Optimized Maximum Intensity Projection (MIP)

We present an improved version of the MIP algorithm that requires roughly 2%–10% of the computational effort of the brute-force, straight-forward version. The algorithm conserves the image quality and requires no pre-processing. We propose four different quality levels: a fast line-traversing nearest neighbour algorithm for previewing, two efficient approximation algorithms that are sufficient for most applications, and an analytical method for images of highest quality. Additionally we present an improved cache memory access scheme. The employed ’sub-cube’ volume data representation provides an additional speed up of at least 50% compared to the classical linear memory access. This is of particular interest for (but not restricted to) parallel volume rendering on shared memory multiprocessor systems.