Forum

Author Topic: Issues with Micro 4/3 cameras?  (Read 4601 times)

bdcrane

  • Newbie
  • *
  • Posts: 2
    • View Profile
Issues with Micro 4/3 cameras?
« on: October 30, 2016, 12:17:27 AM »
Are there any known issues or problems with using images taken with a Micro 4/3 camera in Photoscan, such as with camera calibration for example?

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 14813
    • View Profile
Re: Issues with Micro 4/3 cameras?
« Reply #1 on: October 31, 2016, 07:17:12 PM »
Hello bdcrane,

Have you tried to process the images in PhotoScan already? If so, then which issues have you faced?
Best regards,
Alexey Pasumansky,
Agisoft LLC

bdcrane

  • Newbie
  • *
  • Posts: 2
    • View Profile
Re: Issues with Micro 4/3 cameras?
« Reply #2 on: November 02, 2016, 05:51:35 PM »
I have used a Lumix G1 Micro 4/3 camera with Photoscan. There aren't any issues that are visually obvious, but then I haven't compared to end results from other cameras. But I had heard that there might be issues not obvious to the eye regarding camera calibration or error reduction. But I can't point to anything specific.

Kiesel

  • Sr. Member
  • ****
  • Posts: 332
    • View Profile
Re: Issues with Micro 4/3 cameras?
« Reply #3 on: November 24, 2016, 10:39:09 AM »
At https://en.wikipedia.org/wiki/Distortion_(optics) you can read that:

Quote
The Micro Four Thirds system cameras and lenses perform automatic distortion correction using correction parameters that are stored in each lens's firmware, and are applied automatically by the camera and raw converter software.

So if you use these corrected images they are corrected a second time by PhotoScan, which can be an (little) issue, because it is allways better to only let PhotoScan do the distorsion correction with uncorrected images. It would be interesting what you get, if you use uncorrected images developed with a raw converter possible to do that, for example with dcraw.

Cheers,

Karsten
« Last Edit: November 24, 2016, 10:41:30 AM by Kiesel »

JMR

  • Hero Member
  • *****
  • Posts: 502
    • View Profile
Re: Issues with Micro 4/3 cameras?
« Reply #4 on: November 24, 2016, 12:12:19 PM »
I've been using m4/3 without any issue since 5 years ago. I actually find some advantages due to their better DOF compared to ff and apsc as well as their lower weight and size.
Regards,
GEOBIT

Kiesel

  • Sr. Member
  • ****
  • Posts: 332
    • View Profile
Re: Issues with Micro 4/3 cameras?
« Reply #5 on: November 24, 2016, 06:06:00 PM »
Thank you GEOBIT for posting the advantages and your postive experience with Mikro 4/3 cameras! So there are no issues with Mikro 4/3 cameras in practice at all. And you are for sure not the only one who is using them!

On the otherside Agisoft recommend to use only photos that are unprocessed before used in PhotoScan. So there is, perhaps only theoretically  ;), a little difference or improvement between only ones or twice corrected images.
Do you or has anybody compared results with images processed in both ways (corrected to uncorrected M 3/4 images in PhotoScan)? Would be interesting how the distortion coefficients differ (I think much) and more important the undistorted images!

Best regards

Karsten

JMR

  • Hero Member
  • *****
  • Posts: 502
    • View Profile
Re: Issues with Micro 4/3 cameras?
« Reply #6 on: November 25, 2016, 04:14:39 AM »
Hello Kiesel: always welcome.

IN CAMERA CORRECTIONS
Almost all modern cameras apply in camera corrections, so, what you get in the photo file (jpg and raw too) is actually a cooked version of the actual image.
If you apply a raw converter made by a third party without any input from the manufacturer, you will see how your photos actually are, and this will make you feel cheated some times. The same happens when you use legacy lenses, just glass and metal, we are so used to "nice things" that the true is dissapointing (Still there are honorable exceptions)
To manufacture lenses to perfection is imposible, and it is much more cheap to correct the image than to correct the lens. For this reason, camera manufacturers have flirted with software makers so well that many cheap lenses today appear to outperform first-class legacy glass.
In camera corrections are generally well implemented, but they correct the bigest part of the problems. I guess the corrections are generally made not considering particular defects of each serial#, like eccentricity, and more likely a supposedly symetrical model is probably applied around a supposed center of symetry in the center of the sensor and not the principal point actually.

OUT OF THE CAMERA CORRECTIONS
Theese can be applied with software tools (lightroom, ACR, photoshop, ptlens,etc.) that are created to make the photos look nicer (straighten lines, no barrel or pincussion  evidence, etc) but again, if this corrections are not made with rigurous math model then they can turn a more or les evident barrel curve into a very straight "something" that is not a straight line. This weird "something"  is often very hard to fit into photogrammetric formulae and that's why Out-of-the-camera corrections are not recommended unless are applied by means of photogrammetry software. In the best case you could obtain the same accuracy with very well corrected images as with their corresponding unprocessed ones... but is more likely you get worse results.

GEOBIT

Kiesel

  • Sr. Member
  • ****
  • Posts: 332
    • View Profile
Re: Issues with Micro 4/3 cameras?
« Reply #7 on: November 25, 2016, 06:27:11 PM »
Hello GEOBIT,

thank you very much for your detailed answer!

Quote
If you apply a raw converter made by a third party without any input from the manufacturer, you will see how your photos actually are, and this will make you feel cheated some times. The same happens when you use legacy lenses, just glass and metal, we are so used to "nice things" that the true is dissapointing (Still there are honorable exceptions)

Yes I have seen such images and thats why I know how the images look like without correction applied.

Quote
To manufacture lenses to perfection is imposible, and it is much more cheap to correct the image than to correct the lens.

I remember those days where you only could use expensive metric cameras for close range photogrammetry. Luckily those days are gone, and Agisoft PhotoScan and others have done a lot for this!

Quote
In camera corrections are generally well implemented, but they correct the bigest part of the problems. I guess the corrections are generally made not considering particular defects of each serial#, like eccentricity, and more likely a supposedly symetrical model is probably applied around a supposed center of symetry in the center of the sensor and not the principal point actually.

Yes that is what I think too. And I don't think that the (very professional) camera industry has photogrammetry in mind when it comes to correct camera-lens imperfections.  ;) There are some coefficients missing and it is not corrected for a single specific lens-camera combination that you own.
And thats why I'am just interested if there would be any difference in the result (for example undistorted images) between processing unpreprocessed images, developed with a third party raw converter like RawTherapee, in one step in PhotoScan or use images as they come out of the camera or out of a professional raw converter.

But that ist just theoretical, I don't own a M4/3 (otherwise I could do a test by myself  :) ) and as you posted, it is a workflow with no issues so little range/space for improving results.

So thanks again,

Karsten

« Last Edit: November 25, 2016, 08:15:00 PM by Kiesel »