To what extent is resizing an image (using Photoshop's bicubic sampling algorithm) acceptable for print?

Question

I'm wanting to print a 300 DPI brochure, but my first attempt at making it fell short of the proper resolution, so I am having to remake it. I've already printed my business cards, which were 1125x675 at 300 DPI, and they have the same picture on it that I want in the brochure.

I was thinking about the different ways to get the brochure the correct resolution (remaking it completely or just resizing it,) but I noticed that the picture wasn't quite the same. It's comprised of 12-15 layers that are difficult to replicate because I made a lot of level, saturation, etc. adjustments.

I decided then to simply remake it, importing the business card's picture as a layer to use as a reference for what my end result should look like. I pasted in the flattened picture from my business card and then resized the layer... from 1125x675 to 2245x1347 (nearly twice the size)... and much to my surprise, it looks just fine, so I'm contemplating just going with that resized picture.

Is this generally considered acceptable practice, or is it at least acceptable in this case since it still looks good?

Here's what the resized image looks like (upper-left corner) : http://www.solarcoordinates.com/images/brochure1_interior_5_spacing.png

Answer 1

As Horatio says, if it looks good, it's probably fine.

There are two schools of thought on upsampling: One says, "Never, ever upsample"; the other says, "Hey, what the heck, upsampling rocks." In almost all cases I side with the former. Upsampling adds nothing but "best guesses" to the image. It specifically doesn't add any image information (I don't care what the algorithm is, software can't create data where none exists), so at best you get a bigger-but-fuzzier image.

For a thorough, approachable, informative (and entertaining) look at resizing in Photoshop, see this video by Deke McLelland.

If you do feel you have to resize, you'll actually get marginally better results resizing to exactly twice the original image size than almost twice the size. Exact multiples of two give the bicubic algorithm an easier time interpolating pixels, so the final result tends to be cleaner in the detailed areas of the image.

After you upsample, though, you MUST sharpen the upsampled image. Pixel interpolation spreads out the original image information, which means that every edge in the image loses a bit of contrast. Since edge contrast is what makes an image appear "sharp" (sharpening algorithms like Unsharp Mask and Smart Sharpen increase edge contrast).

For offset printing, you are likely to be quite safe at 200 to 270 dpi. Ask your printer what "line screen" they use (typically 133 lines-per-inch, but do ask) and multiply by 1.5 to get the working minimum resolution you should aim for.

Answer 2

The answer is highly dependent on what the picture is of and how it's used; 300ppi is a rule of thumb.

It's nothing like enough for one-colour line diagram on good paper, it's more than enough for a defocused photo used as a watermark on textured paper.

Answer 3

What is and isn't acceptable quality is really up to person paying the bills. I know people who are perfectly happy taking a 800x600 compressed jpeg and printing it at A4.

Generally speaking if you scale up a smaller image you will get a 'fuzzier', more blurred image than one that was created at the proper resolution. Whether or not the 'fuzziness' is within your tolerance is an entirely subjective call.

For what its worth; dependent on time/cost/effort i will sometimes do exactly what you have done for my own personal work.

Answer 4

I know you are looking for a specific number, but the plain answer is: "looks good" and "acceptable" are generally considered synonyms.