Analytic Isosurface Rendering and Maximum Intensity Projection on the GPU

Abstract

It is well known that isosurfaces implicitly represented by volumetric data can be analytically rendered if a trilinear interpolation is assumed to be applied for the continuous reconstruction. However, to the best of our knowledge, it has not been investigated yet how this approach can be efficiently implemented on current GPUs and how much the analytic intersection point calculations slow down the rendering process compared to the traditional discrete approximation. In this paper, we propose a GPU friendly first-hit ray-casting algorithm that (1) minimizes the number of texture fetches, (2) significantly simplifies the arithmetic operations, and (3) avoids error accumulation during the ray traversal. We show that our analytic isosurface rendering optimized for the GPU is even faster than an equidistant discrete sampling, if the sampling frequency is set such that a comparable image quality is obtained. This is true even if the empty blocks of voxels are not processed along the rays. Therefore, the analytic approach can completely replace the traditional first-hit ray-casting implementations. Additionally, we show that the core of our algorithm can also be used for analytic Maximum Intensity Projection (MIP).