Kriegman Research Group
Home / People / Research Projects / Publications / Software / Links


Image-BasedRendering
Rendering and Reconstruction of Surfaces with Arbitrary BRDFs.

Surface properties of many real-life objects often cannot be effectively captured by any existing lighting models (such as Phong). Not only can the reflectance properties be arbitrary, but they can also vary over the entire surface. This project deals with this problem, specifically, how to reconstruct the surface of an object with arbitrary and spatially varying BRDF, and how to render synthetic images of that object under novel illumination.

Our method works by exploiting the fact, that the radiance along a ray of light in empty space is constant. The method requires a calibrated light source and camera setup to gather a set of images under controlled lighting, where the light source is moved on a surface around the object. By double-covering this incident light field by moving the light source over two non-intersecting surfaces around the object, it is possible to recover the surface depth:

Sample image: Recovered depth map: Sample rotated views of the depth map with a texture:

Once the geometry of the scene is known, it is possible to render new images under new lighting conditions using a modified image-based rendering technique. Unlike other image-based methods, we can accurately render the appearance of the object lit by nearby light sources:


Two nearby point sources on two sides of the object.

A line source extending horizontally.

Two point sources one near the surface and the second far away.

A square grid casts a shadow.

A field of grass illuminated from behind by an area light source casts a shadow.

Light passing through a stained glass.

It is also possible to render scenes with a mix of real and artificial objects. Below is an example of a scene rendered using a ray tracer. The object in the center is a real object captured using this technique and rendered into the scene using a custom surface shader:

This method can be also extended to video footage. Image-based models can be automatically composed into dynamically-lit real environments. The method preserves all illumination effects such as shadows and shading:

people publications downloads
Last updated : January 12 2007
Research support