LEVELS-OF-DETAIL GENERATION METHOD FOR SKELETAL MESHES

Abstract

This paper is devoted to the development of an algorithm for Levels-Of-Detail generation from skinned meshes. Animated meshes, unlike static ones, cannot be simplified without redistributing or recalculation bone weights. In some cases, objects of rendered scene have redundant details. It happens when their size on a screen, the distance from a virtual camera and other factors are such that there is no sense to display these objects in their full complexity, as it may significantly impact time for rendering one frame. One of the solutions is to create a set of Levels-Of-Detail for each object – a set of meshes and/or texture which represent same object, but with lower level of detail – and change the original object with them, when it is necessary. The simplification of visual models is especially important for visualisation of digital twins of real-world objects, subjects, or processes within the digital twin technology. An analysis of existing algorithms for Levels-Of-Detail generation for animated meshes is presented and discussed. An improved method for Levels-Of-Detail generation is introduced and discussed. The proposed method is based on Houle and Poulin animated mesh simplification. However, there are the following core differences in the proposed method: weights of resulting vertices are interpolated, not just copied; multiple poses are used for simplification input. These new features allow to achieve the animated meshes simplification without significant drawbacks in animation quality and mesh optimization.