What is the Best Way to Make a 4x10 Image From Two 4x5 Images?

[r.e.]

I raise the question here because the answer starts from the camera as a piece of hardware.

I am interested in doing this because the 4x5 camera and 360mm lens that I have will result in the composition that I want if two frames can be combined. I don't want to buy a 4x10 camera and I want to know whether this is a good or bad alternative to buying a 600mm lens for my 8x10.

Thanks.

[wfwhitaker]

Have you considered simply doing a diptych? (Pan the camera between shots, of course.) Diptychs and triptychs can be very effective, if not overdone, and the hardware required beyond what you already have is nil.

[r.e.]

Have you considered simply doing a diptych?

Thanks, yes I have considered it.

[Ken Lee]

I plan to try some 5x8 shots. If things work out, I'll show an example. It can't always work, but it's cheaper than a 5x8 camera and holders.

[cjbroadbent]

A diptych is like a double-page spread with a gutter down the middle - harder to compose than a triptych where the wings are there to add peripheral vision to a main subject and the gutters do not intrude.
Scanned 4x5s can (with a little hassle) be stitched together with Adobe Stitcher, just take care of the nodal point with the tripod and the focal length setting with the software. With two shots in landscape mode you would get 4x7, three shots, 4x10.5.
So it might be better to go vertical mode for three shots to get 5x9.

[Frank Petronio]

The Nodal point concerns lessen the closer to infinity focus you get. I suspect longer lenses lessen the concern as well. In practical application it depends on how detailed and precise you want to be -- for most landscapes all that matters is that the horizon lines meet seamlessly.

[Ken Lee]

"With two shots in landscape mode you would get 4x7, three shots, 4x10.5. So it might be better to go vertical mode for three shots to get 5x9."

OK - Thanks !

[Ed Richards]

You need to shoot three sheets, not two - the more overlap, the better the job the software does. Nodal points really do not matter with modern software, esp. for long lenses - assuming, as Frank points out, you are close to infinity. Level really matters for the rotation, and you probably do not want to use a polarizer. The best way is to use the index on your pan head and work out the position for each shot ahead of time. Do not crop tightly - leave a little extra at each end to give some latitude in stitching. Then you can just stop down and start shooting, rotating the camera to the next index mark for each shot.

[r.e.]

Thanks, these responses are very helpful. I'll be using a 360mm lens at infinity.

People who, like me, are completely new to this might find it useful to have a look at a paper by Doug Kerr, who used to participate on this site, called The Proper Pivot Point for Panoramic Photography: http://doug.kerr.home.att.net/pumpkin/Pivot_Point.pdf

I'm not sure how to square what he says about parallax shift with references that I've seen to using shift rather than panning around the nodal point (or, as Mr. Kerr prefers, the entrance pupil). He does say: "In reality, unless there are objects that lie quite close to the camera, the parallax shift caused by even a substantial error in selecting the position of the panoramic pivot axis will be negligible". If one is using a long lens on a distant subject, such as a town, are there any advantages to moving the camera laterally on a rail (basically, more capacity for what amounts to front shift) instead of panning?

[jb7]

Wouldn't that lens cover 4x10?
You could just shift the back, and take two pictures within the same image circle?
Same nodal point then, no messing about...

[jb7]

It will, of course, be a different picture-
I can't see how you can compare a 4x10 using a 360mm to a crop from 8x10 using a 600-
the smaller picture is going to need some enlargement anyway-
it's going to be a 6" wide picture, according to what you've been talking about on the other thread-

Edit- if your lens is a tele, you're back to rotating around the nodal point-
but if your Arca rail is long enough, you might be able to position the camera all the way back,
with the nodal point over the tripod head
and use a monopod for extra support...

[Greg Miller]

You definitely need more than 2 images to make a 4x10. This is because you need some overlap in order to do the stitching; meaning with overlap you will end up with something closer to 4x8.

I have done thousands of stitched panoramas, and for landscapes I have worried about parallax less than 10 times. I do carry an inexpensive rail with which I can position the camera (lens) correctly for occasions where parallax will be a meaningful concern. But for landscape work it rarely comes into play.

I would agree with Kerr that parallax is a practical concern only when you have objects very near the camera. But this rarely happens because including both nearby and far objects with a wide angle lens will also distort the dimensions of the nearby object yielding some pretty strange results.

If you have access to a DSLR, I would recommend doing some (free) practice panoramas to see for yourself if Parallax is meaningful to you. And doing some practice scenarios will help you dial in the in the field process and stitching process. Then when you have a scene that you really care about, your odds of success will be higher.

[Greg Miller]

I'm not sure how to square what he says about parallax shift with references that I've seen to using shift rather than panning around the nodal point (or, as Mr. Kerr prefers, the entrance pupil). He does say: "In reality, unless there are objects that lie quite close to the camera, the parallax shift caused by even a substantial error in selecting the position of the panoramic pivot axis will be negligible". If one is using a long lens on a distant subject, such as a town, are there any advantages to moving the camera laterally on a rail (basically, more capacity for what amounts to front shift) instead of panning?

Any easy way to experiment is to play around with the human eye. Close one eye and hold your right pointer finger (close to your nose) and left pointer finger (at full arms length) vertically and in line with each other in front of your nose. Then pivot your head and watch to see if the fingers stay in line. The closer you position the near finger to your nose, the more parallax you will notice when you pivot your head.

This works because your eye is not pivoting directly above your neck, but at a point in front of the pivot point; much like a lens positioned in front of the pivot point on a tripod.

[r.e.]

Joseph, yes the lens covers 4x10 with room to spare. The image circle at f/22 is 494mm. The rear standard of the camera allows 5cm/2" of shift to each of the left and right.

Greg, your suggestion to experiment with a DSLR is spot on. That is a neat human eye experiment - makes some things crystal clear.

I'm not wedded to precisely 4x10. It can be a bit less or a bit more provided that in the end I have about one inch of height to about 2.5 inches of width.

I gather that Christopher's suggestion of shooting vertically has to do with giving the software more edge to edge real estate to work with. His numbers seem to recommend a fair bit of subject overlap. Joseph, where does your number 6" come from?

I'm quite speeded up about trying this; and if it works, it could save me both a significant amount of money and reduce - maybe not eliminate, but reduce - concerns about lens movement/vibration.

I'm also interested in adding into the mix Christoper's ideas about mounting the camera on a leveling base with panoramic plate instead of on a regular head.

[jb7]

6"?

It's the relative magnification-

On the other thread, you mentioned that the picture you wanted required a 600mm lens on 8x10.

Using a 360mm lens would give you the same picture at 6" wide,
if my mental arithmetic is correct-

A 360mm lens at 10" wide would be a much wider shot-

[r.e.]

Got it.

[AJSJones]

Joseph, yes the lens covers 4x10 with room to spare. The image circle at f/22 is 494mm. The rear standard of the camera allows 5cm/2" of shift to each of the left and right.

Greg, your suggestion to experiment with a DSLR is spot on. That is a neat human eye experiment - makes some things crystal clear.

I'm not wedded to precisely 4x10. It can be a bit less or a bit more provided that in the end I have about one inch of height to about 2.5 inches of width.



One of the factors I considered when buying my 4x5 was back-cross. My Ebony has more than your camera but even with yours you can get a 4x9 (2 + 5 + 2) chunk of the image circle by using the extremes of the movement. (3.6 x 9 has the 2.5 aspect you seek) This will provide plenty of overlap to paste one onto the other in PS. No stitching required, no parallax issues. If the exposures are right and the lighting consistent and the scanning protocols are identical, the limiting factor is the computer! The image below is from Lundy Canyon and each half was 600MB in 48 bit from my AgfaT2500, so the pasting was slow - the slowest part was the rotation tweak to get perfect alignment but the measure tool made it a simple operation, if slow :( The final image is 9,230 x 21,450 for 560 odd MB. I plan to upgrade from my Dual2GHz G5 to the next MacPro soon (I hope) and I expect the operation to go about 10x faster then :)

http://www.fototime.com/3577A692082C97D/standard.jpg

[r.e.]

So at 300 dpi that will produce a 2.5' x 6' print. That is something that I'd like to see.

Now I'm trying to figure out the difference between stitching and pasting. Is stitching indicated only if the camera is panned in an arc rather than shifted laterally (e.g. by shifting the film plane)? I've come across one or two posts, including one by Sandy King if I recall, that refer to "manual stitching". Is manual stitching actually pasting?

[domaz]

I don't think there is a difference between stitching and pasting. All stitching programs do is find overlap in 2 or more images and use that overlap to determine where to put the images in relation to each other correct? I'm sure some do fancy things like distortion correction etc.. but probably not really necessary.

[AJSJones]

r.e. I have a 20x44 hanging on the wall from my 7600 - mmm nice :D

The extent to which the various images need to be manipulated to create a seamless composite varies. If the lens moves between shots, it gets more complex (my 4x10 or flat stitching from DSLR shift lenses need little help). Long lens shots panned simply, will often need little help. Multi row multi image requires correction of perspective distortions from the body tilt etc.

domaz
I made the distinction because the good stitchers will remove lens distortion and perspective distortion and do various projections and will distort overlapping frames to maximize overlap and will ensure colors and tones are equalized , generate masks and blend layers etc etc etc - they are very sophisticated and complex. In my case, it was two samples from a fixed image projected by the lens (Nikon 210 mm), so only pasting required since there was no distortion introduced.

[Greg Miller]

Is manual stitching actually pasting?

In my mind, pasting generally refers to simply aligning layers in Photoshop. This generally applies when scanning one piece of film in sections, because there are no issues with lens distortion or lunimosity or color balance issues between the layers. Manually stitching takes it a step further and aligns and blends the overlapping images by using tools layers masks, transforms, and adjustment layers.

[r.e.]

I'd like to express my thanks to everyone who has contributed to this thread. The comments are really helpful in getting a handle on this subject.

[venchka]

My first efforts were thwarted by framing too tightly. Old habit. I hate to crop. Stitching software robs you of image top and bottom. Allow extra space. Another reason to work out the shooting with digital. If you don't have Photoshop, Microsoft offers a free program ICE.

http://research.microsoft.com/en-us/um/redmond/groups/ivm/ice/

[walter23]

I raise the question here because the answer starts from the camera as a piece of hardware.

I am interested in doing this because the 4x5 camera and 360mm lens that I have will result in the composition that I want if two frames can be combined. I don't want to buy a 4x10 camera and I want to know whether this is a good or bad alternative to buying a 600mm lens for my 8x10.

Sorry if I'm stating the obvoius and probably being a bit pendantic too, but from the point of view of lenses, if you're talking about using a 600mm lens on 8x10 and cutting down to 4x10 vs. stitching two 4x5 frames at 360mm, the field if view is still going to be different... you're using the same film area in both cases, so you'd need the 600mm lens even if stitching. Alternatively you could cut down 4x5 using your shorter lens, then you do get the same field of view (300mm on 2x5 vs. 600mm on 4x10).

As for combining them I think the obvious answer is to scan and merge, but maybe there are analogue techniques that will work.

[r.e.]

Thanks, my choices are not entirely dictated by arithmetic and I can live with the discrepancy. I happen to own a 360mm. If I go with an 8x10 and a longer lens, 600mm, especially a telephoto design, is for reasons relating to rail and bellows length and general hassle, something I'll live with, whereas an 800mm, which is the next available focal length, is not. On a low lying peninsula open to the North Atlantic, about 7 meters/20 feet above sea-level, wind is something of an issue :)

I'm going to give both pasting and stitching a go and see how they work/differ. I want to try these techniques anyway.