Detail-preserving compression for smoke-based flow visualization

Smoke is a useful method to visualize flows from scientific experiments or physical simulations in many applications of science, engineering, graphics, and virtual reality. In visualizing flows, smoke evolution inside a flow field creates 3D, high-resolution, and time-varying data sets. The large data size imposes challenge on storing and transmitting the smoke animation results where good compression techniques are demanded. Furthermore, small-scale smoke details play important roles in visualizing and conveying realistic flow behavior. They should be well preserved in compression and reconstructed to create smooth animation effect in decompression. This requirement impairs the direct adaptation of existing techniques in video and volume compression. In this paper, we design new techniques to enable effective smoke visualization compression with smooth detail preservation. The motion estimation between key frames of density fields is implemented by a special bidirectional advection. The intermediate frames are built from advected key frames with specific blending. The advection is driven by motion vectors over nonuniform blocks, which are created with an adaptive simplification of velocity field to reflect the heterogeneous velocity variations. Moreover, pertinent intra-frame compression techniques are integrated. Our approach eventually achieves good compression performance and quality with easy control.