View-dependent conifer LOD models

High compression of plant geometric data is an important aspect in fast rendering of a forest. Conifer trees are typical species, in which the number of leaves is so big that existing simplification processing is still time-consuming. In this paper, we present an efficient view-dependent algorithm to construct continuous level of detail models for coniferous foliage geometry. The models are defined with cylinder and line representations. At close distances, coniferous leaves are represented by cylinders, which are polygonized into prisms. With progressive increased distance, the cylinders will transform into lines, and for views even farther, fewer lines will be used to represent the crown by randomly sampling. Experiments show that our approach has three main advantages: smoothness of the transition between cylinder and line models, high realism for any views, and efficiency in data compression.