FOV of reversed lens?

A forum to ask questions, post setups, and generally discuss anything having to do with photomacrography and photomicroscopy.

Moderators: ChrisR, Chris S., Pau, rjlittlefield

Post Reply
lauriek
Posts: 2402
Joined: Sun Nov 25, 2007 6:57 am
Location: South East UK
Contact:

FOV of reversed lens?

Post by lauriek »

Is the field of view of a lens affected when you reverse it?

I'm using a reversed 50/1.8 at f5.6 on around 1/2 - 2/3 bellows extension. PTAssembler wants to know the FOV or the focal length & multiplier of the lens. I have absolutely no idea of the former. I know the latter but I don't know if the fact I've reversed the lens matters to this. Any advice?

Cheers!

rjlittlefield
Site Admin
Posts: 20779
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

I think what you're really asking is "How do I get PTAssembler to align my images?"

The simplest and safest approach is to just lie. Tell PTAssembler that your lens has a fairly small FOV, say 5 or 10 degrees. Pick one image in the middle of the stack as your "anchor". Then optimize to align all the other images, allowing FOV to be optimized separately for every image except the anchor. PTAssembler will end up gradually changing FOV from one end of the stack to the other, just the same way that CombineZM adjusts its scale factor from one end of the stack to the other.

Regarding all the other parameters, I recommend to lock all lens distortion parameters a=b=c=0. If your mechanics do not allow the camera to rotate, then also lock roll=0. Regarding the pitch/yaw and d/e shift parameters, you have to lock at least one pair. For macro stacks, the formally"correct" approach is to lock pitch=yaw=0 and optimize the shifts d and e. But if your FOV is small, it is almost exactly as good to lock d=e=0 and optimize pitch and yaw. (That's why I suggested that setting a small FOV is simple and safe.)

For panorama stitching, it is important to have an accurate FOV so that the various pictures can be mapped smoothly onto a sphere. But for stacks, really all you're trying to do is get the pictures scaled, shifted, and rotated so they line up properly. It's what the panorama community calls a "flat stitching" problem.

Regarding the exact question that you asked, see this paper, particularly pages 11-16. I won't try to summarize here, except to say that despite the terminology, PTAssembler doesn't really want to know field of view, it wants to know angle of view, and the angle of view can vary widely between different lenses even if they have the same field of view and even the same focal length! :? If you care about this (and there's no compelling reason you should), then go read the paper -- hopefully it will help make sense of a very confusing situation.

Hope this helps!

--Rik

lauriek
Posts: 2402
Joined: Sun Nov 25, 2007 6:57 am
Location: South East UK
Contact:

Post by lauriek »

Rik,

I think you read a bit much into my question there, but having said that I'm not averse to the tips! I'd basically worked out how to use it but I was concerned that by filling in the lens parameters for a 'normal' 50mmm lens with 2x factor, that the fact I'd reversed the lens meant these values were not correct and would therefore cause issues. Also I've been using image 1 as the reference, using an image in the middle makes a lot of sense. Would you say 'mathematically' in the middle, or 'in the middle' as in a shot with the most in focus detail (which in the case of my fly would be around 1/3 of the way into the stack I guess - the front of the body/face is certainly not 'in the middle' of the stack if you see what I mean!

As always - many thanks for the pointers!!

I will try to take a look at that paper later on, although I do find a lot of this optical science quite tough!! (in fact I obviously don't even understand properly what FOV means, as I would have assumed this was directly related to angle of view or focal length and sensor size!!) ;)

rjlittlefield
Site Admin
Posts: 20779
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

Laurie,

Sorry for the brain dump. :) For questions surrounding PTAssembler, I'm used to working in other forums where questions like yours usually mean that somebody has gotten strange results and needs to know how to fix them. I'm very relieved to hear that you're getting reasonable results and just wanted to know why!

Which image you pick to anchor doesn't matter a great deal. The issue is only that alignment adjustments accumulate, so starting with a frame in the middle of the stack will keep your output looking most like your input. If you start with one end or the other, you allow more opportunity for the image to grow, shrink, and rotate. It makes a lot of sense to anchor the most detailed frame, or whichever one you care most about, regardless of where it is in the stack.

--Rik

lauriek
Posts: 2402
Joined: Sun Nov 25, 2007 6:57 am
Location: South East UK
Contact:

Post by lauriek »

No, feel free to brain dump as much as you want, it's all good stuff!! :)

I seem to be having more problems now but I'm really not sure what's happening. I did get something out of it the other day which was not perfect but was not too bad, but now it's coming out with really weird results in the preview, like the whole image has been morphed down to an orange segment cross section or something! I suspect I've followed most of your instructions but missed out some key one which is mucking everything up!!

I'll have another go in a bit with your instructions and see how I get on!

rjlittlefield
Site Admin
Posts: 20779
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

That bit about "orange segment cross section" sounds like what happens when the abc lens distortion parameters get weird. Set 'em all back to zero, and be sure they're kept out of the optimization. Optimizing for abc in a stack is just begging for trouble -- there's absolutely nothing to keep them sane, and lots of little errors to drive them crazy.

--Rik

Post Reply