1. Reminder: Please use our affiliate links for holiday shopping!

Embedded lens profile distortion correction: is it more correct?

Discussion in 'Image Processing' started by T N Args, Oct 17, 2016.

  1. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    call me Arg
    I suppose the answer is yes in the general. But does the correction sometimes make an image look worse?

    An example. One of the two images below is uncorrected. The other is corrected using the embedded lens correction as read by Lightroom. But the corrected one looks a bit more distorted to me?

    PA130832.

    PA130832-001.

    In particular, look at the motorcyclist's face and hair:

    Crop2-130832.

    Crop2-001.

    What do you think?

    Ignore the different 'look' of the two different raw processors.
     
  2. fredlong

    fredlong Just this guy...

    Apr 18, 2011
    Massachusetts USA
    Fred
    "Correcting" the inherently curved field of a wide angle lens to a flat field has to stretch the image in some way. Whether it's done optically or with software it can look unnatural.

    Fred
     
    • Agree Agree x 2
  3. Leighgion

    Leighgion Mu-43 Regular

    134
    Aug 1, 2012
    Madrid, Spain
    Leigh L Pang
    Definitely. In general, and in your particular example.

    Different strokes for different folks, but I personally pretty much never bother with distortion correction as I find it a lot of sweat and tears for a very theoretical benefit. I can see the need for things like architecture if you're doing a job that involves flattering your client's building, but that's never my situation. The way I see it, if a distortion-free photo is really important, I take care of it in the lens selection stage or in the case of a zoom, by picking the optimum focal length that minimizes distortion.

    If you get close with a wide angle, distortion is going to happen and in the end, a computer can only do much.
     
    • Like Like x 1
  4. Turbofrog

    Turbofrog Mu-43 Hall of Famer

    Mar 21, 2014
    Careful now, it sounds like you may be confusing perspective distortion with rectilinear (i.e. barrel or pincushion) optical distortion.

    The auto-correction only applies to the optical distortion. Perspective distortion is solely a product of the distance to the subject, regardless of the focal length or field of view.
     
    • Agree Agree x 1
  5. Wisertime

    Wisertime Mu-43 All-Pro

    Aug 6, 2013
    Philly
    Steve
    I assumed this feature was only used to correct vertical lines in architecture/landascape (at least for me) photos or say photos with Light poles, flag poles etc that are badly distorted. I would never have considered using it for a portrait or people scene.
     
  6. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    call me Arg
    To both of you, thanks for commenting, but note that I am not personally applying any distortion correction in the LR example. I am talking about (and showing) the embedded m43 lens correction that is automatically applied (by LR and most other raw processors). That is the issue here. In the corrected (top) example, the rider's face looks a bit hollowed out and the top of her head stretched. Despite her not being anywhere near to the edge of the image circle.

    I did not specify the lens used for this example, because I have noticed the same concern with this 'distorting' effect of the automatic lens correction with other lenses too. I did not want discussion to focus on this particular lens, because the issue is more generic than that. Fred correctly noticed that this is a wide angle photo, 12mm in fact, but not ultra wide.
     
  7. wjiang

    wjiang Mu-43 Hall of Famer

    12mm is easily wide enough for rectilinear perspective distortion to make people look bad, I notice it even at 15mm and 17mm with inappropriate subject placement! Her head in this case is actually quite close to the edge, it's certainly nowhere near centre.

    The embedded correction is usually going to do barrel distortion correction for wide angle lenses. This can tend to make the rectilinear perspective distortion worse as, well, it makes it more perfectly rectilinear.

    The most extreme case is when you get really really wide with a fisheye. In most cases people will look more natural with the original fisheye projection or even better, a stereographic projection that preserves local angular relationships. If you try to 'de-fish' to rectilinear, you can be fairly certain people will look really badly distorted unless they are in the middle of frame.

    So yes, I suppose in some cases it might suit the subject to 'leave in' the barrel distortion, but it would look really bad generally if that distortion was left in. In any case it's up to the RAW converter whether they give you the option to turn it off.
     
  8. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    call me Arg
    Her head is actually closer to the centre than to the edge of the image circle.

    @wjiang@wjiang, do you know any raw converters that give you the option to turn off m43 embedded corrections?
     
  9. David A

    David A Mu-43 All-Pro

    Sep 30, 2011
    Brisbane, Australia
    When I saw the title of the thread in the new posts list—"Distortion correction: is it more correct?"—I thought the post was going to be about correcting perspective distortion such as keystone distortion. I looked at the images first without reading Arg's comments and I spent a long time looking at the images and can see no evidence of that. When I finally read the text, Arg's second paragraph made it clear that he is posting about lens distortion correction based on the lens correction profile which gets embedded in the metadata of RAW files (not in JPEGs because it is applied in camera as part of the JPEG conversion process). I think what is under discussion might have been a bit clearer if the thread title had said "lens correction" or "lens profile correction" rather than "distortion correction", or at least it would have been a bit clearer to me.

    So, assuming we're talking about the application of the lens profile (which is the way the manufacturer intends the images to be rendered but manufacturers aren't always right) I still have concerns about what I'm looking at. Arg said "Ignore the different 'look' of the two different raw processors" but it looks to me as if the images have been processed to differing degrees. That's always an issue when comparing results from different RAW processors because the image you get presented with when you open the file has always had some level of default processing applied and different applications have different defaults plus they often let you change the defaults to suit your own preferences. I see increased saturation and contrast in the first image in each pair, and I also see a smoothing of skin texture which could be a result of noise reduction in the first image in each pair.

    I'm assuming the first shot in each pair is the image processed in Lightroom. I don't know if Arg has set his own default settings for how Lightroom initially renders his images but if he has and he has increased contrast as part of his defaults, then that might account for the increase in saturation because increases in contrast in Lightroom also increase saturation slightly. I tried taking the lower of the second images into Lightroom and playing with the contrast slider and increasing contrast produced a similar kind of saturation change to that seen in the upper image in each pair and also resulted in a bit of smoothing of skin texture as skin tone became more uniform. I don't know if Arg has adjusted his Lightroom import setting to also include a bit of noise reduction and I don't know anything about the defaults in the other, unnamed, application or whether those defaults have been altered.

    The other thing I wonder about is export settings for the 2 processors. Lightroom's export settings include a sharpening setting which has only 3 options, low/standard/high. I don't know what the other processing app does here.

    And then there's the old point I've made many times when confronted with a "comparison" of RAW conversion apps. Conversion does not take place when the image is first displayed, conversion only takes place when you save the file to an image format such as JPEG or TIF. All of your editing adjustments up to that point are not edits to a converted RAW image, they are adjustments to the parameters to be applied to the RAW conversion when it is eventually done. Even if we're comparing images made at the original default settings for each application, we're not really seeing a comparison of how each application would convert an image if no adjustments were applied because each application always applies some default adjustments and the defaults are different from application to application.

    So, to get back to Arg's question about whether correction is more correct, I have no idea based on these images. They have obviously been subject to different processing apart from the application of the lens profile and that is unavoidable. One of the images looks worse than the other to me, the first image, but I think that the bulk of the difference can be put down to differences in contrast, saturation, and sharpening/noise reduction which could be adjusted in that application to deliver a similar result to the lower image. If the first image were so processed, so that it was more to Arg's tastes, would he then be asking about whether lens correction is causing the differences he sees here?

    I don't think the lens correction is causing the difference. I think the cause of the differences is elsewhere in what each application is doing.
     
    • Like Like x 1
  10. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    call me Arg
    I have added 3 words to the front of the thread title to assist you and others.

    The difference I am talking about is geometric. It is completely unaffected by any settings I made in each image regarding saturation, contrast, sharpening, noise reduction. That is why I asked for such adjustments to be ignored. I can go back and re-post the unmodified raw imports from LR and RPP, but it won't make the slightest difference to the geometric properties of each image, and that is what I want to discuss.
     
  11. David A

    David A Mu-43 All-Pro

    Sep 30, 2011
    Brisbane, Australia
    I'm having difficulty seeing a geometric difference. Before anyone starts jumping up and down about that statement and telling me I need my eyes tested, I do need to get my eyes tested again soon. I think my next checkup is due in a month or so. I suffer from astigmatism which is corrected by spectacles but I think I'm getting close to needing a new pair of spectacles so how I see geometrical shapes and proportion may not be the way others see shapes and proportion. I could offer Arg my eyes but, problematic as they are, I think I need them more :) They aren't perfect but they are the best I've got.

    Seriously though, differences in colour and luminance can affect our perception of size and shape. Light coloured walls make a room appear larger than the same room will appear with dark coloured walls. Interior decorators work with colour and lightness/darkness to work all sorts of effects on how we perceive the size and shape of a room and the relative sizes and shapes of the furniture within the room. The same effects from differences in colour and lightness/darkness can be seen in our images as we process them, and often the sense of space, scale and 3 dimensional volume/solidity can be quite noticeably different in an image if you do a side by side before/after comparison. Portrait photographers rely on this and pay a lot of attention to studio lighting in terms of direction and hardness/softness and how it falls on the subject's face to make often quite dramatic changes to how a face looks in terms of it's fullness and shape. You're choosing a face to demonstrate what you see in this image and the face is showing those sorts of differences

    Looking at the first pair of images which are matched for size (that's also important) I can't see a difference. Lens correction corrects for barrel or pincushion distortion and looking at "straight" lines in both images I can't see any difference in their "straightness". I would expect to see that if the correction was having a significant effect, and I have seen corrected/uncorrected shots of test targets which are intended to show lens distortion and the accuracy of the profile correction. Moving the browser window sideways and lining up the top of a 'vertical' such as any of the many posts in the first image with the edge of my screen and looking at the second image, the top of the same vertical is intersecting the edge of the screen at the same point. I can't see a noticeable difference in the vertical plane. The images aren't side by side so I can't apply the same rough test for 'horizontal lines using the bottom of the screen as my guideline. If I look at lines instead of shapes, I can't see a significant difference or problem. Lines won't show the effects that differences in colour and tone have on our perception of shapes and 3 dimensionality.

    On the other hand, some parts of the image do 'look' a different size in each image and the 3 dimensional contours look different. The face looks a little 'flatter' to me in the second image than in the first image but the differences in the skin tones and contrast (the darkness of the shaded areas of the face) are going to affect how we see the face, the 3D modelling of the features, and those differences may make one version appear "distorted" relative to the other. The differences in colour and tone here aren't as dramatic as you can get by changing lighting setups in a studio but I do think those differences are large enough in these images to affect our perception of the shape and proportions of the face.

    I'm not convinced that the differences are due in any significant way to the lens correction or lack of it. I still think the differences are largely due to the differences in colour, saturation, and perhaps sharpening/noise reduction.

    I also don't think posting the unmodified import presentations from LR and RPP will help. If you were going to post different versions of the images I think you need to process them so that the colours and tonal gradations in each image are identical. What we need if we're going to say that the lens correction is a contributor to the difference is a comparison between images which are absolutely identical in every respect apart from the application of the lens correction. You can probably only do that if you have an application which allows you to choose whether or not to apply the correction and you compared 2 versions with the same processing apart from the application of the lens correction. If you did that, what I think you would find is that whether or not you think application of the correction is right/wrong or correct/incorrect may very well boil down to which lens you use and just how bad it's uncorrected distortion is and how good it's profile correction is.


    Addition: Brian just posted the sort of comparison I am suggesting while I was busy typing my reply. I prefer the corrected versions which do look "correct" to me and I don't see any sort of downside to the correction which would make me want to turn it off.
     
    • Appreciate Appreciate x 1
  12. fredlong

    fredlong Just this guy...

    Apr 18, 2011
    Massachusetts USA
    Fred
    @David A@David A
    The distortions visible in the corrected version include a stretched out head, helmet, front tire, trash bin and sewer grate. All of these are not distorted in the uncorrected version. But that one shows the back of the parking lot as very curved as opposed to straight in the corrected version.

    Fred
     
    • Agree Agree x 1
  13. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    call me Arg
    Thanks Brian. I am not suggesting that the lens corrections are not generally better; if I may repeat the opening statement of the thread, "I suppose the answer is yes in the general.".

    What I am concerned by is that they are automatic, yet sometimes create stretching and other geometric effects that can make an image look worse, not better.

    Regarding your comments on OV3, if I understand you correctly, you are not sure whether the Auto Distortion Correction checkbox applies the embedded correction, or something different and additional? Interesting point. Perhaps comparing the in-camera JPEG (viewed in OV3) with the OV3 treatment of the raw file, with and without the check box ticked, will shed some more light?
     
  14. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    call me Arg
    @David A@David A , thanks for your extensive discussion, much appreciated, but I assure you, there are geometric differences in my example, and they are due to the lens correction and independent of colour and tonal treatment. :friends:
     
    • Agree Agree x 1
  15. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    call me Arg
    You may need to take another photo of a window with that lens, camera set to raw + JPEG! :p
     
  16. Ross the fiddler

    Ross the fiddler Mu-43 Hall of Famer

    Without reading all the "essays" posted on the topic (sorry, but I didn't have time to read them all), I notice that Capture One shows the image (at 12mm) from the 12-60 SWD lens corrected to 100% with the same result & by returning that to 0% it then looks correct, so it is worth playing with that when editing a raw file in these programs.
     
    • Appreciate Appreciate x 1
  17. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    call me Arg
    OK, so it looks like there is software that allows on/off selection. Pity it's OV3! :hide:

    And, doesn't help Panasonic users.
     
    • Like Like x 1
  18. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    call me Arg
    I feel differently (compared to m43 lenses) about the conscious application of correction to lenses that do not have embedded correction profiles and were not designed with that in mind, and for that reason probably need less correction anyway, and the photographer makes PP decisions as required. And the same decisions will apply to JPEG files using such lenses. The 12-60 SWD Four Thirds is one such example.

    In Micro Four Thirds Native Lens World, however, this happens automatically and rather invisibly using most software, and irreversibly for JPEG shooters. Yet, sometimes, there is a better image lurking within!
     
    Last edited: Oct 18, 2016
    • Like Like x 1
  19. Brian Beezley

    Brian Beezley Mu-43 All-Pro

    I just did the RAW + JPG experiment. The camera JPG image is the same as the RAW image with Distortion Correction off.

    Now I'm totally confused.

    Brian
     
    • Appreciate Appreciate x 1
  20. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    call me Arg
    me too!