Espino,F.J., Boo,M., Amor,M., Bruguera,J.D.

**Abstract:**

NURBS surfaces are widely used in computer graphics applications like CAD/CAM, virtual reality, animation,... because of their excellent accuracy of design and low memory requirements. Real-time rendering of such surfaces is commonly performed by tessellation, i.e. generating meshes of triangles that are rendered by current graphics hardware. In order to generate less triangles for high quality surfaces, adaptive tessellation algorithms are better. However, the geometric tests used by these algorithms perform vector computations of high latency that decreases the performance of the algorithm. We propose an adaptive tessellation algorithm that avoids vector computations for testing purposes, replacing them with scalar computations over a matrix of distance values. This way, latency of tests is reduced and therefore, performance improved.