Flash Light Distance

[Sigmund]

The objects I try to document are erected stone slabs. I have to document carvings in the lower parts of the slabs, near to the ground. For this reason, I often have problems to keep the distance between flash and stone surface. The distribution looks like that:

 

 

Surprisingly, I can not see any decrease in quality. The RTI viewer does its job, I see no difference. When I make a PTM of the same set of images, however, I notice that some light directions are underrepresented. In particular in the mode Specular Enhancement.

 

Does anyone have similar problems? Is it better to leave off certain parts when the distance can not be kept (as I did) or do you always take the longest distance possible, even if it is relatively short?

 

Sigmund

[Taylor]

Hi Sigmund,

 

CHI usually recommends a light distance of 3 times the width of the object you're capturing, but you can reduce the distance to as low as 2 times the width and still get a good result (I measure the diagonal of the image area for this calculation).  There are always trade-offs, and your choices will depend on the purposes for which you're planning to use the RTIs, particularly if you plan to use the information quantitatively or qualitatively.  The method is very robust, so you have quite a bit of leeway before you get big errors in the normal directions.  Limitations on the light distribution are very common with larger objects, because of the ground surface, floor, walls, or ceiling. 

 

The best way to improve the light distribution that I can suggest (other than to move the subject to an easier location for the capture) is to reduce the area of the subject you're capturing, by moving the camera closer or using a longer focal-length lens.  This can give you more detail, but of course the trade-off is you need more RTIs to cover the entire subject.  As you said, you can leave out some light directions and still get a very usable RTI for qualitative interpretation. 

 

I used a spreadsheet, which I'd be happy to share with you,  to help me decide what distances to use for capturing RTIs of a large painting.  There is also a tool that was developed by ISTI to plan the placement of lights before the highlight method was developed.  It creates a very nice 3D representation of the room for planning RTIs of larger objects.  I got a copy of the PTM Planner tool from one of the authors of the paper, "High Quality PTM Acquisition: Reflection Transformation Imaging of Large Objects," here:

http://vcg.isti.cnr.it/Publications/2006/DCCS06/

I could ask them for permission to share it with you, which I think they wouldn't mind.

[Sigmund]

Hello Taylor,

 

thank you for your reply, it was helpful. Good to hear, that others made similar experiences. Did I understand correctly, that I can use different distances between light and object in one capture, if necessary? Do you also reduce the light intensity in those cases?

 

Reducing the area of the subject and making some more RTIs seems to be the best way. My best results are always with quite small portions. Another reason for making smaller RTIs is: Most of the time I'm alone and without any assistance. If I had longer distances, measuring becomes extremely difficult or even impossible. The areas I choose now are about 30cm, the distance then is 90-120m. This works, even when I'm alone.

 

Testing the spreadsheet and the PTM planner would be wonderful!

 

Sigmund

[Taylor]

Hi Sigmumd,

 

I didn't mean to suggest that you could use different distances for the light in the same RTI capture sequence.  Once you choose the distance, you should be as consistent as possible for all the light positions in the sequence, and don't change the flash power settings.  Although CHI recommends using 3 times the width of the image area for the flash distance, they've said that you can reduce the distance to 2 times the width.  I've always measured the width on the diagonal. 

 

I'm not sure exactly what are the effects of reducing the flash distance to 2 times the image width, but my guess is that there could be a moderate increase in the error windows for the coefficients in the bi-quadratic polynomial functions that calculate the normals.  There might also be less room for error in your positioning of the light source, such that small variations in pointing the light at the subject might lead to greater errors in the normal calculation.  You'll see that moving the light closer to the subject makes it harder to get an even distribution of light over the surface.  You'd have to experiment to find out if this will adversely affect the appearance of your RTIs, and whether these errors are acceptable for the purpose for which you're using the RTIs.

 

I've sent a note to one of the authors of the paper I mentioned above, and I'll let you know when I hear back from him regarding permission to share the PTM Planner tool.  I don't expect this would be a problem.  Remember that the PTM Planner was developed before (or while) the Highlight RTI method was being developed, so it's not used very much.  There have been other templates developed for positioning the light source.  For example, if you can position the subject horizontally instead of vertically indoors (this might not apply to your stone subjects), here's a nice, detailed setup:

http://www.si.edu/content/MCIImagingStudio/papers/QuadriFlashMantis_RTI_Technical_Note.pdf

 

I've attached a copy of my planning spreadsheet to a previous post on this topic, here:

http://forums.culturalheritageimaging.org/index.php?/topic/210-planning-for-high-resolution-and-multispectral-rti-capture-and-photogrammetry/

 

You'd need to modify the spreadsheet for your camera, sensor, and lenses.  Try it out and see if it works for you.

[Sigmund]

Taylor,

 

thank you so much! It is great to have people here who share their experiences.

 

Sigmund

[Taylor]

You're welcome, Sigmund!

 

I heard back from Matteo Dellepiane, one of the authors of the paper I mentioned above, and he says it's ok for me to post the link for you to download the executable file for the PTM Planner tool:

http://vcg.isti.cnr.it/~dellepiane/PTMPlanner.zip

 

It downloads as a zip file.  After you open the zip file (using a program like Winzip or 7-zip), you can run the executable just by double-clicking it.  I've only used it in Windows XP (32-bit); so I don't know how it runs on other platforms. 

 

Matteo also requested that I post the citations for both papers that describe how they used the PTM Planner.  The papers also discuss the relative accuracy of PTMs using different numbers of light positions, compared to a 3D scanner:

 

High Quality PTM Acquisition: Reflection Transformation Imaging for Large Objects, by Matteo Dellepiane, Massimiliano Corsini, Marco Callieri, and Roberto Scopigno

http://vcg.isti.cnr.it/Publications/2006/DCCS06/

 

Or:

 

Reflection Transformation Imaging on Larger Objects: an Alternative Method for
Virtual Representation, by Massimiliano Corsini, Matteo Dellepiane, Marco
Callieri, and Roberto Scopigno

http://vcg.isti.cnr.it/Publications/2006/CDCS06/Reflection.pdf

[Sigmund]

Thank you!