Accelerating volume rendering with quantized voxels

Abstract

We present here a new algorithm for accelerating volume rendering with an orthographic projection. Because volume rendering handles huge data sets, a reduction in the computational cost of voxel projection is required to obtain interactive volume rendering. We satisfy this issue by using the possibilities of orthographic projection that allows the quantization of voxel positions by subdividing the pixels. The same projection properties are given for all the voxels with the center falling within the same pixel subdivision. In contrast with classical algorithms that require several instructions to compute either the next traversed voxel or the next rasterized pixel, our method needs only one addition instruction and one addressing instruction that is sufficient to determine one projected pixel. Splatting can also have a decisive advantage of it. Our algorithm is well suited for low-end platforms when no hardware acceleration is available. Experimental results show that our rendering rate is better than other existing methods. This algorithm might allow obtaining real-time volume rendering on conventional computers soon.