Abstract: The advent of GPGPU technologies has allowed for sensible speed-ups in many high-dimension, memory-intensive computational problems. In this paper we demonstrate the effectiveness of such techniques by describing two applications of GPGPU computing to two different subfields of computer graphics, namely computer vision and mesh processing. In the first case, CUDA technology is employed to accelerate the computation of approximation of motion between two images, known also as optical flow. As for mesh processing, we exploit the massively parallel architecture of CUDA devices to accelerate the face clustering procedure that is employed in many recent mesh segmentation algorithms. In both cases, the results obtained so far are presented and thoroughly discussed, along with the expected future development of the work.
Authors: S. Marras, C. Mura, E. Gobbetti, R. Scateni, R. Scopigno.
Two examples of GPGPU acceleration of memory-intensive algorithm.
EuroGraphics Italian Chapter 2010, 49-56.
Genova, Italia, Novembre 2010.
Abstract: Recent advances in acquisition and modelling techniques led to generating an exponentially increasing amount of 3D shapes available both over the Internet or in specific databases. While the number grows it becomes more and more difficult to keep an organized knowledge over the content of this repositories. It is commonly intended that in the near future 3D shapes and models will be indexed and searched using procedure and instruments mimicking the same operations performed on images while using algorithms, data structures and instruments peculiar to the domain. In this context it is thus important to have tools for automatic characterization of 3D shapes, and skeletons and partitions are the two most prominent ones among them. In this paper we will describe an experience of building some of this tools on the top of a popular and robust library for manipulating meshes (OpenMesh). The preliminary results we present are promising enough to let us expect that the sum of the tools will be a useful aid to improving indexing and retrieval of digital 3D objects. The work presented here is part of a larger project: Three-Dimensional Shape Indexing and Retrieval Techniques (3-SHIRT), in collaboration with the Universities of Genoa, Padua, Udine, and Verona.