Painting BRDF Projects
In order to realistically display paintings in a virtual environment, we have developed a method to measure and model the reflective behavior of such objects. There is a large difference in the reflective properties of different types of paint and varnishes. Using a digital camera we measure the BTF of the entire surface, by taking a few thousand pictures of the painting from different directions. Using different sampling techniques it is possible to reduce the number of measurements needed for the acquisition of the BRDF by a factor of 40 compared to regular sampling, with a reduction of Signal to Noise Ratio of only 1dB. The acquired reflectance measurements are fit to the Lafortune BRDF model and rendered for display using a ray tracer. The current method shows promise, but needs to be expanded further to take into account important factors such as surface structure.
It was investigated how the quality of the BRDF degrades with reduced number of light samples from the original surface, and also which reflection directions are most important in this process. Figure 1 shows a possible subsampling scheme. A very important result is that it is possible to reduce the number of samples by two orders of magnitude while loosing only about 1dB SNR in quality of the BRDF (Figure 2).
| Figure 1. Example of a sparse sampling pattern around the perfect reflection direction. | Figure 2. Horizontally the number of samples, and vertically the quality of the BRDF (SNR), for blue metallic paint. |
Finished. Ran till november 2005
|
Heuberger, J. S. (2005). Image Based Acquisition and Rendering of Paintings. Master Thesis, Delft University of Technology, November |
|
Heuberger, J. S., & Pasman, W. (2005). Data Fitting Report. Technical Report, Delft University of Technology. |
|
Heuberger, J. S., & Pasman, W. (2005). Sampling Strategies. Technical Report, Delft University of Technology. |
This project was supported by Netherlands Institute for Cultural Heritage (ICN).
The reflection of a surface can be stored in a Bidirectional Texture Function (BTF).
To evaluate the quality of a BTF scan of a painting, a realtime visualization tool is of prime importance.
This project developed a real-time BTF renderer based on the Lafortune reflection model, augmented with
features to also use a detail height map of the surface and a high resolution high dynamic range environment map
to enable arbitrary lighting simulation.
Furthermore, this project worked on a more accurate BTF generation.
The fitting by Heuberger suffered from artifacts at transitions between different types of paint. In
combination with an incorrect color correction and insufficient alignment corrections to get a single patch of the
actual painting into a single BRDF, the result did not resemble the real painting.
To address these issues (Figure 3), two modifications were done to the BTF: each image patch of the painting now gets
a height offset and a normal vector in addition to the Lafortune lobe.
The height offset from the idealized perfect plane results in a height map, and height mapping is used to render the microgeometry.
Figure 3. Comparison of Heuberger (left), Snel (center) with a photo (right)
Previous scans of the painting done by Jurriaan Heuberger (above)
have a number of disturbing artefacts and are a bit low resolution to do this properly.
Therefore the acquisition was redone, using an improved scanning set-up (Figure 4, 5).
| Figure 4. Impression of the scan setup during scanning. | Figure 5. Sample painting photo. |
Here are a video of the realtime renderer and another video showing the improvements that were made. Of course for the real impression you need to steer the light and your viewpoint yourself but it gives some idea. This is an early prototype of the system, still showing some artefacts in the images.
Finished in April 2007.
|
Snel, E. (2006). Realistische en Interactieve weergave van Schilderijen. Technical Report, Delft University of Technology, January. | |
Snel, E. (2007). Acquisition, Fitting and Rendering of Paintings. Master Thesis, Delft University of Technology, April. |  |
Snel, E. (2007). Master Thesis Software. | |
Snel, E. (2007). Short Manual for Master Thesis Software. Technical Report, Delft University of Technology, June. |
This project was supported by Netherlands Institute for Cultural Heritage (ICN).