Adaptive Dynamic Filtering for the Visualization of Phase Portraits

Several methods have been proposed for representing vector fields (glyphs, particles, icons, Streamlines, Hyperstreamlines, Spot Noise, LIC, etc.), being streamlines and LIC the two most important. In previous work we already discussed thoroughly the advantages and drawbacks of these methods, and proposed a new, parametrized method that combines successfully the good properties of both. In this paper we present a different texture-based approach. Instead of convolving the texture along the integral paths as in LIC, the idea is to locally apply a spatial convolution filter to the texture, whose orientation is determined by the local gradient in the vector space. Thus, the spatial features present in the input texture get reinforced along the directions tangent to the flow, and faded out in the orthogonal directions, and hence the same overall result as in LIC is obtained, with a significant saving in computation.