Multi-view isosurface similarity analysis for transfer function design in direct volume rendering

Transfer function (TF) design is crucial in direct volume rendering, yet it faces challenges due to the lack of semantic information about volumetric data. In this paper, we propose an approach to extract the semantic information by clustering isovalues based on a novel isosurface similarity measure and an optimized clustering strategy. The measure is derived from the visual appearance of multi-view rendered images rather than spatial properties. It is designed to more closely model human visual perception mechanisms and supports efficient computation via GPU acceleration. The clustering strategy incorporates both isosurface similarity and isovalue distance to classify volumetric structures and guide semi-automatic TF design. Our proposed approach facilitates the identification of representative isosurfaces and enables users to interactively refine the TF. We demonstrate the effectiveness and generality of our approach across diverse datasets, including medical imaging, industrial CT scans, flow simulations, and combustion data.