Fast Planarity Estimation and Region Growing on GPU

We present a fast approximate planarity calculation implemented on a Graphic Processing Unit (GPU). The approximate planarity of an image patch is calculated by combining the output of a number of planarity filters. We also demonstrate the use of the local planarity as a criterium for region growing. This region growing is then further optimized using a parallel implementation. The sparse nature of these filters and the inherent parallelism of the filter bank allow a fast implementation on a parallel processor architecture such as the Compute Unified Device Architecture (CUDA) from nVIDIA.