The tools for fitting of model parameters are independent from Radiance too. The examples start with the Ward Gaussian model, which happens to be the main model in Radiance. Other models or simulation programs may be covered in the future.
Your invited to use the data and come up with better mathematical functions in light simulation. If you're kind enough to sent us a copy of your paper, this will be much appreciated. Please note the disclaimer for our data.
For questions on results, please see the conclusion FAQ.
choosing parameters for the Radiance plastic material
Easy. You can choose between simple ways and more accurate ways. Here's a list:
- beginners: very crude & easy approximation: ideal diffuse, grey surfaces
Set specularity and roughness in your materials to zero and set all three RGB values to the direct-hemispherical reflectance value (Greek letter rho ρ).
Crude estimates for RGB values are 0.9 for white painted walls, 0.5 for medium grey and 0.15 for dark grey.
On the BME data pages for specific materials, you'll find ρ in the left column of the summary table. The RGB value may depend on the incident direction (incident angle theta_in), at this level of estimation, an average may be used.
For a white painted standard room, the calculated illumination levels will provide a first approximation. Surfaces in the image will look a bit ''dull'', since they are ideal diffuse. Glare values from these images would almost certainly be wrong, since there are no ideal diffuse surfaces in reality. Surfaces which act as light transport, e.g. light shelves, indirect reflectors, etc, should not be modelled this way, since resulting light distributions will be very incorrect as well.
- advanced: estimate specularity and roughness
Same as above, but get an estimate for specularity and roughness from the three left columns of the summary table. Again, these values may depend on the incident direction (incident angle theta_in). At this level of estimation, a suitable average may be used.
- near expert: estimate the errors between model and material
The two right columns in each summary table give the quality and the error of the plastic model in comparison to measured values of a real material:
The chi-squared value is an indicator of the differences between measured data and model, generated by the algorithm which finds the optimum parameters. It is mostly useful to compare results for two materials, or two different type of models for one material. The value itself is a bit hard to interpret, the lower, the better.
The max/min factor is easier for practical use: It specifies the maximum and minimum factor between model and measured data for each incident direction. The ideal value would be 1. For example, a factor of 2 indicates that the model reflects twice as much light as the material really reflects into a certain outgoing direction for this specific incoming direction (theta_in). A value below 1 indicates that the model under estimates the reflected light. See below for more details on where the differences between model and material are.
- expert: understand the errors between model and material
On the web page for each material, the big table lists the details for each measured incident direction. Each row has the incident angles, the fitted parameters for the model and the differences between model and material for all measured outgoing directions. The two right columns display a 2D plot along the scatter plane.
The min/max errors in the summary table are the minimum/maximum values of the plot in the 'vis model/data' column.
This data offers an in-depth estimate how well the use of the Radiance plastic model reproduces real materials.
- expert-plus: make RGB, specularity and roughness depend on incident direction
For most measured materials, the RGB, specularity and roughness parameters depend on the incident direction, mostly on theta_in . This by itself indicates that the plastic model is not an optimal functional model of light scattering. The ideal model would have parameters that are independent of the incident direction, and depend only on the material itself.
As a work-around and approximation, two plastic models can be linearly combined using a Radiance mixfunc, which itself depends on the incident direction. For now, we leave the details as an exercise to the reader.
So, what next ? Click on a material at the gallery page.
parameters of the Radiance trans material
The functional model for the transmitted light in trans is essentially the same as for plastic, the Gaussian Ward model. See the technical pages. See gallery page.
getting the measured BSDF data directly
Yes, the measured BRDF,BSDF data can be downloaded for each incident direction by clicking on the 'file' link in the 'BSDF data' column. It's yours to do whatever you want with it, provided the license is followed.