dialog box label change in 1.5.1 for "Photo-Invariant parameters"

[andyroo]

.....So For Photo-invariant parameters option would it not be better to use instead Photo-variant parameters as in this case a different parameter will be estimated for each photo?

I agree with Paulo's comment in the pre-release topic - I had to dig into the forum to clarify what this was because it is not explained in the manual, and in the version of English I learned, "invariant" means "unchanging" so I was worried that with "Photo-invariant parameters = none" Metashape was letting all enabled parameters vary for all photos.

My suggestion would be either what Paulo said - change the label to "Photo-variant parameters" or maybe something using a slightly more explicit wording, like "Per-photo variable parameters"

This is an interesting setting. Seems like it could be useful for collections where autofocus or image stabilization or shutter priority was enabled...

[Alexey Pasumansky]

Hello andyroo,

The term "photo-invariant" here is introduced as opposed to "block-invariant" that is applied by default even though not mentioned in the interface directly.

So basically, by default all the calibration parameters are "block-invariant" (or "calibration group invariant" in Metashape terminology) - i.e. the parameters are identical for all the cameras in the calibration group. While "photo-invariant" option, that has been implemented in the version 1.5, means that the parameter applies to each camera individually, even if the camera is a part of a larger block (calibration group).

This term ("photo-invariant") is actually used quite widely in photogrammetric community and literature.

[andyroo]

Hello Alexey,

TLDR; "photo-[in]variant" has contradictory usage in the literature, how about "image-variant"?

May I propose a term with less confusion, ambiguity, and contradictory usage? How about "image-invariant" (sensu Tecklenburg et al 2001)?

I understand what you introduced with the option to allow certain parameters to vary by image, and I have a good grasp of the word, "invariant", but the term "photo-invariant" appears to have contradictory usage in the literature and as far as I can tell it should be hit on the head with a shovel and buried, maybe along-side "photo-variant" with some kind of ode to the photograph. Then we can embrace the fact that we live and work in a digital age, and Metashape works with images, even if some of them are images of  photographs.

I mean no disrespect to you or the field of photogrammetry, but I believe that "photo-variant" is the correct usage, and I now know way more about the history of these two words than I ever intended to, including that many people with way more experience and clout in photogrammetry than lowly me will probably fight (each other?) to the death over "photo-invariant" vs. "photo-variant".

A 2014 paper by Habib et al (2014) discusses this confusion of the issue (and arguable misuse of the term "invariant" explicitly, "Small, and hopefully insignificant, variations in the IOPs may occur within the course of a single data collection campaign. These variations are referred to as photo variant (the term used in the cited literature is actually “photo invariant”; however, the authors consider the term “photo variant” to be more appropriate..."

Also - I read a bunch of papers trying to figure this out, and I'm waving the white flag - I can totally see why you used the term "photo-invariant" even though I don't like it and don't agree with it, and I'm not even quite sure how "photo-invariant" came to be acceptable in the literature.  As far as I can tell, the concept of per-photo varying parameters was originally introduced in the literature as "photo-variant" but within a few years the term somehow degenerated to "photo-invariant" in much of the literature. Since I'm a geologist and not a photogrammatrist, I ... I don't even know. I need a beer.

After taking a deeper dive into the literature than I intended, it looks like the term "photo-variant" was originally(?) introduced by Moniwa (1980/1981), as a way to address shortcomings of "block-invariant" calibration. "Photo-variant" calibration was even specifically contrasted with "photo-invariant", aka "block invariant" calibration by Faig and Owolabi (1988), but apparently their explicit distinction between the two (and equation of "photo-invariant" to "block-invariant") was overwhelmed by a landslide of people using "photo-invariant" to contrast with "block-invariant", including the doctoral dissertation of Mostafa Seyed Madani in 1987, and papers by Shortis and  Wester-Ebbinghaus in 1988.

To further complicate the issue (this is about where I gave up), a 2010 paper by Nakano and Chikatsu refers to "photo-variant" in the text, but in a figure 8 they use "photo-invariant" as a category! In the same paper they also describe "image-variant" and "camera-variant" calibration, because clearly there isn't enough confusion yet.

To be fair, the term, "image variant" appears to be a modern derivative of  "photo-variant" used specifically in the context of digital imagery, and introduced (possibly) in the 2001 paper by Tecklenburg et al, Camera Modelling with Image-variant Parameters and Finite Elements. That paper, and the reference to "image-variant" is where I finally saw a way out of this mess that might not make anyone mad.

References:

Faig, W. and Owolabi, K., 1988. The Effect of Image Point Density on Photo-variant and Photo-invariant Bundle Adjustment. International Archives of Photogrammetry and Remote Sensing, 16.

Habib, A., Detchev, I. and Kwak, E., 2014. Stability analysis for a multi-camera photogrammetric system. Sensors, 14( 8 ), pp.15084-15112.

Madani, M.S., 1987. Accuracy potential of non-metric cameras in close-range photogrammetry (Doctoral dissertation, The Ohio State University).

Moniwa, H., 1981. The concept of “photo-variant” self-calibration and its application in block adjustment with bundles. Photogrammetria, 36(1), pp.11-29.

Nakano, K. and Chikatsu, H., 2010. Camera calibration techniques using multiple cameras of different resolutions and bundle of distances. Dimensions, 154(69), p.44mm.

Shortis, M.R., 1988. Precision evaluations of digital imagery for close-range photogrammetric applications. PHOTOGRAMM. ENG. REMOTE SENS., 54(10), pp.1395-1401.

Tecklenburg, W., Luhmann, T. and Hastedt, H., 2001. Camera modelling with image-variant parameters and finite elements. Optical 3-D Measurement Techniques V.

Wester‐Ebbinghaus, W., 1987. Simultaneous calibration of a photogrammetric stereopair. The Photogrammetric Record, 12(70), pp.519-523.

  _{PS - I will never mention this again}

[Alexey Pasumansky]

Hello andyroo,

Thank you for your research on this matter. We'll try to consider the renaming of "photo-invariant parameters" to something less confusing.

[Alexey Pasumansky]

In the version 1.6.0 pre-release "photo-invariant" parameters have been renamed to "image variant" parameters.

Hope it wouldn't result in any further confusion.

[Kiesel]

Thanks, it's clearer now.