next up previous
Next: Shading Algorithms Up: Algorithms for Finding Edge Previous: Silhouette Finding Summary

Other Methods for Calculating Edge Lines

Trying to find these silhouettes based on geometry may be too slow for very large models. The Srf-Node method requires sufficiently refined surfaces and bogs down on large models like the Bezier teapot model which has 23 surfaces.

At SIGGRAPH 1998, Cassidy Curtis presented a technical sketch entitled ``Loose and Sketchy'' [9]. Based on 3D geometry he calculated edge lines using a depth map. The depth map is converted into a ``template image,'' which approximates the edge lines in the image. In the template image, each pixel represents the amount of ink needed in its immediate neighborhood. The template image is created by calculating the magnitude of the gradient of the depth map, thresholding it to give binary values, and then blurring the result. This technique may be faster than using the geometric information to get edge lines. Curtis then uses this template image and another image called the ``force field,'' also created from the depth map, along with a stochastic, physically-based particle system to create sketchy lines. This method is not interactive, and his algorithm takes about 10-60 seconds to generate ``loose and sketchy'' images, but may be interactive if used only to calculate the template image.

Other methods for implementing interactive/real-time edge line should be explored, especially using OpenGL, both in software and hardware. For example there is a sample program called ``silhouette'' which uses the OpenGL API's stencil buffer [26], but this method misses interior silhouettes.


next up previous
Next: Shading Algorithms Up: Algorithms for Finding Edge Previous: Silhouette Finding Summary