Calculate this

[Ben Rubinstein]

Hi guys,

If I want to achieve the FOV of a 50mm lens on a FF DSLR by using 36 frames total then what focal length do I need? Is there a calculator somewhere that can do this kind of thing for me? I'm trying to replicate exactly the tonality of 4X5. Now on a 4X5 to achieve that angle of view I'd be looking at around 150mm but not sure if it's as simple as that for stitching.

Thanks!

 

[Doug Kerr]

Hi, Ben,

If I want to achieve the FOV of a 50mm lens on a FF DSLR by using 36 frames total then what focal length do I need? Is there a calculator somewhere that can do this kind of thing for me? I'm trying to replicate exactly the tonality of 4X5. Now on a 4X5 to achieve that angle of view I'd be looking at around 150mm but not sure if it's as simple as that for stitching.

I'm not sure I understand the "setup" from your description. Can you elucidate a bit more.

Are you perhaps saying that you want to make a composite image of 6 x 6 frames taken with a camera with a full-frame 35-mm size sensor (equivalent to a format size 216 mm X 144 mm), and want it to have the same field of view as a single frame taken on with a 50 mm lens on a camera with a full-frame 35-mm size sensor?

If that's the case, then approximately, you would need to take your 36 frames with a 300 mm lens (so each frame covers 1/6 the height and width that a 50 mm lens would cover on the camera).

Thanks.

Best regards,

Doug

 

[Murray Foote]

Just a quick response. Doug will probably come in with much more precision.

36 exposures will be a grid 6x6. A 50mm FX lens will have an angle of view of about 40 degrees (horizontally). Assuming 30% overlap, you get one-sixth of 40 degrees multiplied by 1.3 which gives you an angle of view of 8.67 degrees. That makes it somewhere round 250mm (200mm is 10.3deg; 300mm is 6.9).

p.s. Doug has already replied.

 

[Ben Rubinstein]

I want to replicate 4X5 film by providing the same amount of sensor real estate (not for resolution but for tonality). When I overlay a 4X5 sized picture with 35mm pictures and include a 25% overlap then I get approximately 36 frames with sufficient overlap over the edges.

Now I'm confused. The FOV of a 50mm lens on a FF sensor is approximately 40 degrees. Using a LF camera the FOV of a 150mm lens is also approximately 40 degrees. It then follows that if I use a 150mm lens on a FF camera and stitch to cover the same image a 50mm lens of FF would provide then I have sucessfully replicated 4X5 imaging real estate. Calculators such as this: http://www.frankvanderpol.nl/fov_pan_calc.htm seem to bear my calculations out at approx 36 frames if I'm giving overlap on the edges to allow me the full 40 degree FOV after vanishing point corrections, etc.

Am I thinking this through wrongly?

 

[Ben Rubinstein]

I don't have a lens longer than 85mm at present but I thought I'd try the same principles. 85mm lens in 4X5 terms is approximately a 28mm lens on my 5D. I framed with a 28mm lens. I then shot 6 frames in either direction with the 85mm varying the angle to suit the camera orientation but computed from the 65 degree angle FOV. This was without even looking at whether I was matching the FOV on the edges. I came out with a frame which was just a bit wider than 28mm, i.e. enough for keystoning correction and the equivelent of 28mm plus a bit on each side, exactly as I'd planned it. Just making up the pano now and I'll compare relative resolution.

 

[Asher Kelman]

Ben,

Are you intending to refocus for each shot. If not your plane of focus is convex and then the software has to make this rectilinear so the angles change.

Doug and Bart,

What do you think of that?

Asher

 

[Ben Rubinstein]

No I won't be refocusing for this project if I can get it off the ground, though modern software copes with this very easily, all my 6X12 stitching work is refocused frame to frame and I've never had a problem.

 

[Ben Rubinstein]

And the results are in!

Basically it worked! If I make a 4X5 frame with the same resolution as my sensor (5X4" @ 3081DPI) the final stitch pretty much reaches the edges. It's about 5% off but it's certainly close enough. A few things I learnt, I need more overlap over the edges, at least another frame each way to allow keystoning correction and 'oh woops I screwed up' correction, plus if I want a 4X5 ratio then I need something better than my camera for initial framing . Alain Briot has a nice little app for the iphone that does just this actually.

So to sum up, if I want the same sensor real estate as a 4X5 negative via stitching then I have to use the same lens I would use on a 4X5 camera for a chosen FOV and stitch accordingly. Voila!

Now someone tell me I've got it all wrong!

 

[Doug Kerr]

Hi, Asher,

Are you intending to refocus for each shot. If not your plane of focus is convex and then the software has to make this rectilinear so the angles change.

Doug and Bart,

What do you think of that?

Quite* (that's why I said "approximately").

*However, whether or not one refocuses is not an actual part of the issue. Field of view is an angular measure, and does not presume focus on any particular plane or curved surface.

A small complication though is that for any given focal length, the field of view depends slightly on the focus state.

Imagine that we have a 300 mm lens on a ff35 sensor camera, and that focus is "distant". Then the horizontal field of view would be 6.87° (assuming a rectilinear lens).

Now imagine that we take 6 shots, allowing no overlap. Then the angular field of view of the whole operation would be 41.2°.

Again assuming "distant" focus, to get that horizontal field of view in a single shot with the same format size would require (again for a rectilinear lens) a focal length of 47.9 mm.

Working it the other way, if we start with a single frame with a focal length of 50 mm, (and "distant" focus), the horizontal field of view would be 39.6°. If we divided this into six frames, the horizontal field of view of each would have to be 6.60°. Still assuming a rectilinear lens (and "distant" focus), this would require a focal length of about 312 mm.

Best regards,

Doug

 

[Asher Kelman]

Ben,

I remember William Harvey's advice, "Don't think, try!". Here, I'd start with the 200 mm focal length 25% overlap and see how this goes. Compare with a lens that gives a ~ 40 degree angle of view in 35 mm. Your 50mm gives 46 degrees.

So I'd just try. What other lenses do you have and why not do 8x6 matrix if you have a clean workflow?

Asher

 

[Ben Rubinstein]

Asher, did you read my posts above? Tried, worked, now confused about Doug's math as my method works for some reason.

 

[Ben Rubinstein]

OK here's my results, forget the stitching error and the keystoning, not relevant here.

This is 25 frames stitched. It would of course be more if I was stitching for a 4:5 ratio rather than the 2:3 ratio here and I've left very little extra room round the edges (will have to be careful of that). The black frame is what a 5X4 sensor would look like if it had the same resolution of my 5D. As you can see I've approximated the sensor real estate of a 5X4 sheet of film with this stitch. The FOV is that of a 28mm lens using my theory that stitching with the 5X4 equivelent focal length (here 85mm) provides the correct amount of frames to approximate a 5X4 size sensor for a given resolution.

 

[Asher Kelman]

And the results are in!

Basically it worked! If I make a 4X5 frame with the same resolution as my sensor (5X4" @ 3081DPI) the final stitch pretty much reaches the edges. It's about 5% off but it's certainly close enough. A few things I learnt, I need more overlap over the edges, at least another frame each way to allow keystoning correction and 'oh woops I screwed up' correction, plus if I want a 4X5 ratio then I need something better than my camera for initial framing . Alain Briot has a nice little app for the iphone that does just this actually.

So to sum up, if I want the same sensor real estate as a 4X5 negative via stitching then I have to use the same lens I would use on a 4X5 camera for a chosen FOV and stitch accordingly. Voila!

Now someone tell me I've got it all wrong!

Ben,

I had missed where you actually did it! Comes from trying to cover many threads. When one uses an 85 mm lens in 4x5 one can put a 5DII on the back and then sample all over the image plane and assemble afterwards. So to do the same thing with your 5DII directly and get the same perspective, you are using an 85mm lens with the same central field. With rotating through the entrance pupil, you are essentially building a virtual LF 85 mm lens, but with a curved focal plane. Your intuition to realize that an 85 mm 5DII lens was the lens to choose for this work is simple and brilliant. Now that you have shared this, it's so obvious. Still, I wish I'd come up with that myself!

Asher

 

[Doug Kerr]

Hi, Ben,

Tried, worked, now confused about Doug's math as my method works for some reason.

Note that my math was (essentially) for a "large" frame of 216 mm x 144 mm (8.5" x 5.7"), since you spoke of 36 tiles and I assumed you meant 6 (h) x 6 (v); I made no allowance for overlap.

That's why I asked for a more complete description at the outset!

Probably if I worked with your actual "4x5" frame size (whatever that is exactly) I would get your result.

Best regards,

Doug

 

[Ben Rubinstein]

Asher, I actually owned a CameraFusion LF adaptor for a while, it's what first 'turned me on' to the 6X12 format, I was using it with a Gandolfi camera. The main problems were focusing and composition. Sperical stitching is of course far quicker, easier, far lighter and of course cheaper!

Sorry Doug, I didn't provide all the details until my follow up post after your request for more information.

 

[Jerome Marot]

So to sum up, if I want the same sensor real estate as a 4X5 negative via stitching then I have to use the same lens I would use on a 4X5 camera for a chosen FOV and stitch accordingly.

Exactly. There are no maths involved and it is independent on the amount of recovery to stitch frames. If you want the exact same effect of a X mm lens at f/Y by stitching, use a Xmm lens at f/Y.

 

[Ben Rubinstein]

Going out with a pal tomorrow to try out the concept, he has a 70-200L so I can try the concept of stitching with the 150mm focal length. I'm going to shoot a scene similar to that which I want to use the effect for and compare it to shooting with my 50mm prime at various apertures so that I can compare the tonality and focus fall off. I'll report back! If it all works as I plan then tomorrow when I've seen the results I'll buy a Mamiya 150mm f3.5N lens (I have a mamiya-eos adaptor that I picked up years ago) to use for the project, it's small, incredibly cheap (especially as I'm incredibly poor at the moment!) and will give a softer rendering being a MF lens than using the usual ultra sharp 35mm lenses. Tonality is crucial here.

Let's hope that tomorrow will start a new chapter of my 'Dream Project' with a different look though still working towards the same goal that I originally bought my lensbaby for. An Ethereal take on my surroundings!

 

[Ben Rubinstein]

OK problem, big problem, how the heck do you stitch OOF frames together?

 

[Murray Foote]

Possibly manual joins in AutoPano Pro if you can work out what corresponds. I was able to get away with manual water-to-water joins at the bottom right of the pano in this post (click on third image)

 

[Ben Rubinstein]

Yeah having to do that though Autopano Pro seems to be much better with OOF sections of a stitch than PTGUI. My main problem however at the moment is that I screwed up with the stitch, didn't leave enough overlap. Hopefully I'll have enough to be able to decide whether on not my idea works in the real world though!

 

[Bart_van_der_Wolf]

OK problem, big problem, how the heck do you stitch OOF frames together?

Hi Ben,

It depends on the amount of OOF. A good stitcher can use several strategies to determine the best fit, but there has to be at least some level of (even OOF) detail.

The better automatic routines use a resolution independent/insensitive method by first aligning downsampled versions of the files. The downsampling will allow to reach a higher per pixel resolution, which may be close enough to produce a match at the original resolution.

An alternative method (borrowed from HDR stacking) could also be used, where one determines the (histogram) median values in the overlap region of the tiles, and produces proxy image versions with a threshold at those median values. Then align the proxies, and copy the settings to the original images.

You may also have to increase the overlap region, to include some detail that can assist in the registration. Sometimes you can stitch rows first, and then stack entire rows with enough detail somewhere along the row to find a match between them.

When you have too few clues in the overlap region, manual placement remains the final option, and there a clickstop rotator shows one of its benefits, because all steps are close to identical in degrees of rotation.

Cheers,
Bart

 

[Ben Rubinstein]

Well I'm still trying to make the stitch work, because I'm an idiot I did my math wrong and...

The idea was to take one image with my 5D and 50mm at f1.4 then shoot the identical image with a 150mm lens at f3.5 and then compare tonality and focus fall off. While the stitch is still being done, here is the original frame and to be honest, for all that it's one of Canon's cheapest lenses, at f1.4 and as long as you aren't dealing with specular highlights, it's actually got a very nice soft rendition drawing with just a very tiny weeny hint of chroma. Even if I don't get the stitch to work or don't like the look, this picture is a keeper in my book!

Just outside the Psalms Museum in a century old courtyard little changed, Jerusalem.

 

[Ben Rubinstein]

Bart, I screwed up though I hadn't realised it at the time, I not only didn't leave enough overlap but actually didn't leave any at all for one of the rows which is what screwed it up so badly.

 

[Bart_van_der_Wolf]

Well I'm still trying to make the stitch work, because I'm an idiot I did my math wrong and...

The idea was to take one image with my 5D and 50mm at f1.4 then shoot the identical image with a 150mm lens at f3.5 and then compare tonality and focus fall off.

Hi Ben,

The focus fall-off is approx. identical if you use proportional focal length and F-stop numbers. So the difference between a 50mm and a 150mm lens is 3x, therefore the effect of f/1.4 on the shorter FL versus f/3.5 on the longer FL should be rather similar. It won't be eactly the same, due to optical design differences, but close enough in practice.

While the stitch is still being done, here is the original frame and to be honest, for all that it's one of Canon's cheapest lenses, at f1.4 and as long as you aren't dealing with specular highlights, it's actually got a very nice soft rendition drawing with just a very tiny weeny hint of chroma. Even if I don't get the stitch to work or don't like the look, this picture is a keeper in my book!

If this is the amount of defocus that causes you stitching problems, then I suggest to first stitch the in focus tiles area. Then prevent adjustment to the found parameters for those tiles, and then add the OOF image tiles one by one. A click-stop rotator is very useful for such scenarios, also for stitching featureless (but with a gradient) blue skies.

Cheers,
Bart

 

[Bart_van_der_Wolf]

Bart, I screwed up though I hadn't realised it at the time, I not only didn't leave enough overlap but actually didn't leave any at all for one of the rows which is what screwed it up so badly.

As they say, s**t happens...

Cheers,
Bart

 

[Ben Rubinstein]

Not sure I agree about the focus fall off, you can take a short lens and a long lens and play the f-stop numbers until the DOF is the same but the 'look' of the focus fall off is very different. At least in my experience.

 

[Cem_Usakligil]

Hi Ben,

I want to replicate 4X5 film by providing the same amount of sensor real estate (not for resolution but for tonality). ...

...The idea was to take one image with my 5D and 50mm at f1.4 then shoot the identical image with a 150mm lens at f3.5 and then compare tonality and focus fall off. ....

From your messages, it is quite clear that you are doing this to replicate the tonality of the 4x5 format. Can you please help me to understand what precisely do you mean by tonality in this context? I admit that I am fascinated by this discussion but I also want to make sure that I understand it correctly.

 

[Bart_van_der_Wolf]

Not sure I agree about the focus fall off, you can take a short lens and a long lens and play the f-stop numbers until the DOF is the same but the 'look' of the focus fall off is very different. At least in my experience.

Okay, if you are referring to what is known as Bokeh (the 'quality' of defocus), then the optical design is responsible for that. It will be hard to match.

Cheers,
Bart

 

[Ben Rubinstein]

It was hard enough explaining this to the guy who was tagging along with me for the shoot. Bascially I want to make available more sensor real estate for describing the transition between tones. The more tonal information available to the final image the smoother the tonal rendition, the transistion between light and dark, will become. I described this to him as if the camera sees a blob then it can say that the blob is 70% grey. If you have a good camera then it may even be able to describe it as 72.5% grey. If you however can dedicate more pixels to the blob then it's no longer a blob but an area with many distinct levels of grey progressing through it.

Something like that anyway, harder to explain than to see even a cheap scan from a 4X5 neg and have the tonality hit you in a truly 'wow' moment.

 

[Ben Rubinstein]

Okay, if you are referring to what is known as Bokeh (the 'quality' of defocus), then the optical design is responsible for that. It will be hard to match.

Cheers,
Bart

It's not even that. It's how the transistion from in focus to out of focus is drawn. A 400mm lens and a 50mm lens, even with identical numerical DOF will have a very different look to the focus fall off. The shorter lens at a wide aperture will show very little in focus with an immediate fall off, a longer lens stopped down more will have the same DOF but the focus fall off is far more abrubt, cut off than the shorter lens.