Real-time content-aware texturing for deformable surfaces

Abstract

Animation of models often introduces distortions to their parameterisation, as the parameterisation has been optimised for a single frame. When mapping textures or displacements on a deforming surface with such a constant parameterisation, distortions manifest visually as texture-mapped features appearing uniformly elastic, and such behaviour is not always desired. In this paper we introduce a real-time technique that reduces such parameterisation distortions in areas specified in a provided distortion control (rigidity) map. The parameter space is warped in an axis-aligned way to minimise a non-linear distortion metric using a hybrid CPU-GPU solver. We also extend the technique to compute arbitrary warps for handling more complex use cases. The result is real-time dynamic content-aware texturing that reduces distortions in a controlled way. The technique can be applied to reduce distortions in a variety of scenarios where highly detailed rigid features are represented on a map, abstracted from the underlying low-complexity deforming geometry they are mapped on. Such scenarios include reusing a low geometric complexity animated sequence with a multitude of detail maps, dynamic procedurally defined features mapped on deformable geometry and animation authoring previews on texturemapped models. RELATED WORKParameterisation distortions are typically minimised using mesh parameterisation algorithms [8,20,5]. The majority of such algo-

Real-time variable rigidity texture mapping

“…In figure 4 we use the MugJug example to compare our approach against Koniaris et al [13], Sheffer and De Sturler's method [24] and Mean Value coordinates. In all of these comparisons, our approach preserves local structure under artistic editing better.…”

“…Koniaris et al [13] use distortion control maps on rectangular areas of a mesh and apply a nonlinear optimisation method over the rectangular domain. The technique can be used in real-time for simple distortion control maps, but scales poorly with map complexity.…”

“…The rest pose is the rightmost image; a mug. The deformed frames are shown from left to right: re-parameterisation using our method, Koniaris et al[13], Sheffer and De Sturler[24], Mean Value Coordinates, and using the original UVs. Our method is the only one that preserves the features of the mug at the right scale.…”