Report: Photon Mapping affects the Visualization of Products in Room

I released the interim report. I would be happy if you suggest anything about the report and my project. I described all of the details of the calculation conditions in the report at the following URL:

Summary: This study tests how photon mapping of Radiance affects the visualization of products in rooms. I placed wine glasses in an entrance of my house, which is well-lighted by natural illumination, and took pictures and referenced them to confirm what phenomena were missing in the computer graphics rendered images. Referring to photographs for improving the visualization is one of the approaches to enhance the CG prediction quality of how products will look before their completion. (I used the same approach to improve the CG lighting simulation of a bridge over a valley in 1998 and 1999.) In this study, I found that photon mapping can show the pattern of the coaster laid under the foot of the wine glass. In contrast, the classic Radiance fails to show the coaster pattern through the glass viewing from certain angles where there is a conical gap between the coaster and the foot of the wine glass.

My website: Experimental Project of Collaboration for Art, Science, and Education beyond the Net: BEYOND THE NET: TOOLS

1 Like

Very nice work, Yoichi!

I’m wondering what happened to the table-top reflection of the window in the photon-mapping version of the entrance. Did the specularity of the material or something change? Does anyone else know what’s going on, here?


1 Like

Hello! Greg,

Thank you for your reply. When using the photon mapping in Radiance, I have experienced that the plastic type’s parameter a4, reflectance, must be larger, and a5, roughness, must be slightly smaller than those of the classic Radiance, for obtaining specular reflection images. (The experience is not so long, a month.) The values a4=0.05 and a5=0.015 are not unfounded since I fitted the polished paint values, measured by the gonio-spectrophotometer in 1998, using the least-squares method. I believe that these values are appropriate for the classic Radiance because I obtained near-photographic images using these values when I published my paper in 1999[1]. However, my optical property model of the box common to the photon mapping and classic Radiance and the commands to execute both methods may be incorrect. Please refer to pages 4 through 6 of my report[2] for details on these calculated conditions.
[1] Yoichi Mizomata - aij_paper5
[2] BEYOND THE NET: TOOLS - Report10

void plastic box
5 0.5 0.5 0.5 0.05 0.015

box colorpict Mat-Box-fr
13 red green blue box.hdr match_u match_v -ry 90 -rz 90 -s .6 0 1 1

box colorpict Mat-Box-bk
11 red green blue box.hdr match_u match_v -ry 90 -s .6 0 1 1

box colorpict Mat-Box-top
15 red green blue box.hdr match_u match_v -ry 90 -rz 90 -rx 90 -s .6 0 1 1

void mirror reflection_of_box
3 Mat-Box-fr Mat-Box-bk Mat-Box-top
3 0.5 0.5 0.5

$ oconv -n 1 -r 16384 chamber.rad coaster.rad wine-glasses.rad skyX.rad window.rad door.rad > wine-glasses.oct
$ mkpmap -apg global.pmap 10m -apc caustics.pmap 60m -apo Mat-Photon-B -apo Mat-Photon-L wine-glasses.oct
$ rpict -ab 8 -ap global.pmap 500 -ap caustics.pmap 3000 -vf camera.vf -x 4096 -y 4096 wine-glasses.oct > wine-glasses-pm.pic
$ rpict -dt 0.001 -lw 0.001 -lr 14 -st 0.001 -ab 8 -aa 0 -ar 0 -ad 4096 -as 1024 -vf camera.vf -x 4096 -y 4096 wine-glasses.oct > wine-glasses-cr.pic

Best regards,
Yoichi Mizomata

Wow, this is an amazing effort – many thanks for posting this, Yoichi! Will take a look at your detailed report too. Incidentally, this would be an interesting topic for the upcoming workshop.

The subdued specular reflection on the tabletop is actually a known issue in conjunction with a mechanism that prevents double counting of caustics, and I’ve been busting my head over how to best reconcile it within the RADIANCE framework.

@Lars_Grobe has a prototype patch he’s been working on which addresses this issue. We plan to integrate it with the next release of the photon mapping code, which will also include photon flow and the new precomputed contribution photon map which I developed in the course of the latest projects at HSLU. Unfortunately Lars and I now have other priorities, so I can’t give you a definite answer when this will actually happen.



To confirm the double counting of this topic, I rendered a model with polished granite material attribute in the same scene and under the same conditions. This model was converted from the actual statue to triangular mesh data using a 3D scanner app on iPhone.

Hello, Greg!
When will you release a version of Radiance with improvements regarding double counting of caustics? I want to participate in a verification test.
Best regards,

Hi Yoichi,
as Roland already pointed out, we are working on this as part of finishing the development of recent photon mapping extensions. Before this is completed, it cannot be integrated into Radiance. It is good to know and highly appreciated that you would contribute to this with testing!
Best, Lars.


Hi Lars,
Thank you for your reply. I look forward to the release.
Best regards,