This is the project homepage for our CS348b Final Rendering Project

• Inam Rahman – imalik@stanford.edu
• Georg Petschnigg – georgp@stanford.edu
  

For our CS348b rendering project we hope to explore realistic rendering of Glass artwork. With its multiplicity of colors, reflections, caustics and refractions, glass is a “poster child” for realistic rendering. In this project we will render a glass vase. For our results to look good we have to answer the following questions:

• How to model the caustics which occur when light passes through glass?
• How to acquire the color texture of a translucent glass piece?
• How to acquire the transparency map of a glass piece?
  

Georg took these images in Seattle, 2001. These particular pieces show how complex caustics arise from  rather simple geometric shapes.


These pieces here a by Andrew Weiner and are currently on display at the Rincon Hill Art Exchange. We have contact the owner Ed Weiner.


This rendering was done by Adam Finkelstein. The shadows are too “sharp” and the image does not contain any caustics. As per our conversation with Adam Finkelstein, it is the caustics that are wrong in his images (Images Inspired by Chihuly Glass). The shadow is too uniform. Instead there should be light and dark areas due to the lens qualities of the glass. Also, these effects should be sharper near the base of the dish and more blurred/blended further away. He suspects that photon mapping can help with these problems.

The following techniques will be useful in our endeavor:

• Photon Mapping and Monte Carlo Integration
• Spherical Gantry for Texture/Translucency Map acquisition
• Inverse Digital Compositing
  

We plan to implement photon mapping to render caustics caused by the glass. Caustics can be one of the important noticeable effects in scenes with glass work, as can be seen in some of above pictures. Caustics are formed when light reflected from or transmitted through one or more specular surfaces (such as glass) strikes a diffuse surface. Under appropriate illumination conditions, light is focused by the glass on diffuse surface and it forms a focused, bright area, the caustic. Thus properly rendered caustics can make a difference to realism of glass work rendering. Caustics and reflections of caustics are very difficult to handle with classical Monte Carlo ray-tracing techniques. In contrast, caustics are very easy to compute with photon mapping. Henrik Wann Jensen’s book “Realistic Image Synthesis Using Photon Mapping” will be instrumental here.

One of the unique aspects of glass art is the intricate swirls of color that are molten in the piece. We hope to acquire a vase’s actual texture and translucency map using Stanford spherical gantry.

The following image shows the intended setup:

The gantry will rotate the vase about its center z axis by 360 degrees, while the camera records an image at each half degree (OPEN ISSUE: need to determine how many samples we need – 720 enough). The final texture is reconstructed from the center column’s of each image.

This process is repeated twice: for the first pass the vase is filled with a white substance (Milk/Chalk Water???) and for the next pass with a black fluid (black ink with water???). Mike or Pat: Do you have any suggestions on what kind of fluid to use?

Once we have two textures we can subtract them to generate the transparency texture.

Alpha * Image + (1 – Alpha) * 1 = IBright
Alpha * Image + (1 – Alpha) * 0 = IDark

Alpha = 1 – (IBright – IDark)

We may have to adjust play manually tweak the gamma curve of the alpha map to account for non ideal white and black substances.