the aid of a welldesigned data structure.

Key words: GPU, unstructured grids;illumination;gradient estimation;volume rendering