Appearance-Space Texture Synthesis

Abstract

The traditional approach in texture synthesis is to compare color neighborhoods with those of an exemplar. We show that quality is greatly improved if pointwise colors are replaced by appearance vectors that incorporate nonlocal information such as feature and radiance-transfer data. We perform dimensionality reduction on these vectors prior to synthesis, to create a new appearance-space exemplar. Unlike a texton space, our appearance space is low-dimensional and Euclidean. Synthesis in this information-rich space lets us reduce runtime neighborhood vectors from 5x5 grids to just 4 locations. Building on this unifying framework, we introduce novel techniques for coherent anisometric synthesis, surface texture synthesis directly in an ordinary atlas, and texture advection. Remarkably, we achieve all these functionalities in real-time, or 3 to 4 orders of magnitude faster than prior work.

Links

• Paper
  
• Video
  
• Talk slides
  
• Movies for talk
  
• Close-ups of paper figures
  
• Comparisons with [Lefebvre and Hoppe 2005]
  
• Supplementary anisometric texture synthesis results 1
  
• Supplementary anisometric texture synthesis results 2
  
• Supplementary surface texture synthesis results
  
• Supplementary texture advection videos: ClipA, ClipB, ClipC, ClipD.