Sylvain Lefebvre

Parallel controllable texture synthesis

We present a texture synthesis scheme based on neighborhood matching, with contributions in two areas: parallelism and control. Our scheme defines an infinite, deterministic, aperiodic texture, from which windows can be computed in real-time on a GPU. We attain high-quality synthesis using a new analysis structure called the Gaussian stack, together with a coordinate upsampling step and a subpass correction approach. Texture variation is achieved by multiresolution jittering of exemplar coordinates. Combined with the local support of parallel synthesis, the jitter enables intuitive user controls including multiscale randomness, spatial modulation over both exemplar and output, feature drag-and-drop, and periodicity constraints. We also introduce synthesis magnification, a fast method for amplifying coarse synthesis results to higher resolution.

Appearance-space texture synthesis

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.

State of the Art in Example-based Texture Synthesis

Recent years have witnessed significant progress in example-based texture synthesis algorithms. Given an example texture, these methods produce a larger texture that is tailored to the users needs. In this state-of-the-art report, we aim to achieve three goals: (1) provide a tutorial that is easy to follow for readers who are not already familiar with the subject, (2) make a comprehensive survey and comparisons of different methods, and (3) sketch a vision for future work that can help motivate and guide readers that are interested in texture synthesis research. We cover fundamental algorithms as well as extensions and applications of texture synthesis.

Perfect spatial hashing

We explore using hashing to pack sparse data into a compact table while retaining efficient random access. Specifically, we design a perfect multidimensional hash function -- one that is precomputed on static data to have no hash collisions. Because our hash function makes a single reference to a small offset table, queries always involve exactly two memory accesses and are thus ideally suited for parallel SIMD evaluation on graphics hardware. Whereas prior hashing work strives for pseudorandom mappings, we instead design the hash function to preserve spatial coherence and thereby improve runtime locality of reference. We demonstrate numerous graphics applications including vector images, texture sprites, alpha channel compression, 3D-parameterized textures, 3D painting, simulation, and collision detection.