Estimating curvatures and their derivatives on meshes of arbitrary topology from sampling directions

Estimation of local differential geometry properties becomes an important processing step in a variety of applications, ranging from shape analysis and recognition to photorealistic image rendering. This paper presents yet another approach to compute those properties, with comparable numerical and accuracy performances to previous works. The key difference in our approach is simplicity, allowing for direct implementation on the GPU. Experimental results are provided to support our statement.