Crop vs FF for landscape photography

[RLPhoto]

Generally, bigger is better with landscape photography. The bigger piece of film/ sensor will always resolve more detail than a smaller format.

But yes, I've shot some fantastic landscapes with my 7D+10-22mm when I had them.

[Chosenbydestiny]

Your three mistakes are as follows:

* Zooming the lens. This is likely inconsequential with these lenses, but it is a mistake none the less.

* Equivalent processing. This is a huge mistake which invalidates your test and your results out right. You do not use identical processing with different sensors, even different sensors of the same format.

* You do not mention if the shots were unlabeled. If your wife knew which came from which before picking, the results are less than worthless, they are misleading. There is no shortage of examples of conscious and subconscious human bias, of people picking what they think they should pick. It's just what we do. Even if they were unlabeled, a strict scientist would discount your results because you knew, and there's no shortage of ways you could have consciously or subconsciously telegraphed the "correct" choice to her.

Actually, those are your three mistaken assumptions.

* Zooming is necessary.  Cropping to match FoV negates the purpose of the test, and moving the camera changes the perspective of the image, i.e. the size relationships of various elements - kind of important to a picture.  What viable alternative would you suggest?

* I stated 'equivalent' - thinking that I meant 'identical' is a huge mistaken assumption.  Equivalent means processed in such a way as to extract maximum detail with minimum noise, correct color as appropriate, etc.  Obviously, that means different processing for different sensors.  The idea was to make both images as good as they could be, not to stack the deck against the 7D. 

* Did I have to mention that? Really? What do you think, that I said, "Sweetheart, the first image in each pair is from my new, top-of-the-line camera, my favorite camera in the world, the second is from the old piece of crap camera that I'm going to give to our 3 year old as a toy...now tell me which ones you like better, wink-wink, nudge-nudge?"  The fact that you even considered something like that, much less labeled it a mistake that I made, is actually pretty damn offensive.  In fact, I watermarked a '+' or a 'o' on each image, each label was randomly assigned to one of the pair, but balanced so there were the same number of each label in total, and I set up a looping slideshow that she looked at when I wasn't in the room.  Before you go to the next level and suggest she peeked at the EXIF, if she did I bet she'd have picked the 7D because 7 is a bigger number than 1. She likes my pictures, supports my hobby, and doesn't give a darn about the gear itself.  But being a Professor of Anatomy, she does have a keen eye for detail in the images.

FWIW, I am a card-carrying 'strict scientist', PhD and a day job in the field (including responsibility for some bioanalytical assays conducted under GLP conditions), and once upon a time I actually conducted neuropsychological testing.  I explicitly stated the above was subjective and completely anecdotal.  You seem to have truncated your quote of my post right before that statement. 

Given your evident bias and unsupported assumptions in calling out my 'mistakes' I don't see any point in pursuing this discussion further.

I literally grabbed some popcorn for this thread, starting from this post.

[marinien]

Your three mistakes are as follows:

* Zooming the lens. This is likely inconsequential with these lenses, but it is a mistake none the less.

* Equivalent processing. This is a huge mistake which invalidates your test and your results out right. You do not use identical processing with different sensors, even different sensors of the same format.

* You do not mention if the shots were unlabeled. If your wife knew which came from which before picking, the results are less than worthless, they are misleading. There is no shortage of examples of conscious and subconscious human bias, of people picking what they think they should pick. It's just what we do. Even if they were unlabeled, a strict scientist would discount your results because you knew, and there's no shortage of ways you could have consciously or subconsciously telegraphed the "correct" choice to her.

Actually, those are your three mistaken assumptions.

* Zooming is necessary.  Cropping to match FoV negates the purpose of the test, and moving the camera changes the perspective of the image, i.e. the size relationships of various elements - kind of important to a picture.  What viable alternative would you suggest?

* I stated 'equivalent' - thinking that I meant 'identical' is a huge mistaken assumption.  Equivalent means processed in such a way as to extract maximum detail with minimum noise, correct color as appropriate, etc.  Obviously, that means different processing for different sensors.  The idea was to make both images as good as they could be, not to stack the deck against the 7D. 

* Did I have to mention that? Really? What do you think, that I said, "Sweetheart, the first image in each pair is from my new, top-of-the-line camera, my favorite camera in the world, the second is from the old piece of crap camera that I'm going to give to our 3 year old as a toy...now tell me which ones you like better, wink-wink, nudge-nudge?"  The fact that you even considered something like that, much less labeled it a mistake that I made, is actually pretty damn offensive.  In fact, I watermarked a '+' or a 'o' on each image, each label was randomly assigned to one of the pair, but balanced so there were the same number of each label in total, and I set up a looping slideshow that she looked at when I wasn't in the room.  Before you go to the next level and suggest she peeked at the EXIF, if she did I bet she'd have picked the 7D because 7 is a bigger number than 1. She likes my pictures, supports my hobby, and doesn't give a darn about the gear itself.  But being a Professor of Anatomy, she does have a keen eye for detail in the images.

FWIW, I am a card-carrying 'strict scientist', PhD and a day job in the field (including responsibility for some bioanalytical assays conducted under GLP conditions), and once upon a time I actually conducted neuropsychological testing.  I explicitly stated the above was subjective and completely anecdotal.  You seem to have truncated your quote of my post right before that statement. 

Given your evident bias and unsupported assumptions in calling out my 'mistakes' I don't see any point in pursuing this discussion further.

I literally grabbed some popcorn for this thread, starting from this post.

Yeah, the problem with some people is that they are full of (negative) assumptions. And when they read a sentence in which the first half somehow "matches" their assumptions, the second half just becomes invisible or disappears. Neuro clearly wrote that his wife had a look at the slide show when he was not in the room in the second half of his paragraph, and dtaylor quoted the first half of the paragraph and commented some ***** on how Neuro presented his photos to his wife!

[TWI by Dustin Abbott]

I personally could tell a big difference when I went from my 60D to my 5D MKII as my primary landscape body.  There simply seems to be both more color information and also a smoother graduation in color.

That being said, I used both the Canon 10-22 and the Tokina 12-24 and love them both for landscape work.

I now use the 17-40L, but would be hard pressed to say that I like it better overall.

[neuroanatomist]

* Zooming is necessary.  Cropping to match FoV negates the purpose of the test, and moving the camera changes the perspective of the image, i.e. the size relationships of various elements - kind of important to a picture.  What viable alternative would you suggest?

We are discussing whether or not there is an IQ difference between sensors of similar resolution but different physical size. All other factors must be equal in testing this, or perceived differences could be due to something other than the physical sensor size.

I will grant that in some situations those other factors may be practically relevant. For example, when the choices were the 5D and 20D there were no good options for UWA on crop. But in judging any of this we must first determine what, if any, sensor differences exist absent all other influences.

Sorry, I'm not seeing a viable suggestion from you.  The sensors are different sizes - something must change for the resulting images to be framed the same.  The three options are crop the FF image, change the distance to the subject, or change the focal length.  The first defeats the purpose, the second changes the perspective of the image. Maybe I should have moved the camera, but arranged for the Museum of Fine Arts building to be moved closer to the foreground?  Or maybe I should have chosen perfectly flat subjects like test charts, obviating perspective?  Have I missed any options?

* I stated 'equivalent' - thinking that I meant 'identical' is a huge mistaken assumption.

No, it's not. http://www.merriam-webster.com/dictionary/equivalent. Note "equal in value".

I did not assume anything, you misspoke. Next time be more clear and precise in your test description.

Equivalent in result - the best possible IQ output.  Are 2+6, 4+4, and 2³ identical?  No - but they are all equivalent in that the result is equal in value.

... the simple fact that you presented the images was a mistake,

So...the comparison should have been done with non-presented, i.e., imaginary images?  Most inane comment, ever.  Maybe I'm misunderstanding, in which case, you should be more clear and precise in your phrasing.

Translation: you can't prove your point with hard evidence, so you're taking your ball and going home.

I was never trying to prove a point with hard evidence. Please go back and reread my initial post, and this time, pay very close attention to the part you ignored in your first response, the part where I stated that my observations were subjective and completely anecdotal.  Your treatment of it as if it were hard evidence was your first and biggest mistake, one that essentially renders the rest of the discussion moot.

[neuroanatomist]

Mmmmm....popcorn.   

[Kernuak]

First off, I haven't done anything even as objective as Neuro's comparisons (even if you think they are subjective). Also, as with any picture (photo, painting or otherwise), it is the subject and composition that largely dictates the success or failure of an image. One person's favourite photo may be hated by someone else. Art is very subjective and always will be. Basically, I don't have any hard evidence and it is my subjective opinion.
In the favour of crop sensors, the corners, usually the weakest part of a lens is cropped out. However, something that hasn't been considered is the diffraction limited aperture and circle of confusion. Granted, many don't consider it important, but I can certainly see the difference in sharpness between my 7D and wehn I first started shooting with my 5D MkII. Yes I could add more sharpening to the 7D, but that doesn't replace sharpness in the original image capture, all it is doing is increasing the micro-contrast in the transition areas to give the appearance of sharpness. Also, when you are sending images to stock libraries, they specifically state that images should not be sharpened, in this scenario, there is a greater risk of a landscape image from a crop sensor being rejected for being soft (although you would have to have made a pretty big error for that to occur). When you take the DLA into consideration, the current (Canon) full frame sensors allow you to stop down an extra two stops or so to get similar sharpness. With my 7D I could easily see loss of sharpness at f/11, whereas with my 5D MkII, I could go to f/22 before I saw similar levels of diffraction blur (I'm assuming the 5D MkIII would give me similar results, but I try to avoid f/22 and rarely need to stop down that much). The lower pixel density of the 40D allowed me to go to f/16 before softness became noticeable. Of course, for most printing or web viewing, the differences in sharpness would unlikely to be noticeable, it's only when you start looking at 100% (which you do for stock images), that you can see the differences. Also, the higher pixel density will show up lens deficiencies (or indeed poor technique) much more than the lower pixel densities), this is the reason that I feel the D800/800E have only minimal advantages over the 5D MKIII, simply because lenses still need to catch up. I can only judge those cameras by their respective Nikon sample images though, so there is room for error there.
Something that is even more subjective, is the look of an image. I much prefer the look of landscapes from full frame sensors over crop sensors, it isn't about dynamic range (if you beleive DxO, tee 7D has just as much DR as the 5D MKII anyway), it's more about the tones. To some degree, you can process a crop sensor to mimic a full frame one, but I don't believe it would be exact. It is this same difference in look that drives people twoards Zeiss lenses or the larger format cameras. Some consider it subjective and not valid in the real world, while others swear by them.
In short, you can capture good landscape photographs with any camera, but in the hands of an expert, the full frame camera will have the edge in most circumstances. Whether that difference is enough to make a real world difference largely depends on use and how that image will be viewed (and also the discernment of the viewer themselves). There is a reason why most professional landscape photographers use full frame or larger over crop, partly due to image quality, but also due to the actual resolution. While the difference between 18 MP and 21-22 MP is small, the latter does allow larger prints at 300 ppi, just like the 36 MP Nikons allow even larger prints, if you have the equipment to achieve it.

[nightsky87]

Popcorn was fine up to a certain point... But I have to agree with neuro on this.

We are discussing whether or not there is an IQ difference between sensors of similar resolution but different physical size. All other factors must be equal in testing this, or perceived differences could be due to something other than the physical sensor size.

I will grant that in some situations those other factors may be practically relevant. For example, when the choices were the 5D and 20D there were no good options for UWA on crop. But in judging any of this we must first determine what, if any, sensor differences exist absent all other influences.

Sorry, I'm not seeing a viable suggestion from you.  The sensors are different sizes - something must change for the resulting images to be framed the same.  The three options are crop the FF image, change the distance to the subject, or change the focal length.  The first defeats the purpose, the second changes the perspective of the image. Maybe I should have moved the camera, but arranged for the Museum of Fine Arts building to be moved closer to the foreground?  Or maybe I should have chosen perfectly flat subjects like test charts, obviating perspective?  Have I missed any options?

Cropping the image and doing side-by-side comparisons definitely isn't the way to go. Assuming that the camera position was not moved and the same lens and focal length were used, you have a 22MP (say the 5D3) at focal length X and an 18MP (i.e. 7D) at focal length 1.6X. So cropping this to produce the same view would mean that your APS-C image has 2.0736 times more pixels than the FF one. You could resample the images but that would give the APS-C image a slight advantage.

Besides, why would you have to deal with the crops? The fact is, the FF cameras out there (on Canon's end at least) have higher resolutions than the APS-C ones. Do our eyes crop the image to equalize the resolutions? Doesn't it make more sense to compare whole images?

Zooming in the lens like neuro makes sense since that at least preserves perspective without biasing the resolution significantly. In fact, if you think about it, some lenses perform poorer on the long end so it can actually be to the FF disadvantage anyway. When you downsample everything to a smaller resolution like neuro did, the difference between 18MP and 22MP should be negligible along with minor fluctuations in lens characteristics (assuming good lenses were used of course). At least that way, there is less bias.

Translation: you can't prove your point with hard evidence, so you're taking your ball and going home.

I was never trying to prove a point with hard evidence. Please go back and reread my initial post, and this time, pay very close attention to the part you ignored in your first response, the part where I stated that my observations were subjective and completely anecdotal.  Your treatment of it as if it were hard evidence was your first and biggest mistake, one that essentially renders the rest of the discussion moot.

Hard evidence was never neuro's point from his first post. But if we're discussing hard evidence, shouldn't the methodology at least be properly established?

FWIW, I am a card-carrying 'strict scientist', PhD and a day job in the field

This means nothing at all. All that matters is how a test is performed, and not who performed it. As a "card-carrying 'strict scientist' PhD' you should know and live by this.

What a classic and intriguing case of human bias and emotion in action, and from a "card-carrying 'strict scientist', PhD" no less. This is why I never automatically trust scientists even in their narrow fields of study, but treat their claims with the same critical eye as I would anyone else. No matter what is claimed in training or degrees (pieces of paper with ink), they are still human, and display all the classic flaws of human nature.

Okay... now that really ticked me off. While it is true that everything should be taken with a grain of salt, it is equally as true that scientists do a lot to get their degrees. These are "pieces of paper with ink" but are they inked so easily? Objectivity is one important aspect that is learned as a scientist. You don't publish works that scream out "BIASED". Those things rarely get past peer reviews.

FWIW, I am an academic researcher myself and I deal especially with image processing. And from what I see, neuro's post was anecdotal but technically sound. This would be much easier if there were a good ubiquitous no-reference quality metric but subjective testing is really the only option at this point and there can never really be a perfect "standard" procedure that everyone agrees on.

Personally, I haven't used a FF yet so I can't tell if there is really a difference. I've been using the 60D and I do feel that there's an extra quality that's missing. Maybe its my lenses. Or maybe its the effective DoF. Heck, maybe its even just me! I'll only know for sure when I get my hands on the 5D3 arriving soon.

[phixional ninja]

Zooming in the lens like neuro makes sense since that at least preserves perspective without biasing the resolution significantly.

As I understand it, zooming as described in Neuro's test preserves field of view, but not perspective.  I don't see that as a problem in the test though, I think it just demonstrates an (admittedly subjective) advantage of full frame over crop in this arena.  In the same way that full frame is seen as having an advantage in portraiture due to shallower depth of field for a given field of view, in landscape shooting I see it as advantageous to get greater distance compression for a given field of view.

If the shot I want is at 100mm on a crop sensor and 160mm on a full frame, that's an extra 60mm bringing the mountains in the distance closer.  At least for my own photography, I pretty much always want mountains to appear larger, it's rare that I have the opposite problem.

Of course that means the test is evaluating the overall shot quality, including composition, and not just image quality concerns such as noise and dynamic range, but again, I don't see that as a problem.

Apologies if I've mucked up any of the terminology, or am just plain wrong on anything.  Until a couple of weeks ago I wasn't even thinking about being able to get a full frame camera, and now I'm probably upgrading next week, so I've been doing... a lot... of reading and thinking about the subject.

[nightsky87]

As I understand it, zooming as described in Neuro's test preserves field of view, but not perspective.  I don't see that as a problem in the test though, I think it just demonstrates an (admittedly subjective) advantage of full frame over crop in this arena.  In the same way that full frame is seen as having an advantage in portraiture due to shallower depth of field for a given field of view, in landscape shooting I see it as advantageous to get greater distance compression for a given field of view.

Zooming changes the focal length being used which changes the FoV. But as long as the distance between the camera and subject are not changed, perspective is preserved. What this basically means is that you're effectively cutting out the outer areas of the image. The perspective change comes when you try to move the camera closer or farther away from the subject to maintain the FoV. In the of between FF and APS-C, you have to zoom the lens further on a FF to get the same frame and perspective as the APS-C. Without moving the camera, that is.

If the shot I want is at 100mm on a crop sensor and 160mm on a full frame, that's an extra 60mm bringing the mountains in the distance closer.  At least for my own photography, I pretty much always want mountains to appear larger, it's rare that I have the opposite problem.

As some people would say in these forums, if you don't really plan on printing your pictures and resolution is not that critical, you can crop a FF image in post. Since there isn't any perspective difference, you can get the APS-C reach by cropping an 8.64MP out of a 22.1MP image. That might sound like a big loss but that's still a good 3600x2400 image! For screen use, only a few displays can even match that kind resolution and that's certainly not the general public!

Apologies if I've mucked up any of the terminology, or am just plain wrong on anything.  Until a couple of weeks ago I wasn't even thinking about being able to get a full frame camera, and now I'm probably upgrading next week, so I've been doing... a lot... of reading and thinking about the subject.

I'm pretty much new to photography (started sometime around April this year) so I know the feeling. Hey, I might even be wrong with what I just said above! In which case, someone please correct/confirm what I just said.

[phixional ninja]

Zooming changes the focal length being used which changes the FoV. But as long as the distance between the camera and subject are not changed, perspective is preserved. What this basically means is that you're effectively cutting out the outer areas of the image. The perspective change comes when you try to move the camera closer or farther away from the subject.

Perhaps I'm not doing a good job of it, but I'm talking about this.  That example is moving the camera to preserve the size of the subject, but that's because it's dealing with a single camera and sensor size.  In my hypothetical though, a full frame and a crop camera in the same position would require different focal lengths to capture equivalent fields of view (as described by Neuro).  (Again, as I understand it) The longer focal length on the full frame camera would offer greater distance compression though, and so a shot with a far away mountain would appear to bring that mountain closer.

Edit to add:

As some people would say in these forums, if you don't really plan on printing your pictures and resolution is not that critical, you can crop a FF image in post. Since there isn't any perspective difference, you can get the APS-C reach by cropping an 8.64MP out of a 22.1MP image. That might sound like a big loss but that's still a good 3600x2400 image! For screen use, only a few displays can even match that kind resolution and that's certainly not the general public!

To clarify, I am not bemoaning the lesser apparent reach of full frame.  My hypothetical assumes available glass to get the same field of view on both sensors without cropping.  I'm less concerned with the overall size the mountain takes up in the frame (solvable with cropping), and more with its relative size compared to everything else (which cropping doesn't change).

[nightsky87]

Perhaps I'm not doing a good job of it, but I'm talking about this.  That example is moving the camera to preserve the size of the subject, but that's because it's dealing with a single camera and sensor size.  In my hypothetical though, a full frame and a crop camera in the same position would require different focal lengths to capture equivalent fields of view (as described by Neuro).  (Again, as I understand it) The longer focal length on the full frame camera would offer greater distance compression though, and so a shot with a far away mountain would appear to bring that mountain closer.

In that particular example, the camera was moved around as you mentioned. Focal length doesn't inherently change the perspective. Its the camera-subject distance that does this. So increasing focal length means your subject is larger. To maintain the same subject size, you step back which then changes perspective.

In the APS-C and FF discussion, the FF has a wider FoV due to the bigger sensor. So to get the same FoV as the crop sensor, you can zoom in (increase focal length) without moving the camera. No camera movement = same perspective and same compression.

To clarify, I am not bemoaning the lesser apparent reach of full frame.  My hypothetical assumes available glass to get the same field of view on both sensors without cropping.  I'm less concerned with the overall size the mountain takes up in the frame (solvable with cropping), and more with its relative size compared to everything else (which cropping doesn't change).

Undeniably, there will be differences in FF and APS-C if you try to maintain the same subject sizing with the same focal length. But assuming you're simply using the zoom of the lens, there will be no difference in relative sizing. So if what you're saying is that you have a lens and use it with the same focal length for both sensors, you have to move to maintain the FoV which then changes perspective.

[phixional ninja]

Perhaps I'm not doing a good job of it, but I'm talking about this.  That example is moving the camera to preserve the size of the subject, but that's because it's dealing with a single camera and sensor size.  In my hypothetical though, a full frame and a crop camera in the same position would require different focal lengths to capture equivalent fields of view (as described by Neuro).  (Again, as I understand it) The longer focal length on the full frame camera would offer greater distance compression though, and so a shot with a far away mountain would appear to bring that mountain closer.

In that particular example, the camera was moved around as you mentioned. Focal length doesn't inherently change the perspective. Its the camera-subject distance that does this. So increasing focal length means your subject is larger. To maintain the same subject size, you step back which then changes perspective.

In the APS-C and FF discussion, the FF has a wider FoV due to the bigger sensor. So to get the same FoV as the crop sensor, you can zoom in (increase focal length) without moving the camera. No camera movement = same perspective and same compression.

Well dang, it looks like you are totally right: http://www.scottbideauphotography.com/myths-about-lens-compression/

I've always heard it referred to as "lens compression" and read stuff like this which make it sound as though the focal length is what is responsible for the phenomenon.  Nope!

Thanks for prodding me into reading further.

[nightsky87]

Perhaps I'm not doing a good job of it, but I'm talking about this.  That example is moving the camera to preserve the size of the subject, but that's because it's dealing with a single camera and sensor size.  In my hypothetical though, a full frame and a crop camera in the same position would require different focal lengths to capture equivalent fields of view (as described by Neuro).  (Again, as I understand it) The longer focal length on the full frame camera would offer greater distance compression though, and so a shot with a far away mountain would appear to bring that mountain closer.

In that particular example, the camera was moved around as you mentioned. Focal length doesn't inherently change the perspective. Its the camera-subject distance that does this. So increasing focal length means your subject is larger. To maintain the same subject size, you step back which then changes perspective.

In the APS-C and FF discussion, the FF has a wider FoV due to the bigger sensor. So to get the same FoV as the crop sensor, you can zoom in (increase focal length) without moving the camera. No camera movement = same perspective and same compression.

Well dang, it looks like you are totally right: http://www.scottbideauphotography.com/myths-about-lens-compression/

I've always heard it referred to as "lens compression" and read stuff like this which make it sound as though the focal length is what is responsible for the phenomenon.  Nope!

Thanks for prodding me into reading further.

You're welcome. I actually used to think the same way myself. When I read about how focal length doesn't actually change perspective (I think it was some post by neuro), I went out to my balcony to confirm it.

[neuroanatomist]

Zooming changes the focal length being used which changes the FoV. But as long as the distance between the camera and subject are not changed, perspective is preserved. What this basically means is that you're effectively cutting out the outer areas of the image. The perspective change comes when you try to move the camera closer or farther away from the subject.

Perhaps I'm not doing a good job of it, but I'm talking about this.  That example is moving the camera to preserve the size of the subject, but that's because it's dealing with a single camera and sensor size.  In my hypothetical though, a full frame and a crop camera in the same position would require different focal lengths to capture equivalent fields of view (as described by Neuro).  (Again, as I understand it) The longer focal length on the full frame camera would offer greater distance compression though, and so a shot with a far away mountain would appear to bring that mountain closer.

Just to drive the point home, I pulled this from a recent post of mine on this issue.

Here's an example from the wiki page on perspective distortion, an example that illustrates some of this confusion:



You see the different focal lengths printed on the images, and you see the relative sizes of the two water bottles changing, and you think that focal length is the reason.  Even the text in the caption of that image suggests the effect is due to focal length.  But consider...the pink bottle is the same height in all three images - therefore, as the focal length is reduced, the camera must have been moved closer to achieve the same framing.  It's the movement of the camera, not the change in focal length, that results in the different perspectives.

Let me try to illustrate what I mean with a similar set of examples, but where I first vary just focal length but not distance, and then vary focal length and distance.  Since I prefer beverages other than water, I selected a different pair of subjects for a similar test.



The two bottles are 18" apart, and the distances in the image are measured from the sensor to the front bottle. All were shot at f/5.6.  As described above, it's normal for the bottle in front to look larger, that's perspective - objects that are further away look smaller, even though we know they are the same size.  Note that the bottle in front is the same height in all the images.

In the left column, the 50mm and 24mm shots were at the same distance as the 100mm images, and the images were cropped to match the framing of the 100mm image.  As you can see, the relative size of the rear bottle is the same in all the images.  The perspective is the same - the two bottles maintain the same relative size, despite the differing focal lengths.  The distance is the same, so the perspective is the same.

In the right column, which is equivalent to the water bottle shots from the wiki page, the 50mm and 24mm shots were taken at successively closer distances to the bottles, matching the framing to the 100mm shot by moving the camera.  As you can see, the relative size of the rear bottle gets smaller as the camera is moved closer.  Different perspective, because the distance is changing.  Comparing the side-by-side 50mm and 24mm images, you can see that with the same focal length but different subject distances, the perspective is different.

So, varying focal length alone, without changing distance, does not affect perspective.  As stated above (by me and others), perspective is determined by distance, and distance alone. 

'Telephoto' compression and 'wide angle' expansion distortion are the same phenomenon.  The reference to 'telephoto' and 'wide angle' there is misleading - it has nothing to do with the focal lengths, only the distance.  But longer lenses are usually used at longer distances, and wide lenses are usually used at closer distances (think of framing a person for a portrait), thus the erroneous association with a lens type when it's really the commonly used distances for those lenses that is the cause of both types of perspective distortion.

Ps. the Chocolate Stout was very good.