Floris' question about sharpening for canvas prompted me to ask a question I've been working through for some time. I've read several times that you sharpen differently for large prints than you do for small prints. However, nobody ever says what you actually do! So do you sharpen more for a large print than a small one or vice versa? I keep doing Google searches for an answer but so far haven't found a single website that gets into specifics--is this some kind of trade secret? Thanks--Larry
Posted by Dan Borzynski on July 17, 2009 at 11:11 PM
The short answer to your questions is that yes, depending on what you want to do with the output, you'll need different amounts of sharpening.
First - output to paper. I have found that the matte papers tend to have a softer output than glossy papers, so sometimes they benefit from a bit more sharpening. For both, very slightly oversharpening tends to work well to compensate for the fact that you lose a bit whe you output to a print. You don;t need a lot of oversharpening, just enough to put you are the border between enough and just a shade too much.
Outputing to a small medium, like web use, requires a lot less sharpening.
The subject matter also makes a difference. Most people don't want a sharp image of themselves, just an in-focus image (think pores, scars, acne marks, etc.). On the other hand, a detailed nature scene may require more sharpening to bring out detail you want visible.
Here is part of a response I gave a while back on sharpening for different size prints:
You have created a file at 12x18, 300 dpi, so the file is 3600x5400 pixels. We'll keep the 300 dpi constant. Now crop out a section of it that measures 8x12, or 2400x3600 pixels. To get the same sharpened end result, you would apply exactly the same sharpening settings.
Now suppose you started with an 8x12, or 2400 by 3600 pixels. To get your 12 by 18 at 3600x5400 pixels, you would need to interpolate the image up a bit. In any interpolation up method, whether in PS, Genuine fractals, or other program, some sharpness will be lost. So now to get to the same degree of sharpness as your smaller print, you have to sharpen as you normally would for that smaller print and add in a little bit more to take into account the losses in going up.
Now suppose you started with the 12x18 at 3600 x 5400. Instead of cropping to get an 8x12, now you want to actually reduce the size of the print by interpolation down to 8x12, or 2400 x 3600, rather than by cropping as in the first print. The amount of USM you need to apply will actually be less.
Why? Keeping in mind that it is 1 am here, hopefully this will make some sense. As you shrink a print, the small defects that lead your brain to believe an image is not sharp also shrinks, fooling your brain into believing that it really is sharp. (This is the notion of circles of confusion) The problems are still there, but you just can't see them. There is no sense in applying the same amount of sharpening as you did the larger print since that will bring them out.
Try again on shrinking the print. You stand 4 feet from a print. In your field of vision, this is your smaller print (the 8x12). It looks sharp, or at least you don't need to use a lot of USM. Now stand 2 feet away. The print is now bigger in your field of vision (the 12x18). Suddenly, your realize this needs more USM, because you can now see details, or their fuzzy presence better.
I especially notice the lesser amounts of USM when I am printing 4x5 scans. Even from a flatbed scan, there is a lot less sharpening needed at reasonable print sizes (under 16x20). It is not because I have a great flatbed, or that it is better than my 35 scanner, but it is because to get an 8x10, I am massively shrinking the file. This has the net effect of hiding the slightly blurred imperfections in the scan, so less USM is needed. Going the other way, from a 35mm to 16x20, greatly magnifies the defects in the scan. I would have to apply more USM to recover the lost sharpness from the magnified defects. (I know, this is too much to be asking from a 35 mm from - just humor me)
Last, sharpening tends to be a bit of art art as well. Ultimately, you are going to have to look at the final output and decide if it enough.
My usual starting point for Espon Premum Luster from a Canon 5D, file downsized to 8x12 at 240 dpi for USM is Amount 180, radius 1.4, threshold 0 and adjust adjust from there, depending on the subject matter.
Posted by David Taylor on July 17, 2009 at 11:15 PM
The first thing to realize is that there are no "fixed" numbers for sharpening - each image requires different sharpening technique, amounts, and values. For me, it's not so much about sharpening differently for larger prints - it's more important how you get your print that large. Correct resizing procedures are paramount, then worry about sharpening. Sharpening should be done at a "reasonable" magnification, like 50% or so. Depending on what your final use is - you may wish to slightly "oversharpen" to create halos, as they won't be apparent in the printed version.
Each image reacts a bit differently when enlarged/cropped and then requires unique sharpening. The best thing you can do is to turn the sharpening preview on and then fiddle with the controls until YOU are happy with the results.
Hope this helps somewhat.
And remember that global sharpening is very rarely needed.
Any reason you know of to sharpen a sky?
Practice selective sharpening on a separate layer using layer masks whenever possible.
This gives you the most control over your sharpening....where to put it and how much to use.
Do your "optimal" sharpening on your master working image at native resolution, and when making a print from a derivative of that file, resize first, then do a final sharpening after resizing/resampling.
I do huge mural sized silver halide prints for a living (partial living right now and growing fast) and really the best way to sharpen in my opinion is custom - still using the USM in PS competently. Mask off a few differing areas that need different kinds of sharpening (high frequency and lower frequency) make them into layers set to 50% opacity, then making sure your monitor is set to its highest resolution, view the image at one notch closer than "print size" in PS. I cant give all the tricks here away but you sharpen just a little crispy for monitor in all the layers or areas. Flatten one of the copies and leave one unflattened. Send it in, get it back and then if you over or under did any of the areas in the image all you have to do is adjust the opacity sliders for that particular layer or area!
BTW the book Real World Image Sharpening by the Late Bruce Fraser, even though a little old now, is packed with highly valuable info!
Discussion
7 repliesFirst - output to paper. I have found that the matte papers tend to have a softer output than glossy papers, so sometimes they benefit from a bit more sharpening. For both, very slightly oversharpening tends to work well to compensate for the fact that you lose a bit whe you output to a print. You don;t need a lot of oversharpening, just enough to put you are the border between enough and just a shade too much.
Outputing to a small medium, like web use, requires a lot less sharpening.
The subject matter also makes a difference. Most people don't want a sharp image of themselves, just an in-focus image (think pores, scars, acne marks, etc.). On the other hand, a detailed nature scene may require more sharpening to bring out detail you want visible.
Here is part of a response I gave a while back on sharpening for different size prints:
You have created a file at 12x18, 300 dpi, so the file is 3600x5400 pixels. We'll keep the 300 dpi constant. Now crop out a section of it that measures 8x12, or 2400x3600 pixels. To get the same sharpened end result, you would apply exactly the same sharpening settings.
Now suppose you started with an 8x12, or 2400 by 3600 pixels. To get your 12 by 18 at 3600x5400 pixels, you would need to interpolate the image up a bit. In any interpolation up method, whether in PS, Genuine fractals, or other program, some sharpness will be lost. So now to get to the same degree of sharpness as your smaller print, you have to sharpen as you normally would for that smaller print and add in a little bit more to take into account the losses in going up.
Now suppose you started with the 12x18 at 3600 x 5400. Instead of cropping to get an 8x12, now you want to actually reduce the size of the print by interpolation down to 8x12, or 2400 x 3600, rather than by cropping as in the first print. The amount of USM you need to apply will actually be less.
Why? Keeping in mind that it is 1 am here, hopefully this will make some sense. As you shrink a print, the small defects that lead your brain to believe an image is not sharp also shrinks, fooling your brain into believing that it really is sharp. (This is the notion of circles of confusion) The problems are still there, but you just can't see them. There is no sense in applying the same amount of sharpening as you did the larger print since that will bring them out.
Try again on shrinking the print. You stand 4 feet from a print. In your field of vision, this is your smaller print (the 8x12). It looks sharp, or at least you don't need to use a lot of USM. Now stand 2 feet away. The print is now bigger in your field of vision (the 12x18). Suddenly, your realize this needs more USM, because you can now see details, or their fuzzy presence better.
I especially notice the lesser amounts of USM when I am printing 4x5 scans. Even from a flatbed scan, there is a lot less sharpening needed at reasonable print sizes (under 16x20). It is not because I have a great flatbed, or that it is better than my 35 scanner, but it is because to get an 8x10, I am massively shrinking the file. This has the net effect of hiding the slightly blurred imperfections in the scan, so less USM is needed. Going the other way, from a 35mm to 16x20, greatly magnifies the defects in the scan. I would have to apply more USM to recover the lost sharpness from the magnified defects. (I know, this is too much to be asking from a 35 mm from - just humor me)
Last, sharpening tends to be a bit of art art as well. Ultimately, you are going to have to look at the final output and decide if it enough.
My usual starting point for Espon Premum Luster from a Canon 5D, file downsized to 8x12 at 240 dpi for USM is Amount 180, radius 1.4, threshold 0 and adjust adjust from there, depending on the subject matter.
Each image reacts a bit differently when enlarged/cropped and then requires unique sharpening. The best thing you can do is to turn the sharpening preview on and then fiddle with the controls until YOU are happy with the results.
Hope this helps somewhat.
Any reason you know of to sharpen a sky?
Practice selective sharpening on a separate layer using layer masks whenever possible.
This gives you the most control over your sharpening....where to put it and how much to use.
I do huge mural sized silver halide prints for a living (partial living right now and growing fast) and really the best way to sharpen in my opinion is custom - still using the USM in PS competently. Mask off a few differing areas that need different kinds of sharpening (high frequency and lower frequency) make them into layers set to 50% opacity, then making sure your monitor is set to its highest resolution, view the image at one notch closer than "print size" in PS. I cant give all the tricks here away but you sharpen just a little crispy for monitor in all the layers or areas. Flatten one of the copies and leave one unflattened. Send it in, get it back and then if you over or under did any of the areas in the image all you have to do is adjust the opacity sliders for that particular layer or area!
BTW the book Real World Image Sharpening by the Late Bruce Fraser, even though a little old now, is packed with highly valuable info!
Good luck.
Mark