R-Simp: Model Simplification In Reverse, A Vector Quantization Approach

Just as colour image quantization can be viewed as simplification of a three dimensional colour in a two dimensional image, model simplification can be viewed as quantization of three dimensional normal vectors on a two dimensional surface. Thus many of the approaches used in quantization can be applied to the problem of model simplification. We present a model simplification algorithm that is based on the splitting algorithm from quantization literature. The algorithm works in reverse by expanding from a coarse to a fine model. It employs large scale curvature measures to determine which portion of the model to partition next, and how to partition it. The algorithm can accept non-manifold input models, and is capable of simplifying topology. It produces good quality simplifications, and is faster than most other simplification algorithms.