Using 3-Dimensional Seed Filling in View Space to Accelerate Volume Rendering

This paper presents and evaluates a novel method for projecting a three-dimensional volumetric 6-connected object onto a 2D view using seed filling in the view space. Resampling a 6-connected volume by ray casting produces a new 6-connected volume, provided that the sampling rate is high enough. This new volume can be projected by a 3-dimensional seed filling algorithm. We implement and evaluate a volume rendering algorithm which performs the resampling and seed filling in one pass. With this algorithm the projection can be calculated by accessing only part of the voxels within the data set. The method was implemented on a general purpose computer architecture and tested with several artificial and medical volumetric data sets. The frame rate achieved by the algorithm allows nearly interactive (4-5 frames per second) rotation of typical medical volumetric images (MR data 256x256x128 was used) on a common single-processor personal computer with no specialized hardware. The algorithm requires a preprocessing step to find the initial seed voxels. Our current implementation supports only parallel projection.