| 摘要 |
Methods and apparatus are disclosed for creating lifelike digital representations of scenes that include numerous fine-grained objects such as hair. Individualized geometric models are defined for a selected, manageable subset of the fine-grained objects. By interpolating based upon these defined geometric models, subject to user-specified hair parameters, geometries are subsequently generated for the full set of fine-grained objects. Rendering techniques are used to generate two-dimensional image projections of these geometries, as seen from a specified point of view. These steps of geometric interpolation and rendering are performed in an iterative manner, such that the numerous fine-grained objects are processed and rendered portion by portion, thereby greatly reducing the computational complexity of the task. Other aspects of the invention include the use of depth information regarding individual hairs for purposes of performing accurate rendering. Selected portions of depth and velocity information are also retained and utilized in order to composite and motion blur, in a reasonably accurate manner, the rendered hair image projections together with other three-dimensional scene elements.
|
