Morse–Smale decomposition of multivariate transfer function space for separably-sampled volume rendering

We present a topology-guided technique for improving performance of multifield volume rendering with peak finding and preintegration with 2D transfer functions. We apply Morse–Smale decomposition to segment the multidimensional transfer function domain. This segmentation helps to reduce the number of cases where sampling in transfer function space should be performed, effectively reducing the rendering cost for equivalent sampling quality. We show that the overall performance is increased depending on the topology of a transfer function.

► We use Morse–Smale decomposition to segment the 2D transfer function space.
► The segmentation improves on the peak finding performance for volume rendering.
► The speedup is in average 10–20% depending on the topology of the transfer function.