Upsizing at a different dpi

[trotter_hardy]

Suppose the human eye can't tell the difference between a print at 300 dpi and

one at 250 dpi. (You don't have to agree--just suppose.) And suppose that the

eye can tell the difference between 300 dpi and 150 dpi. You have a picture that

prints satisfactorily 8 x 10 at 300 dpi. You want to upsize it to, say, 16 x 20.

Assume the print will be viewed from the same distance in all cases.

 

Is it better to upsize it at 300 dpi, or 250 dpi? Same question for 300 dpi and

150 dpi.

 

The higher dpi contains more "detail," but it's made up by the software. The

lower dpi contains less "detail" but less has to be made up by the software.

 

Thanks for any insights.

[v.anisimov]

In your hypothetical case 250 is better, if "...the human eye can't tell the difference between a print at 300 dpi and one at 250 dpi..."

 

I don't think "the higher dpi contains more "detail"... The software merely fills gaps between neghbouring pixels, it doesn't create new detail.

[bruce_watson1]

Neither. It's better to upsize to the exact needs of the printer in question. This will keep the printer driver from doing yet another interpolation which will further degrade your image.

 

IOW, find out the printer's "native resolution" and upsize to that. In the case of a lightjet, I think it's 304.8 ppi or some other obscure number (depends on the lightjet model). In the case of an Epson inkjet printer, 360 ppi. Etc...

 

At least be an even factor of 2 away to make the math extremely easy for the driver (that is, minimize artifacts). IOW, if you aren't going to give an Epson printer 360 ppi, give it 180 ppi or 720 ppi.

[Ed_Ingold]

You can see the difference between 250 and 300 ppi - in the subject's eyes, particularly highlights. Highlights look "angular".

 

For best results, I resample to the exact print size and, for an Epson inkjet, to 360 ppi, and save the results as a JPEG. Epsons print to a modulo 360 resolution, consequently any resampling is done using an integer factor. Epson resampling is quite good, and OK for proofs or non-critical work. I am more concerned about sizing than resampling.

[trotter_hardy]

So ... sounds like going for the printer's native resolution is most important. I'll try working with that. Thanks for the answers.