Triangular mesh simplification on the GPU

Ternary Sparse Matrix Representation for Volumetric Mesh Subdivision and Processing on GPUs

“…Despite numerical simulation on GPUs being a widespread and well researched topic, GPGPU volumetric mesh processing is far less common. Surface mesh processing in the form of parallel mesh simplification has been examined by various authors, such as the works by DeCoro and Tatarchuk, Papageorgiou and Platis, and Odaker et al [DT07,PP15,OKV15].…”

Parallel simplification and compression of reality captured models

“…While many existing methods use a triangle soup representation in their work, a more recent work utilized the half-edge data structure in their parallel simplification. Though our method adopts the triangle soup similar to [69], the idea of using a probabilistic multiple-choice scheme to obtain better quality can also be incorporated into other data representations such as the half-edge representation.…”

“…The main difficulty for such parallel collapse is the problem of thread racing: edges chose by multiple threads may have an overlapped influence area (Figure 4.1). In [69], a super independent vertices mechanism was used to tackle this issue. Super independent vertices are vertices that are at least 3 rings away from each other.…”

“…Unlike QSlim, we cannot select the single edges with global minimum cost, since we must perform the simultaneous multiple collapses across the various super independent region even if some candidate edge cost is suboptimal. In [69] the author proposes to sort all the found edges and choose only a subset of these edges with smaller edge costs for collapsing. However, this will incur more computation as we will then need to do the extra sorting.…”

“…However, different from [69] which chooses a single edge in an area defined by super independent vertex, we adopt the multiple-choice selection method to improve the decimation quality in a probabilistic sense.…”