A triangular mesh of the scene is reconstructed from a set of reference images, and used for
shadow computation. The algorithm uses hardware z-buffer as shadow buffer. The shadow
computation is linear in the number of reference images, and once the mesh is reconstructed
it is reused until the relative position among objects changes.
Adding Shadows to an Image
Using a Single Reference Image
The upper left image was used as input for the reshading process. The other three
images were obtained projecting shadows cast by a virtual light that is moved
Notice the shadows on the hat, face, and shoulders and finger.
Using Multiple Reference Images
Two different reference views of the same object.
Computing shadows using the two reference views.
Two extra reference views to provide more details for the tip and the handle.
Final shadow obtained from the four reference images.
Oliveira, Manuel M. and Gary Bishop.
Dynamic Shading in Image-Based
UNC Computer Science Technical Report TR98-023, University
of North Carolina, May 31, 1998.
Paper (PDF, 0.2 MB)
Last Modified 31 May 1998