There seems to be a misunderstanding among some members that the size of a crop hurts image quality. I have seen the comment several times on my and other's photos that "it is good detail for so small a crop". Under the specific size requirements here for posting large and small images, quite the reverse is true, questions of balance and composition aside.
If, for example, an image of 4288 pixels x 2848 pixels is resized to 1200 p x 797 p, a reduction and interpolation of pixels by a factor of 12.8 occurs. In other words each pixel has to do the work of almost 13. By the same token, if that same is first cropped to 36% of it's original size, then reduced, each reduced pixel then only has to interpolate for 4.6 missing pixels.
To put it succinctly, it is not the crop that hurts image quality because the pixels remain the same size. So a severe crop at the same size pixels has a net effect in resizing that is less than the net effect of a resize of the full frame.
Of course, sensible restrictions have to be considered, such as the necessity to include parts of the image which would get cut by the crop, and the rules of composition which to an extent dictate the shape and dimensions of the frame. Where it works for me is when my 400mm won't quite reach a bird on a uniform background. I will crop to a lesser size to get as much detail as possible on the resized image.
This has nothing to do with printing, just to posting on the web via downsized images. Of course, the exact opposite would hold for upsizing and printing. You would want as large a frame to begin with as you could get.
Posted by Mia McPherson on February 9, 2011 at 5:14 PM
Philip,
I'll be honest, I don't see much fine detail in either image you have presented on this post but I do see more in the one where the bird is smaller.
It is not a misunderstanding to say that the more an image is cropped the less detail there will be. I can see it clearly with these examples.
How much did you crop these images from the full frame in percentage? Are you taking that into consideration?
Posted by Richard Ditch on February 9, 2011 at 6:30 PM
Cropping and resizing are completely different functions, and you are confusing the effects.
Cropping always causes a loss of detail and resolution. The more pixels in the capture the more information potentially available, assuming that all pixels are equal (not true when you compare different classes of cameras such as a DSLR vs. a point-and-shoot).
Capturing the subject with the greatest number of pixels at the time of exposure will yield the most detailed image. Cropping the original capture so a smaller subject fills the cropped frame is not the same as filling the frame without cropping in terms of detail.
Cropping makes each pixel "work harder" to "cover" the fine detail in the subject. It also effectively reduces the resolution of the lens by "spreading" the lines per millimeter performance.
Resizing (downsizing) is not the same as cropping. Yes, it tosses out detail so you have less than you started with before downsizing. But cropping and then downsizing tosses out detail in both steps. The cropped image has less detail to start with and downsizing it only throws away more.
By cropping you are making each remaining pixel look bigger in a web post, but you are also making each bit of noise look big as well. And if you then are forced to use Noise Reduction to a greater extend to tame the noise you are also taking out more image detail.
If the goal is to show the most detail in a frame-filling image in a web post you will get the most detail if you fill the frame when you take the image. Cropping is not a good substitute to getting it right in the field.
I agree with Mia that neither of the two versions you have included here show a lot of detail. Here's a Hooded Merganser from 11/17/07 taken with my D200 and 300 with 2x.
Mia, the image was cropped to 1200 x n, then resized to 720 x n. The original crop was 36% of the full frame.
I have used math to indicate may skepticism of the statement. Despite what you offer in your rebuttal, the second image was a crop out of the 1200 image. It the smaller image has more quality, and I confess I cannot see it as you claim, it could be due to resizing sharper.
I also have been somewhat imprecise, which my be where you have misunderstood me. What I have tried to say (and I think succeeded) is that an image that is cropped strongly at the get go will yield better quality than one not cropped, insofar as an image resized to a specific dimension. I did not say it would be better period, just that it would be better than the uncropped image in terms strictly of resizing and interpolation between pixels. It, in the example, is a matter of a factor 12.8 reduction and interpolation vs a factor 4.6 reduction and interpolation.
A cropped image, including the sample image, is only reduced in number of pixels. The size of the pixels and their mathematical relationship to each other remains the same. A resized image makes all the remaining pixels have to work that much more to interpolate the quality of the pixels missing after resizing. That it seems to me is factual. What you say about the detail clarity of the two images is anecdotal and not based in mathematical fact. Prove me wrong mathematically and I will agree with you.
I am not arguing, as Richard put it, that cropping is the solution to image quality. He is right in that getting it right in the original frame is best. He is also wrong about my confusing cropping with resizing. What counts is two-fold: the number of pixels in the subject and and the amount of downsizing of the whole frame. The more pixels in the subject, regardless of the frame size, the better the image quality.
Posted by Mia McPherson on February 9, 2011 at 7:06 PM
Philip,
If you want to argue that large crops do not cause loss of detail you may want to use better images for your examples.
I don't need a mathmatical equation to see that the image where the bird is smaller in your file has more detail and that the one where the bird is larger has less. I only need my eyes. And they work well.
If I take an image that is full frame, sharp and in focus out of the camera and resize it to 1200 pixels I can gaurantee it will have more detail than the exact same image cropped to 25% and then resized to 1200 pixels. Not cropped FROM the other image, but cropped to 25% in ACR.
Posted by Richard Ditch on February 9, 2011 at 8:22 PM
I am not arguing, as Richard put it, that cropping is the solution to image quality. He is right in that getting it right in the original frame is best. He is also wrong about my confusing cropping with resizing. What counts is two-fold: the number of pixels in the subject and and the amount of downsizing of the whole frame. The more pixels in the subject, regardless of the frame size, the better the image quality.
I did not say that you are confusing cropping with resizing; I said you are confusing the effects of cropping with the effects of resizing.
I agree, showing the subject at larger scale in the posted version will best show detail captured in the image. But it also makes the noise in the image easier to see as well as making missed focus easier to see, and subject motion easier to see, so the actual image quality isn't necessarily better.
Posted by Brian Gatlin on February 9, 2011 at 10:14 PM
Phil, I think Rich has captured an important point in his last post. Magnification (showing a smaller part of the image at larger size) may reveal more of the detail in the original file, but will equally reveal the flaws in it.
If you fill the frame with a bird, and then resize it for web presentation, that resized file will virtually always (examined closely) show finer, better detail than if you filled 1/4 the frame with the bird, cropped to 25% of original, and resized that for web presentation. Hence the reasoning for statements such as, "pretty good detail, considering the crop."
You say:
an image that is cropped strongly at the get go will yield better quality than one not cropped, insofar as an image resized to a specific dimension.
If this were borne out in reality, why aren't we all taking distant shots and just cropping them to present at 100% pixel view, rather than wasting all that time trying to get close, fill the frame, and present the image (resized) at as close to 100% of full frame as possible?? I think you're getting lost in your numbers and losing track of what's really happening?
If this were borne out in reality, why aren't we all taking distant shots and just cropping them to present at 100% pixel view, rather than wasting all that time trying to get close, fill the frame, and present the image (resized) at as close to 100% of full frame as possible?? I think you're getting lost in your numbers and losing track of what's really happening?
Brian, I don't think I am offering a better formula for image quality through cropping. I am simply saying a crop of a small subject is better than no crop, because of the ratio of crop size to final image. I do not recommend it as a means of processing sorry images to maximize quality. My original stipulation was based on what I thought (and still do) was a misunderstanding about the effects of cropping on image quality.
I agree, showing the subject at larger scale in the posted version will best show detail captured in the image. But it also makes the noise in the image easier to see as well as making missed focus easier to see, and subject motion easier to see, so the actual image quality isn't necessarily better.
I actually agree with Richard. But turn that on it's head—does showing the image at a smaller scale in the posted version make "noise in the image harder to see as well as making missed focus harder to see, and subject motion harder to see, so the actual image quality is necessarily better?"
Posted by Richard Ditch on February 10, 2011 at 10:48 AM
I actually agree with Richard. But turn that on it's head—does showing the image at a smaller scale in the posted version make "noise in the image harder to see as well as making missed focus harder to see, and subject motion harder to see, so the actual image quality is necessarily better?"
Composing to use as much of the frame as possible and minimizing cropping maximizes image quality. If you are concerned about quality then you need to address every link in the chain. That means avoiding higher ISO settings that introduce noise (different for each camera model), exposing properly to avoid blown highlights and the need to increase exposure in raw conversion that introduces more noise, selecting the right combination of shutter speed and aperture to stop motion and provide sufficient depth of field, avoiding harsh light by shooting early or late in the day, accurately focusing on the subject (or the subject's eye), properly supporting the camera and lens to prevent introducing unwanted blur, shooting in the highest quality setting available in the camera (e.g. raw). It also means care in raw conversion: adjusting the histogram to get the most information out of the capture, avoiding the clarity slider that can introduce haloing in the image. And care in post processing: avoiding excessive saturation, minimizing noise reduction that robs detail, removing content with cloning that should have been avoided at capture, relocating content that should have been dealt with at capture.
Yes, keeping the subject smaller in the frame by avoiding cropping can make the overall image quality look better by not emphasizing problems in the image. That's one reason why anyone reviewing a posted image should look at the larger 1200 pixel version of any post as the larger image will help bring out such issues.
Posted by David Bostedo on February 10, 2011 at 3:12 PM
Phil - I think maybe the comparison you actually want to make is this (although I'm making some guesses here - sorry if I'm off track) :
Image Type 1 : The area of the bird comprising 50% of the 720 pix frame; arrived at by shooting the full frame with the bird filling 50% of the area of the original image.
Image Type 2 : The area of the bird comprising 50% of the 720 pix frame; arrived at by shooting the full frame with the bird filling 25% of the area of the original image, cropping to increase the size of the bird, and re-sizing.
That way whether you had to crop or not (i.e. the pixels available before resizing) is really the only variable.
Comparing these two, it will almost always be plain that image 1 is "better" in some way. The fact that cropping hurts image quality is absolutely true under most definitions of "quality". I think when people comment on "nice detail considering the crop", they generally mean that for being an image of type 2, you did a good job making it look like type 1.
Otherwise, you're showing the subject at two different sizes, as in your example. This introduces all sorts of vagaries and variables as far as what comprises a "quality" image and makes things much more subjective. Maybe some people prefer a less detailed but larger bird - making what detail is there easier to see. Maybe some prefer a more detailed overall image but with a smaller bird in frame.
In that case with your images above, you could probably have a lot of people look at them, poll them as to which was higher "quality" and get a variety of answers. I'm guessing that for some images, the original will be preferred, and for some, the cropped image will still have good detail and low noise and be preferable. It depends on people's preferences and on how much things are cropped.
A couple of side notes :
To put it succinctly, it is not the crop that hurts image quality because the pixels remain the same size.
This is only true if you then display those pixels at the same size relative to the original image. (Most people crop and then display less pixels in the same area as the original image.) Pixels technically have zero size - they are just bits of data. But how they are displayed is important. Cropping hurts image quality in the same way showing 100% views hurts image quality. You're using less pixels in the same space.
And to address the math : yes you're interpolating less after you crop. But you've already thrown a ton of information away by having the bird smaller in the frame and cropping. Also, I don't think that there is really a direct relationship between "quality" and amount of interpolation since it seems to be very image dependent. I'm not sure you can generalize - a 5x factor in one image might look great, and a 2x factor in another look terrible. There are too many variables.
I realize the images are unfortunate examples. That being said, the math exists and the possibility exists, and fortune ought to favor the least reduction. That it quite often doesn't because of other variables is true also, but I never said it wasn't. Just that the generalization of "quality despite crop" is as subject to specifics as my examples.
A further note, and I will bid this goodbye—no one argues, least of all me, that display after cropping at the same size as the original usually hurts image quality. We don't do that; we display at 720 pixels x n and 1200 pixels by n. That is where apparent quality is seen and critiqued.
Discussion
12 repliesIf, for example, an image of 4288 pixels x 2848 pixels is resized to 1200 p x 797 p, a reduction and interpolation of pixels by a factor of 12.8 occurs. In other words each pixel has to do the work of almost 13. By the same token, if that same is first cropped to 36% of it's original size, then reduced, each reduced pixel then only has to interpolate for 4.6 missing pixels.
To put it succinctly, it is not the crop that hurts image quality because the pixels remain the same size. So a severe crop at the same size pixels has a net effect in resizing that is less than the net effect of a resize of the full frame.
Of course, sensible restrictions have to be considered, such as the necessity to include parts of the image which would get cut by the crop, and the rules of composition which to an extent dictate the shape and dimensions of the frame. Where it works for me is when my 400mm won't quite reach a bird on a uniform background. I will crop to a lesser size to get as much detail as possible on the resized image.
This has nothing to do with printing, just to posting on the web via downsized images. Of course, the exact opposite would hold for upsizing and printing. You would want as large a frame to begin with as you could get.
I'll be honest, I don't see much fine detail in either image you have presented on this post but I do see more in the one where the bird is smaller.
It is not a misunderstanding to say that the more an image is cropped the less detail there will be. I can see it clearly with these examples.
How much did you crop these images from the full frame in percentage? Are you taking that into consideration?
Cropping always causes a loss of detail and resolution. The more pixels in the capture the more information potentially available, assuming that all pixels are equal (not true when you compare different classes of cameras such as a DSLR vs. a point-and-shoot).
Capturing the subject with the greatest number of pixels at the time of exposure will yield the most detailed image. Cropping the original capture so a smaller subject fills the cropped frame is not the same as filling the frame without cropping in terms of detail.
Cropping makes each pixel "work harder" to "cover" the fine detail in the subject. It also effectively reduces the resolution of the lens by "spreading" the lines per millimeter performance.
Resizing (downsizing) is not the same as cropping. Yes, it tosses out detail so you have less than you started with before downsizing. But cropping and then downsizing tosses out detail in both steps. The cropped image has less detail to start with and downsizing it only throws away more.
By cropping you are making each remaining pixel look bigger in a web post, but you are also making each bit of noise look big as well. And if you then are forced to use Noise Reduction to a greater extend to tame the noise you are also taking out more image detail.
If the goal is to show the most detail in a frame-filling image in a web post you will get the most detail if you fill the frame when you take the image. Cropping is not a good substitute to getting it right in the field.
I agree with Mia that neither of the two versions you have included here show a lot of detail. Here's a Hooded Merganser from 11/17/07 taken with my D200 and 300 with 2x.
I have used math to indicate may skepticism of the statement. Despite what you offer in your rebuttal, the second image was a crop out of the 1200 image. It the smaller image has more quality, and I confess I cannot see it as you claim, it could be due to resizing sharper.
If you want the particulars, the image is posted in the avian gallery at http://www.naturephotographers.net/imagecritique/ic.cgi?a=vp&pr=166359&CGISESSID=5852eefb56d6ce4a6180a91c39bdd1a7&u=4575.
I also have been somewhat imprecise, which my be where you have misunderstood me. What I have tried to say (and I think succeeded) is that an image that is cropped strongly at the get go will yield better quality than one not cropped, insofar as an image resized to a specific dimension. I did not say it would be better period, just that it would be better than the uncropped image in terms strictly of resizing and interpolation between pixels. It, in the example, is a matter of a factor 12.8 reduction and interpolation vs a factor 4.6 reduction and interpolation.
A cropped image, including the sample image, is only reduced in number of pixels. The size of the pixels and their mathematical relationship to each other remains the same. A resized image makes all the remaining pixels have to work that much more to interpolate the quality of the pixels missing after resizing. That it seems to me is factual. What you say about the detail clarity of the two images is anecdotal and not based in mathematical fact. Prove me wrong mathematically and I will agree with you.
I am not arguing, as Richard put it, that cropping is the solution to image quality. He is right in that getting it right in the original frame is best. He is also wrong about my confusing cropping with resizing. What counts is two-fold: the number of pixels in the subject and and the amount of downsizing of the whole frame. The more pixels in the subject, regardless of the frame size, the better the image quality.
If you want to argue that large crops do not cause loss of detail you may want to use better images for your examples.
I don't need a mathmatical equation to see that the image where the bird is smaller in your file has more detail and that the one where the bird is larger has less. I only need my eyes. And they work well.
If I take an image that is full frame, sharp and in focus out of the camera and resize it to 1200 pixels I can gaurantee it will have more detail than the exact same image cropped to 25% and then resized to 1200 pixels. Not cropped FROM the other image, but cropped to 25% in ACR.
I did not say that you are confusing cropping with resizing; I said you are confusing the effects of cropping with the effects of resizing.
I agree, showing the subject at larger scale in the posted version will best show detail captured in the image. But it also makes the noise in the image easier to see as well as making missed focus easier to see, and subject motion easier to see, so the actual image quality isn't necessarily better.
If you fill the frame with a bird, and then resize it for web presentation, that resized file will virtually always (examined closely) show finer, better detail than if you filled 1/4 the frame with the bird, cropped to 25% of original, and resized that for web presentation. Hence the reasoning for statements such as, "pretty good detail, considering the crop."
You say:
If this were borne out in reality, why aren't we all taking distant shots and just cropping them to present at 100% pixel view, rather than wasting all that time trying to get close, fill the frame, and present the image (resized) at as close to 100% of full frame as possible?? I think you're getting lost in your numbers and losing track of what's really happening?
Brian, I don't think I am offering a better formula for image quality through cropping. I am simply saying a crop of a small subject is better than no crop, because of the ratio of crop size to final image. I do not recommend it as a means of processing sorry images to maximize quality. My original stipulation was based on what I thought (and still do) was a misunderstanding about the effects of cropping on image quality.
I actually agree with Richard. But turn that on it's head—does showing the image at a smaller scale in the posted version make "noise in the image harder to see as well as making missed focus harder to see, and subject motion harder to see, so the actual image quality is necessarily better?"
Composing to use as much of the frame as possible and minimizing cropping maximizes image quality. If you are concerned about quality then you need to address every link in the chain. That means avoiding higher ISO settings that introduce noise (different for each camera model), exposing properly to avoid blown highlights and the need to increase exposure in raw conversion that introduces more noise, selecting the right combination of shutter speed and aperture to stop motion and provide sufficient depth of field, avoiding harsh light by shooting early or late in the day, accurately focusing on the subject (or the subject's eye), properly supporting the camera and lens to prevent introducing unwanted blur, shooting in the highest quality setting available in the camera (e.g. raw). It also means care in raw conversion: adjusting the histogram to get the most information out of the capture, avoiding the clarity slider that can introduce haloing in the image. And care in post processing: avoiding excessive saturation, minimizing noise reduction that robs detail, removing content with cloning that should have been avoided at capture, relocating content that should have been dealt with at capture.
Yes, keeping the subject smaller in the frame by avoiding cropping can make the overall image quality look better by not emphasizing problems in the image. That's one reason why anyone reviewing a posted image should look at the larger 1200 pixel version of any post as the larger image will help bring out such issues.
Image Type 1 : The area of the bird comprising 50% of the 720 pix frame; arrived at by shooting the full frame with the bird filling 50% of the area of the original image.
Image Type 2 : The area of the bird comprising 50% of the 720 pix frame; arrived at by shooting the full frame with the bird filling 25% of the area of the original image, cropping to increase the size of the bird, and re-sizing.
That way whether you had to crop or not (i.e. the pixels available before resizing) is really the only variable.
Comparing these two, it will almost always be plain that image 1 is "better" in some way. The fact that cropping hurts image quality is absolutely true under most definitions of "quality". I think when people comment on "nice detail considering the crop", they generally mean that for being an image of type 2, you did a good job making it look like type 1.
Otherwise, you're showing the subject at two different sizes, as in your example. This introduces all sorts of vagaries and variables as far as what comprises a "quality" image and makes things much more subjective. Maybe some people prefer a less detailed but larger bird - making what detail is there easier to see. Maybe some prefer a more detailed overall image but with a smaller bird in frame.
In that case with your images above, you could probably have a lot of people look at them, poll them as to which was higher "quality" and get a variety of answers. I'm guessing that for some images, the original will be preferred, and for some, the cropped image will still have good detail and low noise and be preferable. It depends on people's preferences and on how much things are cropped.
A couple of side notes :
This is only true if you then display those pixels at the same size relative to the original image. (Most people crop and then display less pixels in the same area as the original image.) Pixels technically have zero size - they are just bits of data. But how they are displayed is important. Cropping hurts image quality in the same way showing 100% views hurts image quality. You're using less pixels in the same space.
And to address the math : yes you're interpolating less after you crop. But you've already thrown a ton of information away by having the bird smaller in the frame and cropping. Also, I don't think that there is really a direct relationship between "quality" and amount of interpolation since it seems to be very image dependent. I'm not sure you can generalize - a 5x factor in one image might look great, and a 2x factor in another look terrible. There are too many variables.
A further note, and I will bid this goodbye—no one argues, least of all me, that display after cropping at the same size as the original usually hurts image quality. We don't do that; we display at 720 pixels x n and 1200 pixels by n. That is where apparent quality is seen and critiqued.