FastLIC (Fast and Resolution Independent LIC), presented by Stalling and Hege [34] in ACM SigGraph 95, is one order-of-magnitude faster than the original LIC algorithm by exploiting streamline redundancy that occurs in LIC as a new streamline is always advected for each pixel despite the fact that many pixels are hit by the same streamline. A Fourth-Order Runge-Kutta integrator coupled with adaptive step size and error control is adopted to achieve fast accurate streamline integration, followed by Cubic Hermite Polynomial interpolation to generate evenly sampled points. Equi-distant streamline sampling (by samplen, the sampling interval) in combination with a box kernel simplifies texture convolution to a texture-avergaging operation. Thus a streamline obtained to determine a pixel value can be reused, with only minor corrections (due to the bi-directional kernel offset along the streamline), for its two immediate neighboring pixels since they have nearly the same correlated pixels as their texture contributors (Figure 1). The statistics on pixel hits are employed to evaluate streamline coverage in the flow field and therefore only a small number of streamlines are actually advected for all pixels to collect texture contributions.
Figure 1. Bi-directional kernel offset along the streamline with only minor corrections to collect texture contributions (© Zhanping Liu).
Figure 2. FastLIC images.
Line Integral Convolution
Flow Visualization