Cropping FF doesn't give sharper images

[DanK]

In a couple of recent threads, there was a debate (in the context of macro photography) about whether all else equal, cropping a FF image to the size of a crop image will provide more detail even if the resulting pixel count is lower. I reasoned that if the image is displayed at a large enough size (or cropped enough) that resolution differences are visible, the crop should provide more detail. Two folks said that I was wrong, but I couldn't see why, so I tried it. My test suggests that I was right.

I put a copy of Bob Atkins resolution test chart on the wall, 8.5 x 11 (roughly A4). Stupidly, I used regular copier paper; the results would have been clearer had I used a smoother paper. I put my 100mm L macro on my 7D and moved it to the distance that had the chart filling the frame. I took one shot and then took another with the same lens on my 5DIII from the same distance. I then cropped the 5DIII image to show the same content as the 7D image and started blowing them up.

The difference is very small when one uses rough-surface paper, but it's visible: the 7D image, which has more pixels, has more detail.

It was clearer in Lightroom, which allowed me to constrain the cropped images to be the same size. When I uploaded them to show them here, they were no longer the same physical size; the software resizes the 5DIII image, which contains fewer pixels, to be correspondingly smaller. I was able to get them reasonably similar by fiddling with the links in Smugmug, but at the So in the images I will post here, it is harder to see the difference; the smaller size of the 5DIII image obscures a good bit of the lack of sharpness of line edges. By fiddling with Smugmug, I think I got links that will show them at almost the same size, at the cost of making them small.

Despite that, if you look at the small number 10 and the lines surrounding it, you can see the difference in sharpness, despite the problem with rough paper.

5DIII:



7D:

[NorthernFocus]

Thanks for posting your test results, Dan. I've argued the same thing forever. It only stands to reason after all. Where detail is of prime interest, all things being equal I'll take more pixels on target every time. The only exception is if high ISO is necessary and the higher pixel count option has really poor noise performance. Cleaning up the noise can destroy the advantage of the additional pixels.

Interesting that you posted this. I was contemplating posting a thread on this topic, albeit without supporting test data. After all, purely philosophical discussions are much more lively.

I think you'll find than many of the proponents of the advantages of cropping lower pixel count images are those who are still shooting cameras that are three or four generations obsolete and trying to rationalize doing so.

[Manfred M]

If the pixel pitch in both cameras is the same, there really shouldn't be any difference. In general the full frame cameras will have larger pixels, but as the image is not magnified as much, the output will generally be cleaner.

In your example, the 7D with its 18MP (4.29 µm pitch) sensor and the 5D Mk III with its 22MP sensor (6.22 µm pitch), I would expect the crop frame to have higher resolution when one crops the FF shot to crop frame size.

If both cameras had the same pixel pitch, there would be no difference, but as crop frame cameras usually have smaller pixel pitch than a full frame camera, they will tend to give one a sharper image versus a FF crop.

[DanK]

Dan,

I agree. That's why I did the test; the reasoning seemed clear. My crop sensor camera is a generation or two behind my FF, and the combination of that and the smaller pixels gives me low-light performance that quite a bit inferior to the 5D III. However, most of my macro work is at low ISO. So in practice, I use the crop for bugs--it gives me more pixels on the subject, it's lighter and smaller (an advantage when struggling with position and focus), and I usually shoot at 100 or 200, almost never above 400. For indoor flower stuff, I usually use the 5D III because magnification isn't usually a limitation.

[DanK]

If the pixel pitch in both cameras is the same, there really shouldn't be any difference. In general the full frame cameras will have larger pixels, but as the image is not magnified as much, the output will generally be cleaner.

In your example, the 7D with its 18MP (4.29 µm pitch) sensor and the 5D Mk III with its 22MP sensor (6.22 µm pitch), I would expect the crop frame to have higher resolution when one crops the FF shot to crop frame size.

If both cameras had the same pixel pitch, there would be no difference, but as crop frame cameras usually have smaller pixel pitch than a full frame camera, they will tend to give one a sharper image versus a FF crop.

Manfred,

your second and third paragraphs are what I had suggested in earlier posts. However, this came up in the context of macro work, and at maximum magnification, the image of the subject--not the image of the entire frame--is identical regardless of sensor size. Therefore, the image of the subject has to be magnified by the same amount to be displayed at any given size.

Dan

[dje]

I agree that the image from the crop camera should be sharper due to the smaller pixel size (in the context of Dan's example). There are two main reasons

Smaller pixels have a higher MTF at any given line frequency

Smaller pixels have a weaker AA filter (the AA filter blur is usually about 0.3 - 0.4 of the pixel size)

Dave

[george013]

Difficult comperisations. If you want to compare 2 shots you must be sure that all the variables are equal, except for that one you want to investigate. And even then you must be carefull.
Using a FF camera and shoot once in FF and once in crop mode will leave everything equal: pixelsize, aperture,focal length. I call this the optical aproach, the game between lens and sensor. It's the same as just cropping afterwards.
But then you've what I call the photographical aproach: the print. The crop has to be enlarged 1.6 times more as the ff. For Nikon 1.5. That has influence on the dof.

It must be possible to make two matrixes with a comparisation between a ff and a crop camera: one based on the same framing and one on a different framing. And showing all the other variables.

I think dx0 is comparing the image quality as you want. You can select your lens and compare it on your 2 camera's.

https://www.dxomark.com/Lenses/Canon...-Mark-III__795
https://www.dxomark.com/Lenses/Canon...on-EOS-7D__619

George

[DanK]

If you want to compare 2 shots you must be sure that all the variables are equal, except for that one you want to investigate.

Yes, if you read my post carefully, I did precisely that. I held all factors other than camera body constant.

And even then you must be carefull.

I was.

It's the same as just cropping afterwards.

Again, read the post. That's what I did. I cropped a full-frame image after capture. I didn't shoot an FX camera in crop mode. (I don't own a camera that will do that.)

But then you've what I call the photographical aproach: the print. The crop has to be enlarged 1.6 times more as the ff. For Nikon 1.5.

That's what I did and what I intended, except that the issue is "display," not print, as I used a computer display rather than a print. Note that I kept the display type and, to the extent I could, display size constant. A fundamental principle in the science of measurement is that a measure can only be evaluated with respect to a specific use or conclusion. The question I was addressing was the sharpness of displays, and I showed that under these specific conditions, the display from the higher-density crop sensor camera is sharper.

I think dx0 is comparing the image quality as you want. You can select your lens and compare it on your 2 camera's.

Read the post. That is what I did. The comparison I made is the one I wanted.

That has influence on the dof.

Irrelevant to my comparison. the chart I photographed was flat.

[george013]

Yes, if you read my post carefully, I did precisely that. I held all factors other than camera body constant.



I was.



Again, read the post. That's what I did. I cropped a full-frame image after capture. I didn't shoot an FX camera in crop mode. (I don't own a camera that will do that.)



That's what I did and what I intended, except that the issue is "display," not print, as I used a computer display rather than a print. Note that I kept the display type and, to the extent I could, display size constant. A fundamental principle in the science of measurement is that a measure can only be evaluated with respect to a specific use or conclusion. The question I was addressing was the sharpness of displays, and I showed that under these specific conditions, the display from the higher-density crop sensor camera is sharper.



Read the post. That is what I did. The comparison I made is the one I wanted.



Irrelevant to my comparison. the chart I photographed was flat.

If it's just and only comparing a crop with ff, then just crop a crop out off a ff. That's what a ff camera is doing. Everything is the same, except the size. But when showing on display or print the crop has to be enlarged more in metric units.
Using 2 different camera's is introducing different variables per definition.
I'm not sure seeing difference in sharpness. There's a slight difference in exposure and contrast.

George

[DanK]

If it's just and only comparing a crop with ff, then just crop a crop out off a ff. That's what a ff camera is doing

I fully understand. It's just not relevant to my post.

But when showing on display or print the crop has to be enlarged more in metric units.

No, in my comparison, the enlargement to display size is identical because the full-frame capture has been cropped down to the size of the crop-sensor camera image.

Using 2 different camera's is introducing different variables per definition.

Yes, unavoidably, but the only relevant one is the sensor. The settings were identical, and the images were raw files, not processed further by the camera.

I'm not sure seeing difference in sharpness.

I do, but it isn't a large difference, and it is partly obscured by the rough texture of the paper

There's a slight difference in exposure and contrast.

yes, despite identical camera settings, there is a slight difference in exposure. What you are perceiving as a difference in contrast, I believe, is actually the difference in sharpness. No contrast adjustments were applied to either raw image.

I'm signing off. The answer seems clear to me.

[xpatUSA]

What does "crop" mean in this thread?

Is there any re-sizing involved?

[DanK]

What does "crop" mean in this thread?

Is there any re-sizing involved?

What I meant is simply cropping the image from the FF so that it contains only the material one captures with the crop sensor camera.

Given that the starting point was the lower pixel density of most FF cameras, the cropped FF image necessarily contains fewer pixels than the crop camera image. That is the whole point of the exercise. Therefore, yes, one has to resize the cropped FF more to reach any given display size--and the same display size is also part of the exercise, since the question was which will give you a sharper image once the desired display is produced. that's why I mentioned that I had to fiddle to get the online images to be similar in size.

[xpatUSA]

What I meant is simply cropping the image from the FF so that it contains only the material one captures with the crop sensor camera.

Thanks, Dan, that's what I hoped was meant.

Given that the starting point was the lower pixel density of most FF cameras, the cropped FF image necessarily contains fewer pixels than the crop camera image. That is the whole point of the exercise. Therefore, yes, one has to resize the cropped FF more to reach any given display size--and the same display size is also part of the exercise, since the question was which will give you a sharper image once the desired display is produced. that's why I mentioned that I had to fiddle to get the online images to be similar in size.

Thanks again, and very well put re: display size. Elsewhere, comparison posts are rife (F****n vs. Bayer, usually) and the output medium rarely gets a mention except for a certain pedant who always mentions his P242W monitor but the implication of the which mention is rarely grasped there.

[Tronhard]

I admit arriving late at this thread but wondered if this article was of any use...
http://www.robsphotography.co.nz/cro...advantage.html

[xpatUSA]

For my own work in comparing two cameras, I try to stay as input-referred as possible. For that reason, I use raw files opened by RawDigger and compare raw exported images with no conversion at all. I hasten to mention that my interest is only in detail, micro-contrast, edge sharpness etc.

Others go to the extreme of raw conversion, post-processing and printing out, say 8x10", and then using those prints to declare that this camera is better than that. An output-referred POV.

Pardon my cynicism . .

[pschlute]

The point here, as has been pointed out but is worth mentioning again, is that it is not the sensor physical size (mm x mm) that makes the difference, but the pixel density. If I take a picture with my camera in FF mode and crop it, then compare to a picture taken in crop-mode, they are identical. Same pixel density in both captures.

If I take a picture in FF mode (36MP), crop it, and compare it with a 6MP crop sensor image, then the FF camera has better resolution. Do the same comparison with a 24MP crop sensor camera and the latter has the advantage.

Of course this assumes that resolution obtained from a specific pixel density is equal across camera models and brands. I am sure as technology advances we may find that this simple equation is not necessarily the case, but is a good basis to work on at present.

[Tronhard]

Hmmm... I agree about pixel density, as long as that is defined in absolute terms of pixels/unit area (e.g. per mm sq).

For example I have a Nikon Df that has a FF frame sensor and a total pixel count of 16MP
I also have a Canon Powershot SX60HS which also has a pixel count of 16MP, but it has a minute sensor (1/2.3" 6.17 x 4.55 mm) in comparison to the FF (36 x 23.9 mm)
Thus the pixels/sensor may be the same but the pixels/unit area are definitely not.
While the Powershot is an excellent camera in its space, it will develop noise quite quickly because of the small pixel sizes (it's actually pretty well if you give it enough light) as the reduced pixel size creates a higher signal:noise ratio. Larger pixels should also provide a wider dynamic range: more detail in both the bright and the dark areas of a photo.

[pschlute]

That is interesting Trev, and unlike a aps-c vs FF comparison we cannot swap lenses between to formats to give an exact comparison. But if you compare as best you can, how does the resolution appear between the two ? Is it a case that all pixels are not created equal ?

I knew I shouldn't have got involved in a form discussion after half a bottle of wine and a few beers

[Tronhard]

Well resolution is often in the eye of the beholder - by which I mean it makes a difference depending on what medium and size you decide to create your output, and where you place yourself in relation to the result. (Pixel peepers stick their noses to the image, most people view images from much further back in reality). The image disparity may not make much difference if you are posting on a web page, but if you create a large digital or print image the difference is plain.

I guess the only reasonable way to do that is to look at it 1:1 for each unit. If I do so the resolution of the smaller sensor is definitely inferior to that of the larger one. Which makes total sense because each pixel in the larger sensor has more light gathering capability, and (as mentioned) has a higher signal / noise ratio, which means less better tonal range.

There is an article on this subject: https://www.photocrati.com/large-vs-...pros-and-cons/ that you might find interesting.

Also this DRP item: https://www.dpreview.com/articles/53...sizes-on-noise

[xpatUSA]

Well resolution is often in the eye of the beholder

You got that right . .

I guess the only reasonable way to do that is to look at it 1:1 for each unit. If I do so, the resolution of the smaller sensor is definitely inferior to that of the larger one.

If the smaller sensor has a smaller pixel pitch, would that not be a superior resolution, i.e. more lp/mm?

Which makes total sense because each pixel in the larger sensor has more light gathering capability, and (as mentioned) has a higher signal / noise ratio, which means less better tonal range.

Didn't quite get that, Trev, sorry.