Interpolation method

[Andrew Teakle]

Now this is probably a tired old topic, but things move on in this fast-paced digital world and it has been a while since I have read anything on the latest interpolation methods.

For several years I have been using Jack Flesher's Uprezzing Digital Images technique that was featured in Digital Outback Photo. For those not familiar with it, basically you upscale your image to 20% larger than your desired output with Photoshop's Bicubic Smoother, Output Sharpen, then downscale it to the final print size using Bicubic Sharper. It has generally given me impressive results and I have not really felt the need to do things differently. Recently, however, I have been commissioned to produce 36 prints at up to 60" from files from 35mm slides, 8MP digital, and 6x7 MF slides. Now I'm not expecting any miracles from the 35mm and 8MP files, and to be fair they won't be viewed from close range, but I want the best possible prints. Thinking about this, I came upon an ad in Outdoor Photographer for a Photoshop plug-in called Blow Up from Alien Skin. It promises 1600% enlargements, "better than Bicubic or any other third party solution". I generally take marketing blurb with a grain of salt but it got me thinking...What is the best contemporary interpolation method? The ones I'm familiar with (by name at least, I have only tried Genuine Fractals) are S-Spline, Extensis Pxl-ate(??), as well as GF.

I'll try the Blow Up demo and report back, but any thoughts?

[David White]

I tried the demo but found that on details, such as a ship's rigging, bicubic smoother in Photoshop, direct to the final size, was better prior to output sharpening.

[Hermie]

Ron Bigelow did some testing:
http://www.ronbigelow.com/articles/interpo...terpolation.htm

[Andrew Teakle]

Thanks David and Hermie for your prompt replies. I have downloaded the demo and will test it ASAP.

[TimothyFarrar]

A new (developed in the past month) method to upsample, here is a 200% enlargement crop,



To take advantage of my strange noiseless workflow, I recently created a new method of upsampling (which runs as an action in Photoshop CS2). While I think the results are impressive in comparison to other popular methods, I would love to get some constructive feedback from others to see where it actually stands.

http://www.farrarfocus.com/ffdd/double.htm - Page with before and after image rollover examples also images enlarged and then sharpened again as well as a 800% enlargement (I personally never use over 200%).

http://www.farrarfocus.com/ffdd/example3e.jpg - A much larger example saved as a JPEG (it's not too big).

http://www.farrarfocus.com/ffdd/examples.htm - Download example3.psd to get the source of the above larger example before upsampling. If you like you can upsample this image using your favorite method and compare to the JPEG above.

Let me know what you think.

[bjanes]

Ron Bigelow did some testing:
http://www.ronbigelow.com/articles/interpo...terpolation.htm
[a href=\"index.php?act=findpost&pid=76567\"][{POST_SNAPBACK}][/a]

These tests are interesting, but an option that Ron did not test is to send the file directly to the printer without any interpolation. In his recent book on sharpening, Bruce Fraser states that he usually does not bother to resample when sending output to an inkjet printer when upresing would be needed to match the native resolution of the printer.

[plugsnpixels]

If you do go with Blow Up, my site offers a $20 discount.

Also check the list of Resampling/interpolation options here. I have samples of most of the results posted (I just took delivery of the new FFDD, and still need to experiment with its interpolation capabilities).

This would be a great head-to-head topic in an upcoming ezine!

[Mark Graf]

I keep hoping something will eventually prove itself out, but every time I try to do a trial on my own images, I end up convinced I don't need anything more than Photoshop CS2.    I just recently did a trial of Genuine Fractals vs. PhotoZoom vs. Photoshop Bicubic Smoother (in one step).

I enlarged a 4000 dpi scan of a 35 mm slide to a 20x30 print at 360 dpi using each.  Both GF and PhotoZoom resulted in what appeared to have more detail at the final size.   However, once I applied SmartSharpen as I normally would before printing - the GF, PhotoZoom, and the Bicubic Smoother samples all ended up looking the same.

So from direct processing, the third parties seem to offer some advantage, but I only care what ultimately looks better in print.   This is what is really hard to get to in all the comparisons on the net - is there a genuine difference in the final print?  And, is it possible to get to the same result perhaps with only a couple extra steps within Photoshop.

[TimothyFarrar]

Mark,

Working from scanned film definatly takes a different type of enlargement/sharpen/denoise workflow that digital. Most film scans are done at such a high ppi (pixels per inch) that individual pixels generate the film grain instead of actual detail.

You might actually have better results reducing the film scan (bilinear resize or gaussian blur + nearist neighbor resize) to a resolution where pixels represent detail instead of film grain, then try a traditional digital processing workflow (reduce noise, sharpen, enlarge, resharpen).

[TimothyFarrar]

BTW, anyone who wants to try the FFDD tools for digital file enlargement, just send me an email that notes your working colorspace (so I can sent you the correct action), and I will email you a demo action.

email to: timothy
at domain: farrarfocus.com

[Tim Gray]

QImage has traditionally had excellent up-rezzing capability.  You can print back to a file if you just want to up-rez and print using something else.