View-based Rendering: Slide 2 of 17.

Click slide for next, or goto previous, first, last slides or back to thumbnail layout.

Click slide for next, or goto previous, or back to thumbnail layout.

In traditional computer graphics one tries to model both the object geometry and how the surface interacts with light.

In image-based rendering one typically doesn't model much. Instead, one collects images of an object or an environment, and then creates new images by morphing and combining the old ones. One motivation is caching rendering computation [we could render images of complicated scenes beforehand and interpolate between them to increase the frame rate for animation or walk-throughs], another is creating realistic images of real objects.

In our approach, which we call view-based rendering, we concentrate on the second motivation. We create view-dependent geometrical models that we texture map using color images. To view the object, we first render a few of these models separately and then combine the partial images into a final image.