Creative Virtual Tree Modeling through Hierarchical Topology-preserving Blending
IEEE Transactions on Visualization and Computer Graphics, 2017, 23(12): 2521-2534.
Yutong Wang, Xiaowei Xue, Xiaogang Jin, Zhigang Deng
A sequence of blended trees. The top row shows a series of morphologically similar yet different trees (middle) obtained by blending between two Salix species (left and right), while the bottom row demonstrates our capability of generating inspiring novel trees by cross-species blending between an artistic Bonsai (left) and a Salix (right).
Abstract
We present a new method to efficiently generate
a set of morphologically diverse and inspiring virtual trees through
hierarchical topology-preserving blending, aiming to facilitate
designers’ creativity production. By maintaining the topological
consistency of the tree branches, sequences of similar yet different
trees and novel intermediate trees with encouragingly interesting
structures are generated by performing inner-species and cross-species
blending, respectively. Hierarchical fuzzy correspondences are
automatically established between two or multiple trees based on the
multi-scale topology tree representations. Fundamental blending tasks
including morph, grow and wilt are introduced and organized into a
tree-structured blending scheduler, which not only introduces the
randomness into the blending procedure but also wisely schedules the
tasks to generate topology-aware blending sequences, contributing to a
variety of resulting trees that exhibit diversities in both geometry and
topology. Most significantly, multiple batches of blending can be
executed in parallel, resulting in a rapid creation of a large
repository of diverse trees.