When you take a picture of someone and the background has that converging verticals look (but obviously the main subject is the person), like this photo:
do you usually correct/lessen the problem by using the Transform/Perspective tool of Photoshop?
(I'm a photoshop beginner so don't expect too much from the post-processing above)
[This message has been edited by viol (edited 03-13-2004).]
hm I dont see whats wrong with the first photo although the second one looks a lot better. I am guessing the "convergence" you speak of comes from the wide lens you're using (unless you mean something else), and aside from color correction the picture looks simply cropped. I crop my pictures whenever I think it necessary, but never treated "fish-eyeness" as a shortcoming, but rather a quality of the lens.
To me, it seems that, since you lose your field of view, transforming a photo in such a way negates the point of having used a wide-angle lens in the first place.
Though I should say that I typically use a wide-angle lens deliberately for such an effect. If correcting the "distortion," then cropping the image down to focus on the subject is what you're looking for, there's certainly nothing wrong with it.
[This message has been edited by Wes (edited 03-13-2004).]
Although the picture was taken using focal length of 24mm (equivalent, in my camera, to 38mm for a 35mm camera), my understanding is that the perspective problem that I'm talking about is not mostly caused due to the wide angle used but due to the fact that the picture was shot while pointing the lens a little upward. Because I wanted the name of the school to complete show up in the picture, I positioned the camera some degrees up, and this causes the vertical lines of the building to try to converge in some point up above (the same effect happens to the horizontal lines that try to converge in some point to the right - both convergence points are far outside the picture bounds).
Usually, when we take a picture of a tall building, we point the camera up, and this causes the building to have a larger base, compared to the uppermost part. It's the effect that professional photographers eliminate optically by using those "shift" lens.
The opinion I wanted is if you think it's worth correcting such effect for pictures where the building is not the main subject, it's just a background. Correcting this effect may introduce distortions to the main subject, so, there is a trade. In the picture above, I corrected the width of my daughter, to match the original, once the perspective problem was minimized.
Such a thing needs to be determined on a case by case basis.
If the perspective is so visually apparent that it distracts from the foreground - then yes.
If the correction of the sharp perspective causes obvious distortions on the foreground - then absolutely not.
In the examples you provide, it's tougher to say because the color correction on the second image is a very significant factor in making it look better than the first.
On the example you mention in your second post of an upward shot of a tall building - I wouldn't see reducing the effects of depth via perspective as "correcting" it. You expect, based on a lifetime's experience of simply seeing tings, that the further away something gets the smaller it will appear.
If you eliminate that basic visual cue, I can't se any possible benefit other than in the field of abstract art...
[This message has been edited by DL-44 (edited 03-13-2004).]
Sorry the misunderstanding but the idea was not to compare one sample with the other. I showed both so to illustrate what I was talking about. The first shot is the original shot, no processing, just resized to 550 px wide. The second is a cut, with some processing (well, the best I could do), originally a wallpaper 1024px wide, resized to 550 px wide to show it here. It's really unfair to compare both, at least imo.
"if you think it's worth correcting such effect for pictures where the building is not the main subject"
I'm pretty much with DL-44 here. I'd say no, as a general rule the benefits aren't worth the effort. We are accustomed to seeing the convergence and interpreting what it means. However ... that's not to say don't do it if it bothers you in a picture. My initial reaction to the two images was that the tighter cropping in the lower image improved it more than "correcting" the perspective effect. But then I thought with the tighter cropping verticals that weren't vertical might be distracting at the edges so - I would say no you shouldn't feel you need to straighten verticals as a general rule and yes you should feel free to when you want to!
quote:On the example you mention in your second post of an upward shot of a tall building - I wouldn't see reducing the effects of depth via perspective as "correcting" it. You expect, based on a lifetime's experience of simply seeing tings, that the further away something gets the smaller it will appear.
Opinions differ on this, though. If there wasn't a serious demand for perspective corrected architectural photos, no one would sell tilt/shift lenses.
I personally prefer the tilt/shift version (in the middle), because the buildings look more natural to me. I walk downtown at lunchtime in Houston with 20-50 story buildings all around, and I know my eye is seeing something more like the left shot (uncorrected). But I think the brain does that trick where you seem to see more correctly than your eyes can sense. When I attempt to remember what the Enron building looks like, for example, it seems perspective corrected in my memory.
To my taste, uncorrected shots like the left one are much less pleasing to the eye than a perspective corrected one (either tilt/shift or Photoshopped). The uncorrected ones look 'wrong', regardless of what my eye may see in real life.
Well, in those cases you're not talking about an "upward shot of a building". You're talking about a shot of a variety of objects, including some tall buildings, in which the perspective is skewed between the varied objects for a possible variety of reasons.
That's a *huge* difference from an "upward shot of a building".
The example was mixed, but it was the only before-after I could find on the web.
The type of shots I was thinking of are cathedral photos from Europe, where there's just the one building, and the shot is upwards from in front of it (usually from pretty close, as there's no way to get a clean shot from a distance). I've seen those both natural and perspective-corrected, and I vastly prefer the perspective-corrected ones.
Here's a perspective-corrected shot of the type I was thinking of: http://c-weng.com/05photo.htm
The from-the-front shot would show severe perspective distortion at that range and angle. I've seen such distorted cathedral shots in snapshots taken by tourists, and didn't like them at all.
This guy apparently used Photoshop to correct it, but traditional film-based photographers would have used a tilt-shift lens. I've read several books that strongly recommend getting tilt/shift lenses if you're really into archetectural photography.
I've even read of photographers forgetting to bring a tilt/shift, and taking a shot dead level from way far away, so the subject is only in the top half of the shot. They have to crop off nearly half of the bottom of the photo, but they avoid the up-angle perspective.
Of course, tastes vary.
[This message has been edited by Das (edited 03-13-2004).]
From: The northeast portion of the 30th star Insane since: Mar 2001
posted 03-14-2004 05:30
If I'm not mistaken T/S lenses have other advantages/shortcomings related to the plane of focus/dof as well. I know view cams do, probably to a greater extent. I think the plane of focus remains parallel with the lens (front element on a t/s?) while the perspective is relative to the film plane? It's been a long time, I need to dig out some books.
Yeah, everything I mentioned could be done with a lens that just shifts, but I always hear about tilt/shift lenses, even if they're only talking about the shift part.
You're right, the tilt part does something cool with the DOF plane, but I don't recall the details.
[This message has been edited by Das (edited 03-14-2004).]
I guess if your'e "getting into" architectural photography a tilt / shift lens would be useful.
If you ARE an architectural photographer, that would be a novelty or a backup. The view camera is their supreme tool. It's a pain now with digital so pervasive - there aren't any 4x5 instant capture backs. Scan backs are a pain and view cameras for tiny instant capture backs are a pain too.
Anyway - a view camera has a flexible lens standard and film standard. The basic rule is, the film plane controls the geometry (if it's parallel to the subject, all lines will be parallel), the lens plane determines focus. If the camera is at an angle to a planar subject, you tilt the lens such that imaginary lines drawn through the subject plane, the lens plane and the film plane all meet at some imaginary point in the distance.
Right. And you look at the image upside down too.
That could explain a LOT bout me....
But viol's question had to do with photos in which the building was NOT the primary subject, so I'm sticking to what I said!
[This message has been edited by Steve (edited 03-14-2004).]
Despite that fact that it's a natural physical phenomenon that things closer to us should look bigger and things far from us should be smaller, as said, our brain seems to have some sort of defense against this because when I look to a tall building, looking up, I don't see this physical phenomenon as clearly as I can see in a regular photo (taken without shift lens, aiming upward). I'm very used to this fact because I was a fan of first-person shooter games like Doom, Doom II, Quake, NOLF and more recently Mohaa.
Well, where was I? Yes, despite all that, I prefer, for pictures where the building is the main subject, that that effect be corrected. It's annoying to me to look at those converging lines because in real life they are not (in some cases they may look nice though). When it's not the main subject, as in my daughter's photo, and that was my question, I still prefer the corrected version, as long as it doesn't introduce distortions to the main subject, noticeable distortions.
Anyway, it's sort of weird because the converging vertical lines annoys me but the converging horizontal lines, in the same picture, do not. I even like it.
So, it seems it's really a personnal taste and very dependent on the context.
"it's sort of weird because the converging vertical lines annoys me but the converging horizontal lines, in the same picture, do not."
ahhh - you beat me to the point I was going to make.
My neighbor is an artist. She feels she has found a flaw in Renaissance perspective theory. As an example, if you imagine you are standing facing a wall or fence that extends as far as you can see in both directions. If you were to DRAW this, the wall could have to be widest directly in front of you and taper to a vanishing point in both directions but this is obviously not what it LOOKS like.
I tried to argue with here that our eyes don't see panoramically, but then I gave up.
Guess bottom line is - do what you need to do to please yourself. (or your customer)
quote:Anyway - a view camera has a flexible lens standard and film standard.
I've never used anything bigger than a 35mm SLR, so I'm curious about the high-end stuff. What can you do with the view camera that the tilt/shift can't do? I get the idea that the tilt/shift lens is a poor man's view camera, but I'm curious how much you lose.
quote:I still prefer the corrected version, as long as it doesn't introduce distortions to the main subject
Can it, though? I assumed that correcting the highly visible perspective distortions in the tall building would correct trivial perspective distortions in the not-as-tall main subject.
Since using Photoshop to correct the perspective makes it as though the camera were horizontal to the building (with no up-angle), wouldn't it also make the perspective as though the camera were horizontal to everything else, including the human subject?
quote:wouldn't it also make the perspective as though the camera were horizontal to everything else, including the human subject?
In a perfect world, perhaps. In reality it basically just skews the image - as a single 2d plane - to mimic effects of perspective.
Skewing one thing to "correct" it will invariably skew other parts of the image in a negative way. The extent to which it will do this will vary greatly of course, and there may well be many times where it will not be noticable.
I understand the tool, but was wondering if the nature of the photograph (say 30 degrees up-angle of person-in-front-of-building) didn't cause an even perspective shift to everything in frame, relative to the position in the frame.
It just seems to my mental picture of how the whole thing works that adjusting the bottom 6 feet of the building to be 'square' to the camera would also adjust a 6 foot person in front of the building to also be square to the camera .. i.e. non-distorted. It's just that the person (and the very bottom of the building) seems non-distorted before adjustment because the distortion is so minor at the bottom of the frame. Now I know you could certainly skew somebody out of true by using the tool, but it just seems like you wouldn't if you were correcting a building.
This is all just my mental image of how it works, mind. I haven't actually tried any experiments on it.
[This message has been edited by Das (edited 03-15-2004).]
From the work I've done with the tool in the past, I would have to say that it is nowhere near as powerful or as "real" as the name implies.
It is meant to give an image - treated as a single object - perspective within the canvas, as opposed to correcting real world perspective. It can handle some of the minor issues, but is not going to solve any major visual problems in an image...
I'm not sure about fixing the issue in PS. I haven't tried it. I can tell you that a wide angle lens will cause this effect. I do this type of photo with a telephoto lens (lens above 50mm) or you can use a shallow DOF (to blur out the background) .
I also don't fully understand why correcting perspective of the background with PS can introduce distortions, instead of also correcting them, in the main subject, after all, as said, the angle between the film and the subjects (foreground and background) is the same.
I guess that the answer is that what happens through the lens, to the image, to the light, is far more complex than what can be corrected by simply stretching or skewing an image. When PS is correcting perspective, imo it's simply stretching the image in different degrees of intensity (skewing). That link that Das provided shows better what happens.
In the case of the picture I posted here, all I did was to stretch the upper part of the image while maintaining the bottom of it unchanged. That's what the perspective tool did. So, my daughter's face got a little wider than in the original picture. I measured the distances in both pictures and then I resized the corrected image to make my daughter's face the same width it had in the original one.
Then I asked myself: why do I believe that the width of the original picture (her face) reflects the reality better than the same width of the perspective-corrected picture? Well, I don't know but if I am to trust something, I'd better trust the optics then the electronics, specially since her face is in the middle of the image, where distortions are minimized.
Here's a small tutorial that I found about the problem of correcting perspective using Photoshop: http://luminous-landscape.com/tutorials/perspective.shtml - BTW, this seems to me a great site about photography. Since I'm new to this forum, most probably this site, luminous-landscape, has already been mentioned here.
If the girl were closer to the building I suspect there would be no objectional distortion. As she is so much closer, the perspective on her is actually dfferent that that of the building.
"What can you do with the view camera that the tilt/shift can't do?"
Here's a very abbreviated summary.
The most flexible (but least portable) view cameras are built around a "spine" that is called a monorail. Can be square or circular in cross section. On either side of the tripod mounting fixtue are two sliding units. One has a clip to hold a lens mounted on a board; the other has a clip that holds whatever it is that holds the film. They both also have a clip that holds the light-proof "body" of the camera. This "body" needs to be flexible for reasons I'll get into later; normally this is a pleated construciton similar to an accordian bellows, though for wide angle optics it can be more shapeless so movement is not restricted. Focussing is accomplished by sliding the two standards closer together or farther apart.
Both the lens and the film standards are articulated in just about every direction in 3 dimensional space. Both can move side to side ("shift"), up and down ("rise and fall"), around the vertical axis ("swing") and around the horizontal axis ("tilt"). Different manufactures implement the nodal point of the tilt differently and claim superiority, but that's the basic range of movements.
Understand that view camera lenses are designed to project a larger image circle than the size of the film. Let's make up numbers: say you have a camera that takes 4x5 inch film, and a lens that projects an image into the camera body of 10 inches in diameter. This means you can move the film around in that image circle. What? Okay - you point a camera at a building and don't get the top. You don't want to point it up for fear of convergance. You keep the camera (or more accurately, the FILM PLANE) parallel to the buiding but RAISE the lens (or LOWER the film) and now you have shifted the part of the image circle that has the top of the building so that it is projected onto the film.
Another example. Let's say I have to take a picture of a mirror in a rectangular frame. I don't want to be reflected in the mirror, so I have to stand to the side. But I don't want the geometry of the mirror to be distorted. I can stand to the side and SHIFT the lens or film standard so the mirror is centered on the film and the geometry is preserved.
All this assumes a lens with infinite coverage which in the real world never happens. Sharpness and exposure falls off at the edges of the circle too (it is possible to move the film outside the area of the image circle - not good!) so there are always compromises.
That all has to do with maintaining geometry. In terms of focus it is possible to get a plane of infinite depth in focus with the lens wide open using tilts or swings. The classic example is the railraod track stretching to the horizon. With a fixed body camera you are reliant on depth of field (a small aperture) to keep it sharp near to far. With a view camera you can tilt the lens and/or the film standard and have that PLANE the railroad tracks lie on dead sharp wide open. Note I said "PLANE" - you still need a small aperture to get the telephone poles that are perpendicular to the plane of focus sharp!
The price for all this flexibility is being bound to a tripod and doing your focus/composing under a black cloth to block out ambient light while looking at the image upside down!
That's a superficial summary of whata's different about a view camera from a fixed body camera. Helpful?
When you take a photo, you are capturing things in 3 dimensions. Perspective is something which we use to simulate the effects of these 3 dimensions, but doesn't control the actual depth of a scene, of course.
Everything that you look at, in that photo, will have the effects of perspective applied in all directions, as each object in the photo is 3D.
Photoshop's perspective transform tool is a very simple symmetrical skewing tool, and can ony simulate a simple distortion to recreate the effects of perspecitve on a single 2d object.
It has no concept of the various effects of perspective which are inherently captured in any photograph, and therefore cannot accurately correct them.
Two objects being on the 'same plain' and therefore being able to be corrected simultaneously only applies if the camera and the foreground object are alreay perfectly aligned with the background, in which case no correction would be needed anyway...
The luminous landscape link helped a lot. So my mental image of what 'should' happen is only right if the human subject is at virtually the same distance as the building that's being corrected (i.e. back against the wall). That boulder really showed how bad the distortion is if the two subjects are at wildly different distances.
Wonder how much a tilt/shift lens runs? I do take pictures with buildings in them, and using a tilt/shift seems like an interesting skill to pick up.
As advantageous as the view camera sounds, I think I'll leave that for others. I'm too spoiled by all the convenience aspects of a digital SLR.
Good grief. I finally got around to following the link to the luminous landscape and realized I could have saved myself 20 minutes by linking to his discussion of camera movements instead of writing it myself.
Even used the same mirror example. Sheesh.
Today at lunch, I confirmed that my eye/brain setup is perspective-correcting. Looking up at the 50 story office building I work in, I was unable to see the perspective of the up-angle view. The facing wall looked perfectly rectangular. I could measure the perspective - using my index finger as a ruler, the bottom of the wall was one fingerlength wide and the top was less than 1/3 fingerlength wide - but I couldn't see it. The edges of the wall and the lines of windows all looked parallel, not converging towards the top. The width of the bottom of the wall looked to be the same as the width of the top, even though I could measure that they were clearly different.
Brains are freaky.
It does explain why I like perspective-corrected photos of buildings more than uncorrected ones, though. The uncorrected accurate photos don't look like what I see in real life, since my brain doesn't do the correction trick on a photo.
~adds "Brains are freaky." to list of favorite phrases~
Makes you stand in awe of the artists and draughtsmen who codified the art of perspective drawing, doesn't it? I mean, what "normal" person would have even seen the need for that, brains processing the information the way brains do.
But I'll be you have seen architectural photos that have gone too far, giving super-tall buildings a "looming" sense. I have. You can't quite put your finger on what's wrong, but you know something isn't right...
posted 03-13-2004 00:06
