Digital BW, The Print

Steven Karafyllakis wrote:


Steve, thanks for your answer on Qimage. I'll add this to my list of 
interpolation options.


> Another way to get into S-spline that may be cheaper: Panotools and 
> PTGui. It has spline-16 and spline-32...

Spline-16 and spline-32 are cubic splines, and not the same as s spline. 
Panotools does not support s spline in any version that I'm familiar 
with. A pity, too; it would be really nice if it did.

The PT releases that I'm familiar with have some mix of poly 3, spline 
16, spline 36, sinc 256, spline 64, bilinear, nearest neighbor, and 
sinc1024. Poly 3 and Spline 16, 36, and 64 are all cubic splines. 
Sinc256 and sinc1024 are Lanczos splines and I'd guess they are the best 
place to start for resizing in PT.


Paul Roark wrote:

> My conclusion, however, is that these programs are just marginal
> improvements to images that really needed to start as higher resolution
> files in the first place.

I think the only real use of interpolation software for increasing photo 
size is to prevent the print driver or RIP from doing it automagically, 
in an uncontrolled and possibly undesirable way, on its own. Exhibit A 
would be my C86 which still prints bands at certain input resolutions 
(including some fairly high ones); increasing the resolution using any 
method and then sending the file to the printer solves this problem. 
Exhibit B would be any very large print from an undersampled original, 
in which increasing resolution allows you to dodge some 
pixelation-related artifacts. As far as adding detail, you are spot on: 
upsampling doesn't add information to an image; it can only destroy it.


Steve Kale wrote:

> Is this the Shortcut Photozoom product?

Yes, I think they have what is considered to be the best (only?) 
commercial implementation of an s spline for photograph resampling, 
outside of a few scientific packages.


Peter Marshall wrote:

> As I understand it, the maths behind GF is published but patented 
> (rather than a trade secret).

That could be. I based my belief about it being a trade secret on my 
inability to find a relevant patent, combined with several reviewers' 
statements to the effect that GF contains "proprietary algorithms." 
Either way, at the present time I've never seen the math and have no 
idea how GF does what it does. If someone has the software, is a patent 
number listed on the box or in the help file?

-- 
Jeff Medkeff
Eagle River, Alaska