Real-time ray casting of volumetric data

In this paper we present acceleration structures and techniques for real-time ray casting of large volumetric data sets, such as those obtained by CT or MRI scans. The techniques used include adaptive sampling and sparse casting. To improve rendering quality we use the regula falsi method and Monte-Carlo ambient occlusion estimation. To enhance the final rendering we use visual effects - screen-space ambient occlusion and depth of field. The algorithms have been parallelized with OpenCL. We compare the results obtained with different methods - isosurface extraction, maximum intensity projection and alpha compositing. We tested our methods on medical data sets, specifically angiograms.