Skeleton-Based Globally Localized Optimal and Consistent Cross Field Synthesis
Synthesis of a satisfied cross field is fundamental and important for many applications such as quandrangulation and non-photorealistic rendering.
However, articulated objects consist of both complex interconnections and flatly and um-bilically tubular structures to induce confusion and torsion.
Therefore, this work aims at automatically synthesizing a globally consistent cross field on an articulated object by assembling locally-generated and topologically connecting ones indicatedby its skeleton. Our system uses the skeleton to decomposean articulated objects into capping, connecting, turning, and branching partitions to transform the global synthesis intolocalized partition-based problems in order to achieve robustness, efficiency, and parallelism.
Additionally, our system extrudes each partition to have a proper length and smooth their jagged bound-aries for consistent crossed-boundary constraints of harmonic line gradients and boundary edge flows while applying type-dependent methods on each partition independently to remove helices and relieve twistings. Finally, our algorithm can efficiently and effectively synthesize cross fields on several complex examplesin a fully automatic manner with a comparative quality while comparing to the state-of-art algorithms. We also show that the synthesized field can be used for quandrangulation and non-photorealistic hatching.
Therefore, this work aims at automatically synthesizing a globally consistent cross field on an articulated object by assembling locally-generated and topologically connecting ones indicatedby its skeleton. Our system uses the skeleton to decomposean articulated objects into capping, connecting, turning, and branching partitions to transform the global synthesis intolocalized partition-based problems in order to achieve robustness, efficiency, and parallelism.
Additionally, our system extrudes each partition to have a proper length and smooth their jagged bound-aries for consistent crossed-boundary constraints of harmonic line gradients and boundary edge flows while applying type-dependent methods on each partition independently to remove helices and relieve twistings. Finally, our algorithm can efficiently and effectively synthesize cross fields on several complex examplesin a fully automatic manner with a comparative quality while comparing to the state-of-art algorithms. We also show that the synthesized field can be used for quandrangulation and non-photorealistic hatching.