I have some high-quality RAW files of artwork that I've processed with some cropping, contrast adjustment, etc. I'd like them to be printed on metal at about 2'x3', which is about twice the size I get when the ppi is set to 240.

I've been looking at various options. I saw a roundup - it was a few years old, but it was the best I could find - which suggested that something called SmartEdge, then still in development, was the hands-down the best algorithm out there.

The test results appeared superior to Genuine Fractals, which is (or was) a paid service.

Does anyone know about these? Whether they can be brought into Adobe programs or only used via stand-alone software? But most importantly, what is the most effective algorithm?

  • \$\begingroup\$ I think he's saying at the size he wants to print his images, they only have enough pixels to be at 120 ppi. \$\endgroup\$
    – Michael C
    Mar 31, 2016 at 6:03
  • \$\begingroup\$ @MichaelClark I am wondering why the OP need 240 PPI : is it because he think he needs it or is it because of the specific print he wants to make. \$\endgroup\$
    – Olivier
    Mar 31, 2016 at 19:56
  • \$\begingroup\$ Yes, the 240 was probably the preset, as Rafael supposed. We're going for a format of about 2 feet by 3 feet; it is artwork and (for full disclosure) will be printed on metal. \$\endgroup\$ Apr 1, 2016 at 5:37

2 Answers 2


I have some high-quality RAW files.

This is important. Let us move on.

I need to have them printable at about twice the size I get with the ppi set to 240.

I guess the 240 ppi is a preset of the camera... Well, you can modify that value to whatever value you need.

Normal people, won't see a difference on a 100ppi photo and a 200ppi at 30cm except in some real little-micro-tiny details or at really close range (less than a feet).

The question is how far normal viewers will be? If you double the distance you can cut in half the resolution.

(This images have an error, it says dpi instead of ppi)


Now. If we start with a 300ppi file viewed at 30 cm we can start making some numbers.

In my experience, the expected viewing distance is somehow equal to the largest size of your photo.


But let us assume that you really need to resample an image. That is a matter of taste.

This test has some time but you can compare different software. The first image is a resize without any resampling. The second one with a Lanczos algorithm, which is similar to the bicubic one. And the other two are from popular resamping software. Benvista and Reshade.


In my opinion, a simple "bicubic sharpen" on Photoshop is good enough. You can use an additional sharpening if you need.

My recommendation is that you just should use round numbers. Resample at 200%, not at 187% for example.

I would never resample anything more than 200%. For a blurred background I could go for 300%.

  • \$\begingroup\$ Thanks, that's helpful. It's artwork. Will be (as noted above) printed on metal. It looks If, as I imagine, the distance is the key parameter, I don't think there's reason to suppose people won't be a meter away, and for artwork I'll then want 300dpi, it seems. \$\endgroup\$ Apr 1, 2016 at 5:45
  • \$\begingroup\$ Looking at those images at a large size and at that distance, I do see jagged pixellation of the teeth on the non-upsampled image. In that image and the Lanczos algorithm image, there are obvious striations in the hair. The other two do look better... no one's familiar with the SmartEdge algorithm? \$\endgroup\$ Apr 1, 2016 at 5:48
  • \$\begingroup\$ At 1 m, you can go away with 150ppi. Almost any modern magazine has the equivalent of 150 ppi when printed at 150 lpi. \$\endgroup\$
    – Rafael
    Apr 1, 2016 at 12:18
  • \$\begingroup\$ I am preparing some "low resolution simulators". If your computer has 14" in the base, you can download this: otake.com.mx/Apuntes/HablemosDeResolucion/… View it at full screen and you will have a glimps of the resolution marked on each page. I still need to prepare the english versions and a lot of diferent sizes. \$\endgroup\$
    – Rafael
    Apr 1, 2016 at 12:23

The best algorithm should attempt to fill in the missing details as best as possible, which means that they must be based on so-called "super resolution" methods. This looks impossible to do using a single image, however, as explained in this paper, one can make use of the repetitions in a single picture, similar objects appear multiple times on different scales, to extract sub pixel information.


Not the answer you're looking for? Browse other questions tagged or ask your own question.