Parallel iso/aniso-scale surface texturing guided in Gabor space

Abstract

This paper presents a parallel texture synthesis over arbitrary surfaces, generating consistent and spatially-varying visual appearances. A novel scaling field is represented to measure the geometry-aware appearance or geometric deformation, therefore the generated textures locally agree with the geometric structure and maintain the coherence during shape deformation. And we compute the Gabor feature space for the 2D exemplars, to determine the k-coherence candidate pixels. Such Gabor feature space is not only low dimensional but capturing multi-resolution texture structures, hence the k-coherence matching guided by Gabor space improves the performance and quality of pixel similarity measurement. We directly apply multi-pass correction for each vertex according to its local neighborhood for order-independent purpose. Experimental results demonstrate that our method produces significantly improved surface textures with parallel performance.