24 70 2.8L Barrel Distortion

Discussion in 'Canon EOS' started by jean_john, Nov 28, 2010.

  1. I just a got a new Canon 24-70 2.8L. I am sometimes annoyed by the barrel distortion exhibited by this lens - especially when I shoot parallel edges or something like that. I am wondering whether the copy I got has more distortion. Please have a look at the picture attached and comment. It is shot with a 5D Mark II @ 24mm. Is there anyway I can measure barrel distortion ?
    00XlJx-306577584.jpg
     
  2. Virtually all wide zooms have barrel distortion. More distortion in your copy would not be a typical defect. There are many trade-offs in lens design and barrel distortion is usually lower priority than other factors (sharpness, size, cost, etc).
    There is a great program that will remove barrel distortion. PTLens.com. It has a database of all common lenses and uses the EXIF information to determine how much distortion to remove.
    Edit: just saw the image --- This is not barrel distortion, it is perspective distortion (not a defect, law of geometry). Barrel distortion causes straight lines to not be straight. Perspective distortion causes parallel lines to not be parallel.
    To minimize perspecive distortion, keep the camera level (pointed straight ahead, not up or down). If this is unavoidable, correct the distortion in Photoshop. But, be careful, sometimes taking it all out looks un-natural, people are used to seeing some perspective distortion when the camera is obviously pointing up.
     
  3. I don't see any significant curvature of straight lines in the photo you posted, and your caption says "No more parallel edges." This leaves me wondering if you are confused about the meaning of the term "barrel distortion" and whether you understand that it is perfectly normal and expected for a rectilinear wide-angle lens, when tilted relative to the subject, to exaggerate perspective.
     
  4. I was expecting the extreme left edge(back left edge of house) to be parallel to other edges(exactly like we see). Isn't this because of some sort of distortion ?
     
  5. Read both of the replies
    And Google "barrel distortion"
     
  6. Jean, Distortion? What distortion? :) P.S. Love the exposure here. Dramatic! A little under really provides a snappy look to the highlight areas. Best, LM.
    00XlLD-306581684.jpg
     
  7. But what you want is not "exactly like we see." Something farther away always looks smaller to the eye than something close up. In your picture, the top of the house is farther away than the bottom of the house, so the top looks smaller, and this forces the sides of the house to seem to converge toward a vanishing point in the sky. Wide-angle lenses make this perspective effect more dramatic, but even telephoto lenses show it to some degree. It's not a defect of the lens, it's just a natural phenomenon that occurs when you map three dimensional spaces onto a two-dimensional surface.
     
  8. Btw, Len is confusing the issue by using Photoshop to simulate the effect of a shift lens. With a shift lens, you keep the camera level (instead of tilting it up as you did) but you shift the lens vertically to form an image that enters the lens diagonally instead of straight on. You can see that Len has managed to make the verticals parallel, but the diagonals are still converging toward vanishing points to the left and right.
     
  9. William Michael

    William Michael Moderator Staff Member

    The image does not display much, if any Barrel Distortion.
    The image displays Convergence of Parallel Lines. The cause is The Low Camera Viewpoint and the upward Camera Viewing Angle.
    "Convergence" is sometimes loosely termed "Parallel Distortion"
    WW
     
  10. William Michael

    William Michael Moderator Staff Member

    Is there anyway I can measure barrel distortion ?​
    Yes - a simple test is to shoot a rectangular object and have the camera at right angles to it in BOTH planes and in the centre of the rectangle: http://www.photo.net/photo/10163236&size=lg
    WW
     
  11. I was expecting the extreme left edge(back left edge of house) to be parallel to other edges(exactly like we see). Isn't this because of some sort of distortion ?​
    Yes, but this is perspective distortion and not an optical flaw (i.e., barrel distortion). It is a feature of every lens and happens when you shoot not from the center of your subject. If you have to avoid it, you must use tilt/shift lenses to get perfectly parallel lines.
    Welcome to the world of photography!
    P.S.: Our eye lens shows exactly the same as a photographic lens. However, our powerful post-processing (human perception system) tricks us and lets us think that the lines are parallel. They are, of course, in reality, but in our RAW vision they are not. So as a photographer have to sometimes use a lot of tricks to make things appear as they do to our brain... For example, that's why they make the T/S lenses.
     
  12. So...I am wondering why our eyes are not exhibiting perspective distortion(Or are they ?).I tried looking at the same house using one eye closed and I can see all the edges parallel.Is it because our eyes have a curved retina ?
     
  13. Bueh already answered your last question. It's human eye/brain post-processing. We also have "automatic white balance" in our wetware.
    Also, the reason "normal" lenses are called that is because they come close to being the field, etc. seen by the very different organ, the eye.
     
  14. Our eyes do see perspective distortion. Doesn't something look smaller to you when it's farther away? That's the same effect.
    Try reading up on perspective at Wikipedia. What you are seeing in your picture of the house is three-point perspective, with vanishing points above, to the left, and to the right of the image because the camera is not parallel to the house along any axis. This might be a good place to start:
    http://en.wikipedia.org/wiki/Perspective_(graphical)
     
  15. So...I am wondering why our eyes are not exhibiting perspective distortion(Or are they ?).I tried looking at the same house using one eye closed and I can see all the edges parallel.Is it because our eyes have a curved retina ?​
    It's because our eyes drift around to scan a picture this wide. You can detect objects in a very wide angle around you while staring at a fixed point (try waving your hand at 90° while staring at the wall in front of you) but you cannot focus on it. The eyes don't like fuzzy parts and move around all the time and if muscular stress is too high you will automatically move the orientation of your head instead. It's like making a stitched panorama with a normal to telephoto lens, each individual shot will have no noticeable converging lines.
    Actually I don't know if the curved retina compensates for this type of distortion, would be interesting to know.
     
  16. Here's barrel distortion
    ;)
    00XlMm-306601584.jpg
     
  17. Hi Jean,
    This is not barrel distortion, it is "perspective" distortion caused by the fact that your lens is pointing upwards. Go out and take the shot again but this time keep the camera perfectly level (the lens is not pointing up or down but horizontal – use a builders level if you want), you will now find that even though you may have to cut off the top of the house, all of the lines you expect to be parallel will indeed be exactly parallel. There is nothing wrong with your lens, in fact your photo looks fantastic, this is normal optical behavior – our brains compensate for it so it won’t look like this when you view it with the naked eye (actually, your eye actually “sees” a fisheye like bulging image but the brain fixes it all up)
    Keep at it!
     
  18. Bueh already answered your last question. It's human eye/brain post-processing. We also have "automatic white balance" in our wetware.​
    Actually, that isn't quite true, or at least, it's a misleading way of describing what's happening. The real issue here is that the mind interprets what it sees and then tends to think about its interpretation rather than on what it is actually seeing. When you see lines converging and you recognize it as a square tilted in space, you tend to think of it as a square rather than as the particular arrangement of non-equal angles and lines that your eye actually sees. This is a natural behavior that is very good for survival -- you wouldn't want to fail to recognize a dangerous predator just because you were seeing it at an angle. But this is also why one of the hardest aspects of learning to draw is to portray what you actually see rather than what you think you see.
     
  19. The 24-70 is designed to be a top end press photographer lens. Its fast, durable and sharp. You'll see it in press conferences and in the pack chasing some celebrity down the street. Its usuallt teared up with the 80-200 and its not optimised to be a landscape lens. To have a lens that minimises perspective convergence, you need to be looking at some of the wider specialised zooms that are for that purpose. The 17-40 is an example, but Canon are only just starting to update their wider range. Ideally you should look for something at the 18-20mm range. High speed is not as much a factor for landscapes and architecture. These are usually shot in the f8-f16 range to maximise depth of field.
    So do the research, read the reviews and consider what sort of photography you mostly do. Then make your purchase carefully.
     
  20. To have a lens that minimises perspective convergence, you need to be looking at some of the wider specialised zooms that are for that purpose. The 17-40 is an example, but Canon are only just starting to update their wider range. Ideally you should look for something at the 18-20mm range.​
    Well... no. Sorry. Wrong.
    Perspective distortion in rectilinear lenses varies with focal length. To say that the 17-40mm f/4L has less perspective distortion than the 24-70mm f/2.8L is simply wrong; they're both using a gnomonic projection. At the same focal length they behave identically in this regard, though other aberrations may vary between them.
     
  21. Are you shooting with a Full Frame camera, because there seems to be allot of distortion in that picture ? That or maybe you were standing to close to the house when you took this shot and had to tilt the camera up thereby causing convergence of parallel lines.
    On a cropped Canon camera, a 24mm lens is the equivalent of 38mm which is slightly beyond wide angle so you should not see that much distortion. I have a 10-22mm which is a super wide angle and if not careful with this lens you can get severe perspective distortion. What I try to do is set an imaginary straight vertical line inside the viewfinder then try to align the edges of the building or structure agains this imaginary line. If your camera has some type of grid or you have a focusing screen with a grid this works even better.
     
  22. William Michael

    William Michael Moderator Staff Member

    "So...I am wondering why our eyes are not exhibiting perspective distortion (Or are they ?)."​
    They don’t “exhibit it” – they “see it” that is if the brain allows us to see what the eyes see - or if it is so strong an effect, that the brain cannot misinterpret:
    Stand in the middle of a long straight road - the edges meet at the horizon - you will "see" that.
    WW
     
  23. Our eyes do indeed see perspective distortion. Stand at the base of a tall building and look up. You'll see its side converge toward the top.
    However, wide-angle camera lenses EXAGGERATE perspective distortion. Take a picture of the same building with a 24 mm lens and the sides will converge more quickly, i.e. at a steeper angle than what our eyes see.
    Software (PS, etc.) can correct for perspective distortion in some cases, but this comes at a cost of resolution. Shift lenses and view cameras can correct for perspective distortion without loss of resolution, but their range is limited. You can correct the church across the street but not a skyscraper when you're standing near its base.
     
  24. What has been noted here is in fact perspective distortion which any lens will exhibit when point off level at something like a building. The closer one is to the building, the more it will show. A 600mm lens pointed and including a skyscraper will exhibit much less perspective distortion, possibly none, because of the great distance it would be from the subject to include it all.
    The nice thing is that a camera with the MP of the 5dII can, in many situations, allow for the software correction of perspective distortion without the "apparent" loss of resolution that will be noticeable with smaller MP cameras. The best results are achieved by moving the bottom and top of the image equally(bottom in, top out), thus reducing the vertical distortion (crunching) of the object and minimizing the actually resampling magnitude at either end of the image. I do this sort of thing quite regularly and still can make huge prints without any apparent loss of resolution--you just need to play with it to find the optimal procedure.
    On the other hand, your 24-70 is going to have barrel distortion--it is there in this photo although minor compared to the perspective distortion. In fact, most lenses have some, even fixed focal length lenses. It will be more pronounced with shorter focal length lenses and may all but disappear the longer you go. PT lens is a good inexpensive add on program, however, PSCS5 now has it built into the raw processor. At some focal lengths, you might get mustache distortion which may or may not be handled by PT or CS5 but something more specialized like DxO will handle it in most cases. Welcome to the world of optics!
     
  25. Hi,
    See: http://www.photozone.de/canon_eos_ff/528-canon2470f28ff?start=1
    Esa Kivivuori
    Finland
     
  26. Well it's been sorted as to what the effect is called. For the sake of others, as well as the OP. what should we do when shooting to work around the prespective (other then by a TS lens). My thought would be a ladder, or something that would change the highth of the shot.
     
  27. Jamie, that can work in some cases, getting on a ladder, but many times a high viewpoint will look odd. If you don't have a tilt shift, the best solution is just fixing the perspective in post. As I said above, I do that a lot. There are things to learn to do it most effectively, but it is one tool that is a must if you want to correct the issue.
     
  28. This excellent thread brings up a related question I've had for awhile: Of the available photo editing software, such as those mentioned, like PTLens, CS5 and DXO, is there anything that will correct lens distortion in a non-destructive way? The ability to correct in CS5 RAW is interesting- did Adobe offer that update as non-destructive, or simply as a convenience to eliminate a step in the workflow? Thanks in advance- there seems to be a lot of very knowledgeable people here.
     
  29. Is there a real distinction here, or merely a difference?
    the mind interprets what it sees and then tends to think about its interpretation rather than on what it is actually seeing​
    as opposed to
    It's human eye/brain post-processing​
    NTIGAD
    And what is this "mind" thingie? ;)
     
  30. Robert, Lightroom 3 corrects lens distortion AND perspective distortion non-destructively.

    Jamie, you can elevate your shooting position, or you can shoot with your camera level. In the latter case you'll have to crop the unneeded part of the frame away. A third possibility would be to take a number of telephoto images of the building and combine them into a single image with a lot of pixels. That way when you correct the distortion, the loss of pixels will not be as serious. Plus you get the added benefit that telephoto lenses create less perspective distortion in the first place.
     
  31. All,Thank you for actively contributing to the discussion. I now have one more distortion to worry about :)
     

Share This Page