Downsizing a big image

[chauncey]

I've photomerged an image considerably larger than necessary (18000 x 12000 dpi)
Don't want to crop anything...want to end up with a 20 x 30 inch print.
Would I increase sharpness by merely downsizing to 9000 x 6000 pixels?

Am I being clear with my question?

[Shadowman]

Tutorials I've read suggest that image will appear sharper but best to: resample image and also apply USM to bring back any sharpness lost in the downsizing.

[IzzieK]

I am with John here...I always do a slight sharpening after downsizing...

[george013]

I've photomerged an image considerably larger than necessary (18000 x 12000 dpi)
Don't want to crop anything...want to end up with a 20 x 30 inch print.
Would I increase sharpness by merely downsizing to 9000 x 6000 pixels?

Am I being clear with my question?

I once did a test with a photo. It contained text that was shot at an angle. At 100% on the screen it looked sharp. Downsizing it became worse. After sharpening even more worse.
I did print it twice. Once downsized to the right size and once not downsized, let the printer do it. In my case the last one was the better. I don't have an explanation for it yet.
Check the printing office what resolution they use for that size. Most unlikely it's 300dpi.

George

[chauncey]

I must have a fundamental misunderstanding event...it would seem that if you compressed a
300 dpi 60 x 40 print down to 300 dpi 30 x 20 print, it would inherently be considerably sharper.

[Manfred M]

How many times have you looked at an image on your camera's screen and found that it looked just fine, but when you opened it on your larger computer screen and found that it was a bit soft?

This is the same thing that happens when you downsize an image. It can look sharp on the computer screen, but zoom into 100% and it will be obvious it still needs to be sharpened. The same effect is occurring in both scenarios. The resolution of the device we are looking at seems to have sharpened the image, but in fact is really has not. You still have to properly sharpen a downsized image as the image can appear sharper, but really is not.

[george013]

How many times have you looked at an image on your camera's screen and found that it looked just fine, but when you opened it on your larger computer screen and found that it was a bit soft?

Never. The first part.

I.m just trying to say that downsizing to the print resolution is not always better as let the printer do the
downsizing. That was the question.
And secund that the print resolution won't be 300 for that size.
George

[Manfred M]

Never. The first part.

I.m just trying to say that downsizing to the print resolution is not always better as let the printer do the
downsizing. That was the question.
And secund that the print resolution won't be 300 for that size.
George

I don't read the question quite the same way that you do. George.

The question is does downsizing the image from 18000 x 12000 pixels to 9000 x 6000 pixels sharpen the image. The answer is clearly "no", even though the image might appear to be sharper on the screen when the whole image is viewed on the computer screen. In both cases, there could be the appearance of sharpening as the image has to be downsized to be viewed on the computer screen. This is totally independent of what happens during printing. I have done and printed a number of panos and always done output sharpening at 100% of the final print size, in addition to any other sharpening processes I have done during my workflow.

I don't know what William is planning to do to prepare the image for printing, but a 9000 x 6000 pixel image implies a 30" x 20" final print size on a printer that has a native 300 dpi resolution. Yes, the printer driver will take care of any resizing / scaling, as required. If the final print is done on a Canon or HP printer, in theory, no further action is required, but if an Epson printer is used (360 dpi native resolution), the printer driver would have to rescale the image appropriately for a 30" x 20" final image.

[Mike Buckley]

it would seem that if you compressed a
300 dpi 60 x 40 print down to 300 dpi 30 x 20 print, it would inherently be considerably sharper.

The cheapest and easiest proof that that is not true is to make it happen by preparing both image files. Just make sure you view both of them at 100% for valid comparisons. Otherwise, you risk being fooled. As an example, downsizing an image to the size of a postage stamp will almost always look sharp without adding any sharpening even if the image was captured out of focus. That's because the final image is so small the lack of sharpness can't be detected by the naked eye.

[dem]

Similar to George I thought the question was: for a 20x30 printout from a 18,000 x 12,000 file do you
A) downscale the image manually to match the printer resolution (300 dpi?) before printing "actual size" or
B) just send the file to print and let the printer "fit to page"?

[Manfred M]

Similar to George I thought the question was: for a 20x30 printout from a 18,000 x 12,000 file do you
A) downscale the image manually to match the printer resolution (300 dpi?) before printing "actual size" or
B) just send the file to print and let the printer "fit to page"?

If this is the question, downsize and prepare the image the way you want it to look. That way you are in total control of the final look. If you don't you are leaving things to the printer driver that will do a generic conversion, which leaves the printer, not the photographer in control of the final image.

[xpatUSA]

I've photomerged an image considerably larger than necessary (18000 x 12000 dpi)
Don't want to crop anything...want to end up with a 20 x 30 inch print.
Would I increase sharpness by merely downsizing to 9000 x 6000 pixels?

Yes, at least according to Bart van der Wolf: http://kronometric.org/phot/iq/Down%...%20methods.htm

Am I being clear with my question?

Yes, William, you are.

[Manfred M]

Yes, at least according to Bart van der Wolf: http://kronometric.org/phot/iq/Down%...%20methods.htm

Photoshop CS was released in 2003, so the software used in the analysis is almost 14 years old. The algorithms doing the upsampling and downsampling had a major update in 2012 (with the release of CS6) . I'm fairly certain that the conclusions reached in the article should be treated as being dated, given the tools used to do it.

[xpatUSA]

I must have a fundamental misunderstanding event...it would seem that if you compressed a
300 dpi 60 x 40 print down to 300 dpi 30 x 20 print, it would inherently be considerably sharper.

It would indeed, assuming we all agree on what constitutes "sharpness" - a tall order, I suspect.

I can take an image with a slant edge in it and measure it's MTF at Nyquist*. Then downsize the image and measure again. The second image will have a higher MTF at Nyquist *.

*Nyquist is 0.5 cycles per pixel.

Since, at higher spatial frequencies, MTF is a measure of "sharpness" (just like lens data), that supports your statement, William.

A comparison for a 47% downsize:

[xpatUSA]

Originally Posted by xpatUSA Yes, at least according to Bart van der Wolf: http://kronometric.org/phot/iq/Down%...%20methods.htm

Photoshop CS was released in 2003, so the software used in the analysis is almost 14 years old. The algorithms doing the upsampling and downsampling had a major update in 2012 (with the release of CS6) . I'm fairly certain that the conclusions reached in the article should be treated as being dated, given the tools used to do it.

Bart's conclusions were of a general nature and are not dated as such; the algorithms and images were illustrations and may well be dated - but that does not alter the main principle of the article:

If an image contains frequencies below the Nyquist sampling rate for that image and if that image is down-sampled to where those same frequencies would then be above the Nyquist sampling rate for the resized image, then there will be aliasing. Sharpening that particular image will exacerbate the appearance of the aliasing.

Anyone that downsizes 'bicubic sharper' or 'Lanczos2' and then routinely applies even more sharpening must a) like "jaggies"/Moiré or b) have shot a soft scene or c) have shot out-of-focus or d) have shot at f/22 or more.

[Manfred M]

Bart's conclusions were of a general nature and are not dated as such; the algorithms and images were illustrations and may well be dated - but that does not alter the main principle of the article:

Agreed to a point. I think your earlier comment is actually what is in play here:

It would indeed, assuming we all agree on what constitutes "sharpness"

I would agree that downsizing does what you / the article suggests, but in the applications most of us look at in photographs, "sharpening" is more related to how the transitions between lighter and darker areas of the image are handled. This is what tools like "unsharp mask", "high pass filter" and a host of other techniques do. This has a greater impact on how people view an image that is sharp or not.

[george013]

I just wanted to say I found situations whereby resizing to print size gave a different result as when not doing it. I did a few tests but that was a few years ago. Maybe I did something wrong, I don't know exactly. But I did see it in the small lines, I my case text lines.

On the other hand, resizing the print and saving again as a jpg means an extra compression on the image, one that wouldn't be done when you let the printer driver do the job. My thoughts. Unless you send the file as a TIF.

A print of 20" by 30" is bigger as an A2. I doubt William has a printer like that. So most likely he will have a printing service have it done. That's why he first has to ask what resolution they use for that size. After all what is the use of this question when you make the wrong calculations?

In general I wonder if the feeling of sharpness when downsizing the image is the same as getting sharper. As usual you'll have to define the circumstances under which you compare and on the monitor that will be full size or screen size.
I tried it with IView, you can open that more times. Load the same picture, view one on full screen and resize the other to match the same magnification. In the picture I just did I didn't see much difference. But adding a sharpening at both, the downsized image became bad in the small lines. I know it's not the best editor.


George