I never resample my original scan. Any sharpening I do to the original scan is done solely to correct the effects of the scanning (or capture) process.
The only time I resample is when I'm going to display it at a much different size (always smaller) than the original scan. I've never had any luck up-rezzing photos and having them look worth a darn, and I would rather just print within the capabilities of the scanner or digital camera.
But consider what Frank said. He used an extreme example--a web display--to illustrate the point. A monitor has a screen resolution in the vicinity of 100 pixels/inch. To give that image just a hint of more detail than meets the eye, I have to sharpened it. I set my unsharp masking parameters to achieve the effect. But those parameters are measured in pixels, in Photoshop. For web display, I may be reducing the linear resolution from 8000 to 800 pixels. The sharpening I did at the original resolution may have a radius of some fraction of a pixel, and that sharpening effect will vanish utterly when I display it on a monitor--it will be a fraction of a fraction of the smallest thing visible on the screen.
So, I need to sharpen it again, this time at the display resolution, so that my settings have relevance on the display device. This time, I want a sharpening effect not to correct fuzziness caused by the scanner optics or the AA filter in a digital camera, but rather to achieve a particular look in the final print. Thus, the settings for that sharpening process have to relate to pixels that are actually relevant at display resolution. Since the main parameter in sharpening is the radius, and since the radius is measured in pixels, I have to resample the image first, so that the sharpening I'm doing for targeting purposes has visual relevance.
A screen display is an extreme example. But extreme examples often lay at the other end of a continuum. If I have a file that is 8000x10,000 pixels, and I want to print it on my Epson at 8x10 inches, the file will have a resolution of 1000 pixels/inch on the print. Any sharpening I do with a radius of a fraction of a pixel is so much wasted effort--it will never be seen. But I want that 8x10 print to have a little more zip than a 16x20 print, because of how people will look at it. So, I first downsample a copy of the file that I'm targeting for an 8x10 print to something like 3600 pixels wide, for a pixel density at print size of 360 pixels/inch. I'm not worried about aliasing the printer resolution--that is too small to see in any cased, though some claim to see it. I'm worried about the sharpening that I do having the visual effect that I want, for an 8x10 print. After resampling the image to be 3600 pixels wide, the sharpening settings (particularly radius) that are in pixel dimensions now mean something relevant at that print size.
As I said before, most of my large-format images have such an abundance of local contrast that the final print doesn't need any of that at any reasonable print size. So, I may not apply sharpening at all during targeting. It happens a lot more when making prints from my digital camera. But what about feathering a selection? That's also measured in pixels. Let's say my experience tells me that at 360 ppi in the print, I need a feather of, say, 10 pixels to provide a visibly smooth gradation rather than a harsh edge or bright line. If I was working with a file ten times my display resolution, I'd have to set that feature to 100, and then it would still be wrong for a print of different size. I might want the narrowest feature I can use that isn't swallowed up when being downsampled either by the printer or for web display. Jim says he doesn't worry about it, and that's why he's had to answer a question or two about the bright edges on his sky selections. I know those are smooth and gorgeous at print size, but he would need a thicker feather to achieve the same effect at web display size.
By resampling the target file before making those changes, I can tailor their settings for print resolution relevance rather than having to do pixel arithmetic in my head.
I used to work with original scans, setting them up for the biggest print I intended to make. But then I bought an Epson 3800, and suddenly my biggest print got bigger. The result is that I have had to rescan some images for larger prints. Had I stored my original scans with corrective adjustments only, all I would have needed to do was make another copy for the new target size, and then make targeting adjustments for my new printer at that size. Would have saved a lot of time with those images.
I've made plenty of prints at much higher than 360 ppi resolution at the printer, and let the printer sort it out without issue. But those were prints that didn't need those pixel-based effects to look right. And I'm also giving the impression that I consider this more carefully and in more detail than I really do. The larger the original scan (and the larger the film being scanned), the less any of this seems to matter in the final print. I've made lots of pictures that I like with a 6-MP APS-C DSLR, and those need very careful, display-resolution sharpening.
Rick "sometimes working right at the original file's limits" Denney
Bookmarks