Forum

Author Topic: Macro SfM  (Read 14327 times)

mwillis

  • Full Member
  • ***
  • Posts: 140
    • View Profile
Macro SfM
« on: December 21, 2012, 11:05:57 PM »
Has anyone had consistent luck creating 3D models of tiny object using a macro lens or a microscope?  The results I get range from awesome to crap and don't seem to be predictable.

Macro photography has an inherently shallow depth of field.  I try and compensate for this by masking those areas out. Is there something else that may be going on that can cause distortion when shooting this way?  I thought the EXIF data might somehow be incorrect when a lens is in Macro mode but that assumption doesn't appear to be right.

Any advice is appreciated,

Mark

Kiesel

  • Sr. Member
  • ****
  • Posts: 332
    • View Profile
Re: Macro SfM
« Reply #1 on: December 23, 2012, 12:47:11 AM »
Hello Mark,

I have done some christmas makro work 2 years ago with my point-and-shoot-camera Canon A590is (look at the attachment).
The tiny sensor of a point-and-shoot-camera has as advantage a much bigger depth of field. You know the bigger the sensor the smaller the depth of field. Often point-and-shoot-cameras have a very good makro modus too - for example from 1 to 5 cm distance.

I have read in this forum that some people used succesful focus stacking (Enfuse, Tufuse or Helicon Focus). And you can shoot a focus series with this little Canon point-and-shoot-cameras if they enabling CHDK or SDM in an automatic way in the field.

Cheers,

Karsten

mwillis

  • Full Member
  • ***
  • Posts: 140
    • View Profile
Re: Macro SfM
« Reply #2 on: December 23, 2012, 01:15:19 AM »
Karsten, nice work with the A590.  That's one of my favorite KAP cameras. The macro I am working with is an old Ricoh CX1. It works but not everytime.  That means I'm taking many many extra duplicate photos.  This is very time consuming in the field.

As for focus stacking.  I am going to try this with a lightfield camera (lytro) and see how it works with that data...  Unfortunately a Lytro is to long for the work I'm doing.  It won't physically fit. 

Thanks,

Mark

Kiesel

  • Sr. Member
  • ****
  • Posts: 332
    • View Profile
Re: Macro SfM
« Reply #3 on: December 28, 2012, 01:23:37 PM »
Hello Mark,

which objects do you want to photograph? Is there so little space between your objects that your lytro lightfield camera with 112 mm length won't fit?
In opposite to your lytro camera image stacks from my Canon A590IS need to be aligned because of the little sideways lens moving when I photograph image stacks with CHDK. This is an advantage of your lytro, the images should be match perfectly. Is there no output of a sharp photo from near to far in the lytro software?

1 cm macro distance with your old Ricoh CX1 is quit good, but if it is not always working  :( . If I were you, I would look for a comparable Canon CHDK camera and shoot with a small tripod my image stacks.

In my first answer I forgot to mention Photoacute which has a very good image alignment, noise reduction, depth of field extension and superresolution image improving.

In Helicon Focus software there is also a option for 3D object generation from image stacks, don't know how accurate that is. But because this is a standard technique in microscopy there should be other solution for this for example in ImageJ (which is free).

As you also mentioned microscopy, an old stereo photo technique in microscopy is to tilt the probe.

I hope it helps a little bit.

Karsten


Traceologue

  • Newbie
  • *
  • Posts: 21
    • View Profile
Re: Macro SfM
« Reply #4 on: January 21, 2013, 01:27:44 PM »
Well, this is an interesting topic, but not for a short reply. I recently submitted a paper to Journal of Archaeological Science, making a review of the available affordable solutions for 3D recording and modeling at low and high optical magnification in the field of use-wear studies, but one of the reviewers, who is certainly preparing a paper on the same topic, stated that "There is no real new approach in this paper, no new experimental data, no new understanding of micro-wear data". As said Pierre Bourdieux, the French sociologist, science is a combat sport...

According to my experience with petroglyphs, bone and lithic tools, photogrammetry can works up to ~ 1.5:1 magnifcation (= macro objective with a short extension ring), but for helping Photoscan, which does an incredible work, it is better to reduce the resolution of the image (from 10Mp to 3Mp in my tests). At higher magnification, the depth of field is not sufficient for making the work comfortable: too many shots for carefully covering the whole volume or surface and then for processing.  When depth of field is missing, the best solution, which is involved in all the low and high power microscopes allowing 3D recording (except the Leica IC3D module), including the confocal microscope, is  focal plane merging, commonly called image stacking. In the intermediate magnification range, when depth of field is no longer sufficient for photogrammetry but still a bit high for getting a good vertical resolution (= enough steps), the solution is to combine the both principles. I already got nice results doing this, but not sure that the geometrical proportions are preserved.

The most complete and flexible software for image stacking is Helicon Focus (which includes Helicon Remote and a 3D viewer). However, its main function is not to provide 3D models. This is the reason why they are made in low resolution for being usable with any computer. Nonetheless, beside the light obj file, it is possible to save the depth map at full resolution. Such depth map can be processed by Interactive 3D Surface Plot, a plugin of ImageJ for getting 3D view, but no other way than screen shot for saving the result. As part of a research project entitled “Trac?ologie tridimentionnelle” (http://lascarbx.labex-univ-bordeaux.fr/Les-actions-de-recherche-en-cours/Traceologie-tridimensionnelle-T3D,i634.html - sorry, in French...), funded by Lascarbx, the lab Archeovison has made two Meshlab plugins for modeling and calibrating the depth maps from Helicon Focus. These plugins will be available for free within a few months. Here attached an example at 200x magnification as seen in Meshlab (use wear on the edge of a Mousterian scraper).         
 
« Last Edit: January 21, 2013, 03:35:44 PM by Traceologue »

Mr_Curious

  • Jr. Member
  • **
  • Posts: 86
    • View Profile
Re: Macro SfM
« Reply #5 on: January 21, 2013, 03:20:44 PM »
Very informative post, thank you!

Greetings,

Mr. Curious