(I know this wasn't shot with a non-canon lens or non-canon camera, but I felt like this belonged in this thread, due to my question about focal lengths, chip/film sizes, and DOF which is often discussed in here)


Doing some experimenting with odd techniques here. This evenings test went down as follows: Basically, it's an 85mm shot close up (about 1.5 meters away or so) at f1.2, multiple shots (about 24) stitched together so that the entire car is in frame. This seems to be is giving the field of view and perspective distortion of a wider lens (I'd guess around 28mm) but it appears to be keeping the same depth of field of the 85mm at f1.2.

the question: I understand that DOF relates to aperture, focal length AND film back size (a wide shot at F8 on 35mm small format will mostly all look in focus, but on large format will still have a thin DOF), but what exactly is happening here? Is this the effect the same if it were shot with a 28mm f1.2 lens (if this lens existed) on the same 35mm chip? Or is this effect (because I'm stitching these together), more or less simulating what would happen when shooting on a much larger chip (like a large format) with a relative wide lens (85mm) at a more realistic wide aperture? (f5.6 or whatever)

The effect is interesting on the full image (which is over 20000px wide), The jist of it is there on the web-sized image, but some of the 'coolness' (lack of a better word) doesn't seem to fully carry over like I think it will on a good sized print. Can't wait to do another test like this with a better image, and print it out nice and big. Should be an interesting print

Perspective is ONLY a result of your (or the camera's) position relative to the subject. Different focal lengths do not alter perspective.

An 85mm lens is what it is, a 28mm f1.2 would give much greater DOF, so your 2nd theory is correct, it is akin to using an 85mm lens on a larger format - though it would have to be a pretty huge chip!

Weird looking shot btw. It seems to be coming out of the frame, and is somewhat cartoon-like, but I think I quite like it!.

It's a great shot. You should post it on the 3D thread because, boy! it has some.

Very cool.
How were you stitching the 85/1.2 shots at that range without hideous parallax problems?

Very nice result. That shows what you could achieve with an 85mm f1.2 lens on a much larger format.

hubsand wrote:
Very cool.
How were you stitching the 85/1.2 shots at that range without hideous parallax problems?
stitching programs are pretty good now-days, the only problem I had was stitching the blurred areas (the frames that were completely out of focus) So the ones that couldn't stitch I had to place by hand in photoshop. (mainly, two shots in the upper left corner)

For my next test on this, I may bring the panorama setup that I use at work, should make it easier on the stitching side.

Forgive me (and correct me) if wrong:

If you stich so many images (so wide field of view) the effect if similar as if shot with a fish eye and then correcting for linear perspective: you get a curved plane of focus, more front-focused at the edges.

Pere

pere marti wrote:
Forgive me (and correct me) if wrong:

If you stich so many images (so wide field of view) the effect if similar as if shot with a fish eye and then correcting for linear perspective: you get a curved plane of focus, more front-focused at the edges.

Pere

yes, the strip of in-focus area on the ground would be curved, I believe, like an arc around the camera. Next time I do this, I'll make sure I get the in-focus area of the ground in frame, I bet that would be even more weird looking

The stitching programs can stitch in various modes. Some of them preserve rectilinearity.

very interesting!

foto-z wrote:
The stitching programs can stitch in various modes. Some of them preserve rectilinearity.
Yes, I set it to project on a plane, but the image looked to stretched. So I had it project onto a sphere and it looked more natural.

someone in another forum asked to see some close-ups, so here they are:


This is one of the photos, full un-cropped (roughly 16.6%) you can clearly see that the headlight appears to be in focus.

Cool! There's always so many interesting things to try with photogrpahy. One of the reasons I'm drawn to it

here's the version that's getting printed 24x36"

You're shooting very close to your subject with a long lens at a wide aperture, so your DOF is tiny. Assuming you didn't change the focus point or aperture for any of the shots, the DOF of any individual frame doesn't change, so this is true of the overall picture.

With 24 captures from a full frame camera, allowing for some overlap, your total capture size is pretty close to the size of a 4x5 large format frame. An 85mm lens is a very wide angle on 4x5. I have a 90mm lens that is hard to use because it's so wide, in fact (though there are lenses as wide as 47mm for 4x5).

So you've effectively taken a superwide closeup shot using an 85mm lens at f/1.2 (there is no LF lens on earth that opens to f/1.2). The DOF is determined by the aperture/focal length/subject distance, and the angle of view is determined by the format size (in your case, thanks to the stitch, considerably larger than the sensor size).

Counterintuitively, for the same focal length / aperture / subject distance, DOF is actually considerably deeper on larger formats. This is easily demonstrated with a DOF calculator. The reason is that larger formats require less enlargement, so the circle of confusion (and therefore range of acceptable sharpness) is much much larger.

But in the practical world this is pure trivia, because an 85mm lens on 4x5 is a totally different animal than an 85mm lens on 35mm. So for the equivalent angle of view, LF has a shallower DOF, but for the equivalent focal length, LF has a deeper DOF.

I've done some maths. Correct me if I'm wrong.

CoC is somehow arbitrary and relates to the format, the size of the print, and intended viewing distance. In this case the format that should be considered is the full stiched image itself at the sensor scale, so the CoC is about the same than 4x5 format. DOF doesn't depend on focal length. So, for a given format, it should be the same, be it 85mm or 28mm. The curious look of this image is due to the (segmented) curvature of the plane of focus.

Pere

Yes, I agree with you, Pere. He has assumed the optical parameters of roughly a 4x5 capture size, including the CoC and angle of view.

My statement about CoC was parenthetical, or off topic a bit, and was meant only to illustrate how DOF changes according to format size when you hold all other variables constant. But you are correct that his stitched image will assume the CoC of the final stitched image size, not the CoC of the sensor size.

And DOF does depend on focal length, very critically in fact, though this might be be mitigated if you move change your focal distance to restore composition. But if you don't accomodate the difference in focal length with a reciprocal subject distance, the focal length will critically influence the DOF. Just take a look at the equations linked below.

The near and far limits of acceptable sharpness are determined by (among other things) the hyperfocal distance and the focal length. And the hyperfocal distance is directly related to the square of focal length. It's the single only variable in the DOF calculation that is squared, making it mathematically the most important.

http://www.dofmaster.com/equations.html

Just think about a short telephoto lens on 35mm versus 8x10. A 100mm lens on 35mm is roughly the focal length equivalent of a 700mm lens on 8x10. These will produce the same angle of view. With the same aperture (let's say f/16) and same subject distance (let's say 20 feet), your calculated DOF is 12.6 feet for the 100mm lens on 35mm, but 1.33 feet for the 700mm lens on 8x10. This is despite a 6.7-fold larger circle of confusion on 8x10 (0.2mm versus 0.03mm) which strongly biases the 8x10 towards a deeper DOF.

DrPablo wrote:
And DOF does depend on focal length...

You are right, provided the distance from the subject is the same. And no 28mm lens in the world is fast enough to provide such shallow DOF with a single shot of the entire car, nor any 85mm, and when stepping back to get the car inside the frame the perspective would had changed. I was thinking about getting similar results with a 4x5 camera. The f number should be scaled (about 3.5x) to compensate for the larger diameter to get the same aperture (and DOF), so about f4 should be used instead f1.2. And of course the effect would not be the same with a single shot and a rectilinear lens. Is this correct?

I have no idea, but I think I like it

I think if you were to get similar results with a superwide camera, you'd need a lens that approximates the angle of view. If he's used an 85mm focal length lens to stitch together a 4x5-sized image, then that should be the same as using an 85mm lens for a single shot on a 4x5 camera. As far as I know there are no 85mm lenses for 4x5 (there are 72, 75, 80, 90, etc), and there are definitely no large format lenses that open up to f/1.2.

But if there were an 85 f/1.2 lens that could cover 4x5 at this focus distance, I would think the DOF would be exactly the same. I mean the format size/CoC, focal length, aperture, and subject distance would be the same.

I'm not sure how far from th car he was, but for a single shot on 35mm with a 28mm lens he'd need a really wide aperture to achieve a DOF like this one. I mean it would need to be below 1.

The f/number shouldn't need to be scaled if you're using an 85mm lens on 4x5, though. The difference between a 4x5 lens and a 35mm lens is mainly the image plane coverage, but for a given focal length the f/stop and focal length don't change.

There would, as you say, be a difference in field curvature with a single shot rectilinear lens, and I'd think that 28mm on 35mm would be fairly similar to 85 or 90mm on 4x5. There may be other differences, because LF lenses cast an enormous image circle to allow movements, so if you don't use movements at all you have basically flawless lens sharpness at the corners of wide lenses. This is true with some wide angle SF lenses, but the sensor/film is much closer to the edge of the image circle on these formats. The LF lenses have a different amount of field curvature as well, depending on design, because they're often a fairly simple symmetrical or near symmetrical design. On the other hand, the stitched 85mm image on SF would be pretty darn flat if you could avoid parallax errors. The only way to get a flatter field than that would be to use a true flat-field lens (like a macro).

Creo que finalmente nos acordamos!

DrPablo wrote:...I'm not sure how far from th car he was...
about 1.5 meters. you can see where I had the tripod placed in the images on the 2nd page

DrPablo wrote:
...I mean the format size/CoC, focal length, aperture, and subject distance would be the same...

...The f/number shouldn't need to be scaled if you're using an 85mm lens on 4x5, though. The difference between a 4x5 lens and a 35mm lens is mainly the image plane coverage, but for a given focal length the f/stop and focal length don't change....


Aperture and f-number are not the same. Aperture = focal lenght / f-number. So, if focal length is scaled, so is f-number.

I don't know if you're correcting me because you think I really don't know these things, or if you're splitting hairs. But strictly speaking, you've also been imprecise about aperture in what you've just said, because what you've just called aperture is really the diameter of the aperture pupil.

Aperture, strictly speaking, is an area and not a diameter. Because it's the area of the aperture opening that directly determines exposure. Diameter, which is divided into focal length to determine f/stop, is a derivative of aperture area. And this is why f/numbers progress by square roots of 2.

So if you really want to be precise about what aperture means, you should go around the forums telling people to only talk about aperture in mm^2 instead of in f/numbers.

But for practical purposes, the f/number is the single most useful statement of aperture that exists for photographers. It is a measurement of aperture that controls for focal length, making it universally applicable to all lenses. I can't think of any situation other than building a lens or a pinhole camera in which it's practically useful to think about the diameter or area of the aperture pupil.




Edited by DrPablo on Jun 13, 2007 at 02:07 PM GMT

DrPablo wrote:
...The LF lenses have a different amount of field curvature as well, depending on design, because they're often a fairly simple symmetrical or near symmetrical design. On the other hand, the stitched 85mm image on SF would be pretty darn flat if you could avoid parallax errors...

Curvature of field in lenses that are suposed to be rectilinear is a defect. But it is not what I was talking about. Imagine 3 shot are taken in a raw in order to stitch them. Their planes of focus are not on the same plane. The ones on the sides are slanted inwards. This was what I meant when I said (segmented) curvature of plane of focus.

His images were not necessarily slanted inward. If you use a shift lens, or if you use a bracket that allows parallel shifting, there is no rotation about a central nodal point. Thus, the plane of focus can remain unchanged depending on technique.

pere marti wrote:
Curvature of field in lenses that are suposed to be rectilinear is a defect.

It's a lot more complicated than that, and I don't think that's a fair generalization. Very few rectilinear lenses are corrected to be perfectly flat at all focus distances, and that is by design. Curvature of field tends to be only completely corrected when close to infinity focus. The only lenses that are actually designed to have truly flat fields of view are copy lenses and some higher end macro lenses -- but their field is only perfectly flat at very close focusing distances. And this is most likely to be the case in lenses with symmetrical designs, of which there really aren't all that many.