Adaptive quantization with error bounds for compressed view-dependent multiresolution meshes

Abstract

We present a simple method to adaptively determine compression parameters according to the given input data. Our method allows to make better use of the available dynamic range (number of bits per component of vertex locations) than with a given fixed-point representation. Moreover, the number of bits per component is also determined adaptively under a user-specified upper bound for the quantization error. Our adaptive quantization method can handle models with largely non-uniform geometric resolution. This is typically the case for manually created models or models derived from different sensors. The experiments we performed demonstrate the benefits of the proposed method.