Lens distortion correction is a resolution killer

[DUSTY LENS]

Surely "correcting" perspective is actually to add distortion? The converging verticals aren't an error -- they're what the camera actually saw. We're all perfectly used to seeing images like that all the time, as most professional images and videos don't deliberately distort pictures by changing the verticals in an artificial way. When you artificially distort the image to make the verticals parallel, you're creating an impossible image, as the camera's position is too low.

Actually the act of straightening converging vertical or horizontal lines is not distoition , nor is it creating impossible images .

It happens all of the time , and it is not distortion or impossible in the case of a shift lens and in view cameras using shift control .

But it also happens with unshifted photos simply by holding the lens angle so that the vertical lines are parallel to the image plane or the sensor / film plane .

It is a normal part of optics and photography . the Human eye is perfectly able to apply distortions to images which are recorded with parallel lines which are reproduced parallel as in the real physical world .

A case in reality could be a block ( or other material ) 6 foot high wall which is straight on the top and a constant 6 feet high from one end to the end which is miles away . The camera can be positioned so that the plane of the wall is parallel to the plane of the side of the wall which is facing the camera . It will be reproduced with the top and the bottom of the wall having parallel and straight lines on the film or on the sensor .

This is a common occurrence in document copy photographs . It is certainly not distortion .

Dusty

It's fine if the rectangular object is at infinity or indeed if it's a simple flat object (like, say, a notice on a wall). I do correct the image to make such signs rectangular, and crop the rest of the image. But if it's of, say, a very high room, where the camera is obviously only a meter or two off the floor (as can be seen from people in the image), then to make the back wall rectangular introduces the impossible viewpoint. To look real, the camera should be placed half the height of the ceiling.

But, in any case, this discussion of stretching converging verticals to be parallel is really a separate discussion from the subject of this thread, which was about whether correcting lens distortion affected image resolution. Clearly, converging verticals in wide angle shots is not a distortion in the lens -- lens is accurately recording what it sees. Some people like to artificially alter the shot in post-processing to make to appear as they think the brain would transform the same scene if viewed directly; others (probably the majority), prefer not to add this distortion, and let the brain do its usual transformational work, as it would with any video or movie and the vast majority of published still images.

Interesting response Digital Nigel,

This is true .

But your statement :

" Some people like to artificially alter the shot in post-processing to make to appear as they think the brain would transform the same scene if viewed directly

Is very interesting because you believe people alter the perspective in PP , to make it agree with what they believe is what the eye would see if standing in the camera's position .

I agree with this . But I do not believe it is artificial anymore so than a photograph is artificial .

And I don't believe most people want to see the converging vertical lines either . Instead I believe most people do not have time or energy to correct every photo they keep so that it appears as they believe it should appear .

And I might agree that the act of correcting converging lines might introduce some loss of sharpness , but I do not agree that it always does , or that some PP Computer Systems do not really loose sharpness detail in corrected images . I have been unable to detect this visually in High magnifications on my computer .

Actually it is not artificial , when vertical lines or other lines are photographically reproduced with no convergence detectable in the resulting image . This could have been done with tilt shift optics and the results would have been identical , So the claim of perspective correction being artificial is just an assumption and not supported by any reality or theory .

You or anyone can claim it is artificial but to what purpose ?

shift lenses can do some things which cannot be so easily done in PP . For example if you photograph a window display in a Hi end Fashion storefront , and you want the image to appear as if shot from directly in front of the display but this would create a reflected image of the photographer and his camera , what can you do ?

You could use a shift lens and make the shot from one side but shift the lens so that it appears much as it would if shot from directly in front of it while you and the camera are realy standing off to one side . This cannot be done in Post Processing . So , I suppose you might say this would be an artificial shot . * # ^ ! @ . Artificial what ?

 

[MusicmanUK]

Is very interesting because you believe people alter the perspective in PP , to make it agree with what they believe is what the eye would see if standing in the camera's position .

I agree with this . But I do not believe it is artificial anymore so than a photograph is artificial .

Exactly.

Any projection of three dimensional space onto a two dimensional surface inevitably involves some sort of distortion, whether we realise it or not.

The conventional rectilinear projection for which most lenses are designed is derived from perspective projections developed in Renaissance Europe. This is the projection to which we have become accustomed through hundreds of years of Western Art, but it is by no means the only way to look at the world.

It imposes distortions of dimension - objects that are nearer appear to be larger than objects that are further away, while it is perfectly obvious that this is not the case in reality. By definition, parallel vertical lines do not converge in the real world.

The benefit is that on a perfectly corrected rectilinear lens, straight lines remain straight in the image, but also at the cost of distorting the area occupied by various objects in the scene. An object at the edge of a photograph taken with a wide angle rectilinear lens will be stretched out in a most un-natural manner.

Engineers and Architects use Isometric and Axonometric drawing projections to overcome this limitation, so that scaling along each of the axes of the projection will always be accurate.

Fisheye lenses preserve the area occupied by objects in the scene, at the expense of making straight lines curved. They also preserve the shapes of circles at the edges of frame better than rectilinear lenses do.

Consequently, the question to be addressed when making a photograph is not how distortions may be avoided, but which distortions are acceptable and which are not for each image and its particular purpose.

--
http://www.cybertects.co.uk/

 

[PVCdroid]

Have you noticed that enabling distortion correction in camera significantly affects image resolution? Interestingly it happens not only in the corners where distortions are the worst, but even the center resolution suffers noticeably.

I did a few tests last night with the NEX-7 with Lens distortion On/Off and find sharper results with it off. I was using the SEL1670z which doesn't have much distortion anyway. Also was testing High ISO NR on/off. I do seem to get sharper images with that off as well. Since I pass all images through LR and touch them at least once it seems rather redundant to use this. Also figured out that long exposure noise reduction must be off if you are going to be taking multiple exposures with slow shutter speed or it will sit there processing for about as long as it took to take the picture.

 

[wfektar]

First picture is uncorrected, I assume straight from camera showing natural lens distortion in relation to focus length. You disagree with that? This is a fact of life, laws of optics in action. Your eye is not the lens and your lens is not the eye, they both see differently. What your eye sees is processed and corrected in your brain, without your knowledge. Cameras do the same but only on demand (or you can do that in PP).

Depends on what you think the subject is. If it's a point halfway up the far wall then the first shot looks more natural. If it's the people at the fountain, then the "corrected" shots do, if they had been properly corrected (that is, a vertical stretch applied as well). That is, a crop of a much wider FOV where the photo is about the people in that space.

Of course distortion correction is a resolution killer, especially as it requires a correction in two dimensions (does the in-camera really only correct one dimension?) That's why they invented bellows/shift lenses.

 

[tesilab]

Correcting Barrel distortion should not affect the very center of the images, but rather the corners which have to be stretched.

Pincushion distortion correction shouldn't have too much effect on the corners but would certainly degrade the center (and the sides).

 

[proxy]

First picture is uncorrected, I assume straight from camera showing natural lens distortion in relation to focus length. You disagree with that? This is a fact of life, laws of optics in action. Your eye is not the lens and your lens is not the eye, they both see differently. What your eye sees is processed and corrected in your brain, without your knowledge. Cameras do the same but only on demand (or you can do that in PP).

Depends on what you think the subject is. If it's a point halfway up the far wall then the first shot looks more natural. If it's the people at the fountain, then the "corrected" shots do, if they had been properly corrected (that is, a vertical stretch applied as well). That is, a crop of a much wider FOV where the photo is about the people in that space.

That is clear as the distortion is stronger towards the edges of the picture. Center is least affected.

Of course distortion correction is a resolution killer, especially as it requires a correction in two dimensions (does the in-camera really only correct one dimension?) That's why they invented bellows/shift lenses.

Sure, 2 corrected pics show significant horizontal stretch and that is done at the cost of horizontal resolution. Add vertical correction and you're already "missing" lots of pixels if image is stretched 2-way. Windows (on 2nd and 3rd picture) are almost double size... 2x width produced with the same input material as the 1st one. With shift lenses the correction is done using as input a live image and image pixels produced are filled with live image information collected by the directly lens what is not the case with PP. That's why it's good to see crops of pre- and post-correction to assess the damage.

 

[MusicmanUK]

Sure, 2 corrected pics show significant horizontal stretch and that is done at the cost of horizontal resolution.

That is undoubtedly true. The building is Hays Galleria between London Bridge and Tower Bridge on the south bank of the Thames in London. I'm very familiar with it as I walk past it every day on the way to work and the 'corrected' images make the space appear considerably broader than it is in real life; the glazed barrel vault has a circular profile in section.

There are much better ways of making perspective changes such as attempted here.

 

[kaiser soze]

Surely "correcting" perspective is actually to add distortion? The converging verticals aren't an error -- they're what the camera actually saw. We're all perfectly used to seeing images like that all the time, as most professional images and videos don't deliberately distort pictures by changing the verticals in an artificial way. When you artificially distort the image to make the verticals parallel, you're creating an impossible image, as the camera's position is too low.

Actually the act of straightening converging vertical or horizontal lines is not distoition , nor is it creating impossible images .

It happens all of the time , and it is not distortion or impossible in the case of a shift lens and in view cameras using shift control .

But it also happens with unshifted photos simply by holding the lens angle so that the vertical lines are parallel to the image plane or the sensor / film plane .

It is a normal part of optics and photography . the Human eye is perfectly able to apply distortions to images which are recorded with parallel lines which are reproduced parallel as in the real physical world .

A case in reality could be a block ( or other material ) 6 foot high wall which is straight on the top and a constant 6 feet high from one end to the end which is miles away . The camera can be positioned so that the plane of the wall is parallel to the plane of the side of the wall which is facing the camera . It will be reproduced with the top and the bottom of the wall having parallel and straight lines on the film or on the sensor .

This is a common occurrence in document copy photographs . It is certainly not distortion .

Dusty

I do not think that you have the correct perspective on this. in fact, I am absolutely certain of it. When you stand over railroad tracks and look straight down, so that you can see the tracks in both directions, you see two curved lines. It is not distortion in any true sense of the word, and as such, anything that you do to straighten out the lines would not correctly be considered "correction". I'm not sure what you should call it, but it absolutely is not correction. If you straighten out those lines, you will ADD distortion, and IT WILL NOT LOOK THE LEAST BIT NATURAL. I have a book about M. C. Escher, by Bruno Ernst, published in 1976. Chapter 8, 16 pages in length, deals expressly with perspective. On one page you see a drawing of a person lying on the ground looking up at telegraph wires. Either direction that you tilt your head, looking down the wires, the two parallel lines converge at a vanishing point. The point of the drawing is that this is absolutely true and that the only way that this is possible, without also seeing sharp kinks in the lines at the midpoint, directly overhead, is if the lines are perceived as curved. This effect is not distortion in any true sense. It is not something caused by the eyeball, or anything like that. It is the effect of perspective. If you lie down on that same spot and take a picture of those parallel telegraph wires with a lens wide enough to see fairly far down the wires in both directions, the picture MUST reveal that the lines converge at vanishing points in both directions. It absolutely must, and for it not to do that, would be the same as taking a picture down a long highway and for the two edges of the road to not converge. The ONLY way that it is possible for the parallel wires in your picture to converge in both directions is if they are represented as curved lines. It makes NO SENSE WHATSOEVER to try and get rid of that effect and make them appear as straight lines. It does not work, and this is probably why the people in the picture ended up so weird looking when the lines that should have been represented as curved lines were transformed into straight lines. Anyone who wants a more complete understanding of this question should go try and find a copy of this book, and this advice applies especially to anyone who is applying "correction" to their pictures so that all the things that they assume are straight in reality end up straight on the image. This pursuit is badly misinformed. In fact it amounts to denial of the concept of the vanishing point. If you accept that straight, parallel lines converge at a vanishing point, then you must also realize that straight lines must be rendered as curved lines. You cannot have one without the other.

 

[Promit]

I don't see people complaining that m4/3 lenses are too soft, indeed they have some of the sharpest lenses around. And nearly all of those lenses are software corrected. So it's apparently possible to get good results from software correction, even if they are not the best possible results.

IMO if you sit down and really look at the effects of distortion correction on sharpness of real lenses in real world images, then decide that it's an unacceptable compromise? You're no longer a photographer, you're just a laboratory junkie.

 

[proxy]

Actually the act of straightening converging vertical or horizontal lines is not distoition , nor is it creating impossible images .

It happens all of the time , and it is not distortion or impossible in the case of a shift lens and in view cameras using shift control .

Dusty

... If you accept that straight, parallel lines converge at a vanishing point, then you must also realize that straight lines must be rendered as curved lines. You cannot have one without the other.

Sometimes corrections lead to unnatural results like we can see in pics 2 and 3. The columns are so straight despite their height that they make you feel uneasy and loose the sense of depth. The point about retaining slight convergence is not without merit, this is how we see and keeping proper balance between technical & geometrical perfection and how we see things in reality is the key.

 

[Digital Nigel]

I think this panorama is a good illustration of that:


London Olympics site, just before the Olympics, from an adjacent railway.

 

[kaiser soze]

I think this panorama is a good illustration of that:


London Olympics site, just before the Olympics, from an adjacent railway.

It does. The panorama is inherently the same as an ultra wide angle, and gives you the same image you would get with an ultra wide angle or fisheye. In this particular picture, the overhead wires are in reality parallel with the tracks. At each end they converge to vanishing points, and since they do, it is necessary for both the wires and the racks to curve. If you tried to get rid of the curvature, you would also get rid of the vanishing points, and that would be completely wrong, because it would mean that even though your image includes a part of the tracks and wires that are very far away in the distance, they appear just as widely separated in the distance as they appear up close. Manifestly, that would be wrong, and would ruin perspective, since things that are further away are obviously supposed to be smaller. apparently, there was a time several hundred years ago when painters thought it was correct to draw lines that were parallel to the canvas as straight lines. This was wrong. Today, it seems odd to us that they thought this was correct. Yet, based on many of the comments I read here and elsewhere about lens correction and wide-angle lens, the typical photographer suffers the same misunderstanding. Yes, there is genuine lens distortion, but it seems to me that what most photographers think needs to be corrected, with wide-angle lenses, is not distortion at all, but is simply the effect of perspective. I actually first noticed this when reading a lens test where a picture of a square grid was taken and the resulting image with curved lines was compared to the original image, and the difference between the two attributed entirely to barrel distortion. This method of assessing lens distortion is wrong. The only way that the grid ought to come out completely square, with no curved lines, is if you stand an infinite distance away when you take the picture. Whoever designed this test was very misinformed, and manifestly lacking the qualification to design a test of that sort.

 

[GaryW]

...the typical photographer suffers the same misunderstanding. Yes, there is genuine lens distortion, but it seems to me that what most photographers think needs to be corrected, with wide-angle lenses, is not distortion at all, but is simply the effect of perspective. I actually first noticed this when reading a lens test where a picture of a square grid was taken and the resulting image with curved lines was compared to the original image, and the difference between the two attributed entirely to barrel distortion. This method of assessing lens distortion is wrong. The only way that the grid ought to come out completely square, with no curved lines, is if you stand an infinite distance away when you take the picture. Whoever designed this test was very misinformed, and manifestly lacking the qualification to design a test of that sort.

What I have found is that this "barrel distortion" makes people photos look better. Otherwise, your perfect wide angle photos will have nice straight lines for architecture, but people will be all stretched and weird as they are closer to the edges, and stretched in a mostly non-flattering way.

So, sure, correct for distortion, then put it back in to fix the people. ;-)

But seriously, here's a case where I think the original 18-55 was a decent kit lens. Distortion wasn't bad over most of the range, and even the barrel distortion wasn't that great at the wide end, but there were some that complained about it anyway. And yet your wide angle people photos will come out better with that distortion.

Now that we have the 16-50 correcting severe distortion, is it over-correcting? Maybe use of RAW will add flexibility at least.

 

[kaiser soze]

...the typical photographer suffers the same misunderstanding. Yes, there is genuine lens distortion, but it seems to me that what most photographers think needs to be corrected, with wide-angle lenses, is not distortion at all, but is simply the effect of perspective. I actually first noticed this when reading a lens test where a picture of a square grid was taken and the resulting image with curved lines was compared to the original image, and the difference between the two attributed entirely to barrel distortion. This method of assessing lens distortion is wrong. The only way that the grid ought to come out completely square, with no curved lines, is if you stand an infinite distance away when you take the picture. Whoever designed this test was very misinformed, and manifestly lacking the qualification to design a test of that sort.

What I have found is that this "barrel distortion" makes people photos look better. Otherwise, your perfect wide angle photos will have nice straight lines for architecture, but people will be all stretched and weird as they are closer to the edges, and stretched in a mostly non-flattering way.

So, sure, correct for distortion, then put it back in to fix the people. ;-)

But seriously, here's a case where I think the original 18-55 was a decent kit lens. Distortion wasn't bad over most of the range, and even the barrel distortion wasn't that great at the wide end, but there were some that complained about it anyway. And yet your wide angle people photos will come out better with that distortion.

Now that we have the 16-50 correcting severe distortion, is it over-correcting? Maybe use of RAW will add flexibility at least.

--
Gary W.

I keep trying very, very hard to make a point, but I cannot help get the sense that my point is being missed, in some cases at least. I suppose it might be reasonable to refer to the effect of perspective as a sort of distortion. I supposes, but the problem I have with that is that it muddles the issue of lens distortion. The effect of perspective, call it distortion if you wish, is NOT an effect of the lens! This is the important point here. What sense would it ever possibly make to use a type of correction that is intended to correct for lens distortion, to correct for something that is not in any way an artifact of the lens? It just does not make sense. Here is another way to look at it. Imagine that your camera is a pinhole camera, with an itty bitty little hole in place of the lens, and nothing else. If the surface plane behind the pinhole extends very far in at least one direction, such that you capture the vanishing point effect, then lines that are straight and parallel in reality but that converge at that vanishing point will appear curved. You can call this a type of distortion if you want, but I prefer not to do that, because if I do, there will be a lot of people, most likely the majority, who will incorrectly infer that this effect is inherently a lens effect and that it would make sense to apply lens correction to get rid of it! The only case where you can get rid of without introducing real distortion as a consequence is if the nature of the picture is such that it does not reveal ANY vanishing points! IF your picture shows no hint of any vanishing points, in that case you can apply that type of "correction" and get away with it, without introducing a more obvious, unnatural type of distortion. In general, when you apply a type of "correction" that has the effect of making the edges of a building look straight and parallel, it is NOT correct to think of what you are doing as a correction that undoes lens-induced geometric distortion. This notion is absolutely incorrect. The effect you are trying to get rid of, in trying to make the edges of the building look straight and parallel, is NOT an effect that has been introduced by the lens per se. Furthermore, the typical lens test that I have seen, that measures the barrel distortion of a wide-angle lens, is bogus. In order for these tests to be at all valid, the image should be compared to how a pinhole camera would capture that same square grid. You then have to find answers to a couple of other difficult questions, in order to establish equivalency between the glass lens and the hypothetical pinhole camera. The ideal way to do this would be through using a ray-tracing algorithm of some sort to generate the image used to compare and assess the image that the lens captures. This is obviously a good deal more difficult to do properly, as compared to the simplistic, essentially worthless test that magazines typically perform, in their misguided approach to measuring the geometric distortion of a lens.

 

[Digital Nigel]

You and me both. I keep trying to tell people that editing wide angle shots to eliminate converging verticals is adding, not correcting distortion. Not only does it distort the overall picture in a very unnatural way, but it can visibly distort people or anything else with a known shape near the edges. There's a good reason why professionals very rarely do this.

However, it is sometimes worth distorting an image that's focused on something rectangular, like a painting or sign board where the camera wasn't perfectly positioned in line with the centre of the frame. Once that's straight and rectangular, the rest of the image can be cropped, so you don't get nasty distortions visible outside the frame.

I like your pin-hole camera analogy and may quote it myself next time one of these arguments comes up (probably, all too soon!).

 

[verybiglebowski]

I don't see people complaining that m4/3 lenses are too soft, indeed they have some of the sharpest lenses around. And nearly all of those lenses are software corrected. So it's apparently possible to get good results from software correction, even if they are not the best possible results.

IMO if you sit down and really look at the effects of distortion correction on sharpness of real lenses in real world images, then decide that it's an unacceptable compromise? You're no longer a photographer, you're just a laboratory junkie.

Tell that to Ansel Adams...

I really don't understand, why technical perfection should not be a part of the artistic expression. The form is always needed to express the content, and as such it is important part of overall perception.

I think it comes to the old - taking or making images point of view. If you are snapping life around you, transforming moments in memories or document things and happenings around you, flexibility of a zoom lens, convenience of the small system and simplicity of capturing (such as jpeg ooc i.e.), are much more important than ultimate image quality.

But if you are planning your image several days, months or years in advance, if you walk 100 times to the same location in 4 am hoping for that unique lighting situation, if you prepare your set-up and your models with lot of effort to materialize your vision, things just gets reversed. You will (or should) strive for every single pixel perfection (or intended imperfection), because that is a part of a creative process.

To say about someone who shot his flower in the living room, arranging background, lighting, framing, DOF etc. for many hours, that he is not using "real lenses in a real world images" and that he is not "photographer but just a lab junkie" is unfortunate misunderstanding of photography itself. Being it perceived as an art form, or just a technical tool for documentation.

--
Don't trust your eyes or mind, they might betray you! Trust only comments posted on the forums, because there is the absolute truth!
http://www.flickr.com/photos/viktor_viktor/
http://verybiglobo.blogspot.com/

 

[Greynerd]

Your eyes just see in detail at any one time a small circular bit in the middle. It is all stitched and perspective/colour corrected in the brain. When I first had to wear glasses after a very late diagnosis of short sightedness when I turned my head all the verticals bent in a terrible fashion. This disappeared after a time from processing in my visual centres and not from any optical miracles in the eye.

Probably the actual stuff coming out of the eyes is pretty dreadful which is hardly surprising considering you have a double lens system comprising the cornea and the other lens being a bag of gunk being stretched to shape. Also it is also being interpreted by the model in your brain which is a guess at what the world is like which could be nothing like the reality.

When you walk through tall buildings your eyes do not see curved vertical walls or beams , they all appear parallel and straight . This is as they are in fact . So when ones vision is confronted with vertical structures it does not make them appear curved or even slanted .

 

[Promit]

I really don't understand, why technical perfection should not be a part of the artistic expression. The form is always needed to express the content, and as such it is important part of overall perception.

You misunderstand. I am not opposed to pursuit of technical excellence. Lens design, like many other things in life, is about compromises. Better sharpness versus less distortion, better center sharpness versus better corner sharpness, etc. And there's a whole host of secondary effects, some of which like bokeh smoothness are not easy to test and compare. I am merely stating that as a compromise goes, the one we're talking about is utterly trivial. (Non-trivial example: some software distortion corrections introduce second order, moustache style distortion.)

Do you understand what true technical perfection costs, by the way? Zeiss is willing to answer the question for you, as was RED for a while, as are a few others. And even then I think you'll find a ZF.2 lens isn't quite as ideal as you might think. You can't have it all. Producing the sharpest possible lens (in the center, or the corners?) has consequences elsewhere. Fixing distortions, whether optically or in software, has consequences elsewhere. Price and weight factor in as well.

If you have a lens that seems perfect in some aspect you care about, eg sharpness, it's likely that the lens is compromised in some way that you are not seeing because you don't know or care to look for it.

This is why I occasionally cringe when Roger Cicala writes something. He's like a scientist, very precise and particular about what he's doing and why. Unfortunately his readers often miss the overarching point or the scope of the work and fixate on some particular laboratory result.

 

[DUSTY LENS]

Surely "correcting" perspective is actually to add distortion? The converging verticals aren't an error -- they're what the camera actually saw. We're all perfectly used to seeing images like that all the time, as most professional images and videos don't deliberately distort pictures by changing the verticals in an artificial way. When you artificially distort the image to make the verticals parallel, you're creating an impossible image, as the camera's position is too low.

Actually the act of straightening converging vertical or horizontal lines is not distoition , nor is it creating impossible images .

It happens all of the time , and it is not distortion or impossible in the case of a shift lens and in view cameras using shift control .

But it also happens with unshifted photos simply by holding the lens angle so that the vertical lines are parallel to the image plane or the sensor / film plane .

It is a normal part of optics and photography . the Human eye is perfectly able to apply distortions to images which are recorded with parallel lines which are reproduced parallel as in the real physical world .

A case in reality could be a block ( or other material ) 6 foot high wall which is straight on the top and a constant 6 feet high from one end to the end which is miles away . The camera can be positioned so that the plane of the wall is parallel to the plane of the side of the wall which is facing the camera . It will be reproduced with the top and the bottom of the wall having parallel and straight lines on the film or on the sensor .

This is a common occurrence in document copy photographs . It is certainly not distortion .

Dusty

I do not think that you have the correct perspective on this. in fact, I am absolutely certain of it. When you stand over railroad tracks and look straight down, so that you can see the tracks in both directions, you see two curved lines. It is not distortion in any true sense of the word, and as such, anything that you do to straighten out the lines would not correctly be considered "correction". I'm not sure what you should call it, but it absolutely is not correction. If you straighten out those lines, you will ADD distortion, and IT WILL NOT LOOK THE LEAST BIT NATURAL. I have a book about M. C. Escher, by Bruno Ernst, published in 1976. Chapter 8, 16 pages in length, deals expressly with perspective. On one page you see a drawing of a person lying on the ground looking up at telegraph wires. Either direction that you tilt your head, looking down the wires, the two parallel lines converge at a vanishing point.

They converge in your mind , and in your eyes , But they do not converge in to one line in the real world , So , something like reproduction size made them look like they converge at the vanishing point on the film and in your eye . It's not distortion of the lens or of the sensor , and it is not distortion in the eyes either . It is only perspective having it's effect on the size of the projected image at different distances making distant objects look smaller than close objects on the sensor , etc. .

But this observation and this fact of reproduction size does not mean that the person observing a photograph of every subject or any subject will appear distorted if it is photographed with shift lens techniques applied so that parallel lines remain parallel on the film or on a print of the photograph . many subjects and I would include tall buildings , are very natural looking when the vertical or horizontal lines are photographed with these lines adjusted to appear parallel in the image . and I believe it can be done almost if not completely as successfully in post processing with computers and software as it can be done with shift cameras or shift lenses .

Natural perspective is also a perfectly natural sight and also can appear just as natural in photographs , except when it causes tall buildings to look excessively slanted inward at the top .

( My personal preference ) , not a law to be followed by every photographer .

The point of the drawing is that this is absolutely true and that the only way that this is possible, without also seeing sharp kinks in the lines at the midpoint, directly overhead, is if the lines are perceived as curved. This effect is not distortion in any true sense. It is not something caused by the eyeball, or anything like that. It is the effect of perspective. If you lie down on that same spot and take a picture of those parallel telegraph wires with a lens wide enough to see fairly far down the wires in both directions, the picture MUST reveal that the lines converge at vanishing points in both directions. It absolutely must, and for it not to do that, would be the same as taking a picture down a long highway and for the two edges of the road to not converge. The ONLY way that it is possible for the parallel wires in your picture to converge in both directions is if they are represented as curved lines. It makes NO SENSE WHATSOEVER to try and get rid of that effect and make them appear as straight lines. It does not work, and this is probably why the people in the picture ended up so weird looking when the lines that should have been represented as curved lines were transformed into straight lines. Anyone who wants a more complete understanding of this question should go try and find a copy of this book, and this advice applies especially to anyone who is applying "correction" to their pictures so that all the things that they assume are straight in reality end up straight on the image. This pursuit is badly misinformed. In fact it amounts to denial of the concept of the vanishing point. If you accept that straight, parallel lines converge at a vanishing point, then you must also realize that straight lines must be rendered as curved lines. You cannot have one without the other.

 

[verybiglebowski]

I am merely stating that as a compromise goes, the one we're talking about is utterly trivial.

I certainly don't agree. Software distortion correction is utterly one of the most destructive processes as it involve pixels interpolation and re-mapping.

(Non-trivial example: some software distortion corrections introduce second order, moustache style distortion.)

True, but in most cases, this effect will be less visible than the loss in resolution. It all depends though, on final output and its purpose.

Do you understand what true technical perfection costs, by the way? Zeiss is willing to answer the question for you, as was RED for a while, as are a few others. And even then I think you'll find a ZF.2 lens isn't quite as ideal as you might think. You can't have it all. Producing the sharpest possible lens (in the center, or the corners?) has consequences elsewhere. Fixing distortions, whether optically or in software, has consequences elsewhere. Price and weight factor in as well.

I know about lenses quite a lot and am testing lot of lenses myself. Compromises are fine, but just as with a tools, you need to choose the best one for the purpose of your idea interpretation.

If you don't have any idea, prior of pressing shutter button, but you want to be ready to capture "real" moment, than software correction might be acceptable.

If you does, than choose the lens that will take you as close to desired result in terms of technical perfection as possible.Your result will be better in terms of technical appearance.

It is the quest for perfection (or imperfection for that matter) that is questioned here.

I am not saying that if you don't use the most expensive and best corrected lens for flat focus plane i.e., you won't be able to make nice landscape shot.
I am saying - that if you can, take the lens with the best correction for the flat focus plane (just as example), even if it is the most expensive one, if you can afford it.

If not, search for the best compromise, but don't say that it doesn't matter which lens (camera, filter, tripod etc.) you will use, because that is not true.

If you have a lens that seems perfect in some aspect you care about, eg sharpness, it's likely that the lens is compromised in some way that you are not seeing because you don't know or care to look for it.

This is why I occasionally cringe when Roger Cicala writes something. He's like a scientist, very precise and particular about what he's doing and why. Unfortunately his readers often miss the overarching point or the scope of the work and fixate on some particular laboratory result.

I agree with most of your statements (except software distortion correction), I just had a feeling that you are saying that software correction is perfectly ok for everyone, and who doesn't think so, is not a photographer but a lab jerk.