An isosurface c an be eciently generated by visiting adjacent intersected c ells in order, as if the isosurface were p r opagating itself. We previously proposed an extrema graph method, which generates a graph connecting extremum points. The isosurface p r opagation starts from some of the intersected c ells that are found both by visiting the cells through which arcs of the graph pass and by visiting the cells on the boundary of a volume.In this paper, we propose an ecient method o f searching for cells intersected by an isosurface. This method generates a volumetric skeleton consisting of cells, like an extrema graph, by applying a thinning algorithm used in the image recognition area. Since i t preserves the topological features of the volume and the connectivity of the extremum points, it necessarily intersects every isosurface. The method is more ecient than the extrema graph method, since i t d o es not require that cells on the boundary be visited.

This paper describes a level-of-detail rendering technique for large-scale irregular volume datasets. It is well known that the memory bandwidth consumed by visibility sorting becomes the limiting factor when carrying out volume rendering of such datasets. To develop a sorting-free volume rendering technique, we previously proposed a particle-based technique that generates opaque and emissive particles using a density function constant within an irregular volume cell and projects the particles onto an image plane with sub-pixels. When the density function changes significantly in an irregular volume cell, the cell boundary may become prominent, which can cause blocky noise. When the number of the sub-pixels increases, the required frame buffer tends to be large. To solve this problem, this work proposes a new particle-based volume rendering which generates particles using metropolis sampling and renders the particles using the ensemble average. To confirm the effectiveness of this method, we applied our proposed technique to several irregular volume datasets, with the result that the ensemble average outperforms the sub-pixel average in computational complexity and memory usage. In addition, the ensemble average technique allowed us to implement a level of detail in the interactive rendering of a 71-million-cell hexahedral volume dataset and a 26-million-cell quadratic tetrahedral volume dataset.