Upsizing | Dan Margulis Photoshop

Ledet training home -> Adobe training -> Photoshop training -> Professional Photoshop training with Dan Margulis

Dan Margulis Applied Color Theory

Increasing Resolution

Up resing?

Posted by: "John Ruttenberg"

Tue Mar 6, 2007 6:46 am (PST)

I know there are a lot of third party packages for this (Genuine Fractals, Fred Miranda...)

There is also this advice someone once gave me. Use PS Image->Image Size repeatedly, increasing by exactly 10% each time. I have no idea if/why this works.

Is there a best way to do this? What are the limitations?

--John Ruttenberg

___________________________________________________________________________

Re: Up resing?

Posted by: "MARK SEGAL"

Tue Mar 6, 2007 8:00 am (PST)

Normally I don't comment on stuff I haven't tested myself, but the more about this I read of others who have tested it and from my own work using Phtoshop only, Photoshop's recent algorithms of Bi-Cubic Sharper and Bi-Cubic Smoother would appear to eliminate the advantage of third-party software as well as any need to adjust with small increments. I have used these algorithms making very large adjustments and frankly the result looks fine to me - my only disadvantage is that I have not done an eyeball to eyeball comparison with other methods, because I haven't felt it important to spend the time doing so.

Mark Segal

___________________________________________________________________________

Re: Up resing?

Posted by: "jaelb17"

Tue Mar 6, 2007 10:26 am (PST)

From what I understand, by upsizing with bicubic resampling in 10% increments, you get a smaller, therefore more accurate sampling of pixels than if you used a large adjustment. IMHO, quality still suffers, so I don't like to up-size and rarely do it, if ever. However, I have used this method, sharpening every so often for a somewhat successful resize. Limitations, I think, depend on the quality/resolution of the original graphic, your output, and the desired quality.

___________________________________________________________________________

Re: Up resing?

Posted by: "Louis Dina"

Hi John. I used Genuine Fractals quite a bit in the past (I do a lot of trade show posters) and have stopped using it. While it does help improve smoothness and reduce some of the noise, it tends to introduce other artifacts and a rather plastic look that I don't like. I'd rather live with the noise. I've tried the 10% at a time trick, and don't find it to be a substantial improvement over standard bicubic resampling. I haven't used the other programs.

Recently I had an image from my Canon 20D blown up to 48 X 72 inches, rather a large blow up. For comparison, I sent the original image without modifying the original resolution (about 49 ppi) and a file that I resampled to 125 ppi using Bucubic Smoother, after which I did some sharpening. Frankly, I couldn't tell the two prints apart, and neither could the printer who did the job. The quality was reasonably good given the size. So, for images that will be viewed from a longer distance away, I plan on using the native, unresampled image in the future. Tests I did with this printer (on wide format VuTek inkjet printers) showed that quality was quite good at 72ppi and above. Things started breaking down below about 72 ppi. But my 49 ppi image looked darn good from 3 or 4 feet away (a more normal viewing distance for this size).

Regards, Lou Dina

___________________________________________________________________________

Re: Up resing?

Posted by: "mac townsend"

Tue Mar 6, 2007 10:28 am (PST)

I think part of the answer lies in what the end result is to be.

If the end result is to be halftoned and output for press, your results will almost certainly be disappointing compared to what you would get if the end result is an inkjet print. If the end result is for screen viewing, you the results will likely be pretty good.

Mac Townsend

Adcom Graphics Digital Imaging

Fairfield, California

www.adcomgraphics.com

___________________________________________________________________________

Re: Up resing?

Posted by: "Hoffner, Randall N"

Tue Mar 6, 2007 10:29 am (PST)

Upsizing in 10 percent increments really does work pretty well; not sure why unless it is just that fewer new pixels are generated at a time this way. There is a way that works better, I believe: first upsize by 20 percent more than you need for your final image, using bicubic smoother; then downsize to your final image size using bicubic sharper. There is a sharpening step I there too, I believe after the first upsizing step. You can download a free automation script to do this procedure on www.outbackphoto.com <http://www.outbackphoto.com/> , under the title "DOP Upsizing".

Beyond that, Delores demonstrated to us that if you are upsizing things with straight, geometrical, edges, the best results are obtained using "Blowup" - but it costs $200.

Randy Hoffner

___________________________________________________________________________

Re: Up resing?

Posted by: "Michael Demyan"

Tue Mar 6, 2007 11:41 am (PST)

Up-resing an image using the 10% method works best because of the calculations that PS uses.

There is no rounding of results - so a more accurate calculation/image results.

Mike

Fine Photography by

Michael Demyan

www.mikedemyan.com

610-758-9769

___________________________________________________________________________

Re: Up resing?

Posted by: "MARK SEGAL"

Tue Mar 6, 2007 2:02 pm (PST)

Mike,

Is this a theoretical proposition based on your knowledge of the mathematical algorithms in Photoshop CS2, or is it based on actual experimentation of trying the 10% approach versus the one-large-gulp approach with Bicubic Smoother or Sharper and comparing the printed results? I'm curious to understand the basis of your observation.

Mark Segal

___________________________________________________________________________

Re: Up resing?

Posted by: "J Walton"

Tue Mar 6, 2007 9:33 pm (PST)

On 06 Mar 2007 14:02:40 -0800, MARK SEGAL wrote:

Well put!

I think I've seen more unsubstantiated speculation on resizing techniques than on any other subject. Those who swear up and down Genuine Fractals is the best have *almost never* done side-by-side tests. Having done them I'm not impressed by GF or 10% increments.

In some cases Nearest Neighbor is the best interpolation, but if you don't experiment you'll never know what the differences are.

J Walton

___________________________________________________________________________

Re: Up resing?

Posted by: "Michael Demyan"

Tue Mar 6, 2007 9:37 pm (PST)

Hi Mark:

It is based upon my experience and trials, and a guesstimate based upon mathematical rounding principals. It is also based upon the assumption that if you expand an image in steps of 10% you 'would' not introduce as much of a computer guess as to what color pixels to insert as compared to the Big Gulp approach.

I do not know the alogorithm that PS uses.

The method works quite well for enlarging an image 3 to 4 times the original size.

At the present I do not 'up-res' images for print. I am using Qimage which will up-size images using a number of different approaches. I use the default method which meets my needs.

Mike

Fine Photography by

Michael Demyan

www.mikedemyan.com

610-758-9769

___________________________________________________________________________

Re: Up resing?

Posted by: "Edward Bateman"

Tue Mar 6, 2007 11:33 pm (PST)

I created a test image (containing photos, type, and a grid) and tried resizing up with the various Photoshop options as well as Genuine Fractals, Extensis PixelSmart, and sequentially increasing by 10% increments. I thought the 10% technique was noticeably the worst. Some fine details just vanished. Different images, however, might give different results. Plus there was some disagreement as to which was the best result.

-Ed Bateman

__________________________

Edward Bateman

Fine Art o Capt. Eddie's Space Camp

www.ebateman.net o www.captaineddie.net

___________________________________________________________________________

Re: Up resing?

Posted by: "Richard Wagner"

Wed Mar 7, 2007 4:28 am (PST)

On Mar 6, 2007, at 9:08 PM, Michael Demyan wrote:

It is based upon my experience and trials, and a guesstimate based

upon mathematical rounding principals. It is also based upon the

assumption that if you expand an image in steps of 10% you 'would'

not introduce as much of a computer guess as to what color pixels

to insert as compared to the Big Gulp approach.

Up-rezzing is interpolation. All of the "new" data points are "computer guesses" based on a mathematical formula. It is certainly conceivable that the "big gulp" approach will give a *better* interpolation, since it "knows" which are "real" data points, whereas with the 10% step approach, previously interpolated values are taken as "real" points. I played with this considerably when CS2 introduced the bicubic sharper and bicubic smoother algorithms, and I found no advantage to the step approach compared with bicubic smoother for up-rezzing. The step approach introduced more artifact.

At the present I do not 'up-res' images for print. I am using

Qimage which will up-size images using a number of different

approaches. I use the default method which meets my needs.

Well, in actuality you are up-rezzing, it's just being done by a program other than PS, and you're not sure exactly what it's doing. I do the same with the ImagePrint RIP, knowing that its base size is 360ppi, and that it will up-res to that size. (For, say, 28 x 42 inch prints, I send ImagePrint a file that is 180 ppi and let it use its own up-rezzing algorithms.) Works well for me.

--Rich Wagner

___________________________________________________________________________

Re: Up resing?

Posted by: "john castronovo"

Wed Mar 7, 2007 7:42 am (PST)

From: "J Walton"

I think I've seen more unsubstantiated speculation on resizing

techniques than on any other subject. Those who swear up and down

Genuine Fractals is the best have *almost never* done side-by-side

tests. Having done them I'm not impressed by GF or 10% increments.

In some cases Nearest Neighbor is the best interpolation, but if you

don't experiment you'll never know what the differences are.

I've found all of what you say to be true. This whole thing is highly dependent on image content, and where one method looks better with one method, it's worse with another. Also, people are usually ignorant of the eye's ability to see pixels at a given viewing distance once printed. Most images aren't viewed from ten inches away, so 300 ppi or more isn't required and the need for such resolution dwindles rather quickly as the viewing distance increases.

john castronovo

___________________________________________________________________________

Re: Up resing?

Posted by: "Mark Segal"

Wed Mar 7, 2007 7:42 am (PST)

Mike,

Thanks for clarifying the basis of your advice. Needless to say we should set aside guestimates and assumptions, but rather focus on experience and trials. However, in this area, from your last paragraph it would appear you are mainly relying on QImage, which of course is fine when satisfactory, but doesn't shed any light about the issue at hand, unless we knew what's going on inside the QImage black-box, or by comparison with Photoshop's algortithms. So that puts us back to square-one one on the question of what up-rezzing approach works best on the basis of side-by-side comparison of prints that are identical except for the up-rezzing technique. I'm wondering whether any one reading this List has done controlled tests using CS2 as the Photoshop comparator?

Mark Segal

___________________________________________________________________________

Re: Up resing?

Posted by: "Owen Colborne"

Wed Mar 7, 2007 6:34 pm (PST)

I was told some time ago by a colleague that if you doubled the area of a file, NOT the file size, that in effect you would double the number of pixels in a file. That way every pixel created would have a real pixel to reference. I worked it out to upsizing to 141.43% (as close as I can figure). It works for me.

You might wish to experiment with that and see if you think the result is better than going up in 10% increment.

Cheers,

Owen Colborne

___________________________________________________________________________

Up Resizing (was: Up resing?)

Posted by: "Pylant, Brian"

Wed Mar 7, 2007 8:42 am (PST)

My observations have been that the 10% method (manually, or with that plugin) results in a generally smoother appearance (less jagged edges, etc.) but sacrifices detail to accomplish this. Conversely, simply upsampling directly to the target resolution often appears much more jagged, but the perception of detail elements is greater.

The source material greatly affects this as well: upsampling a small 72 PPI image that didn't have a lot of detail to begin with is wholly different than upsampling a press-ready 300 PPI image to a greater print size (say, from 3 x 5 to 11 x 17).

End result: as with most things, there is no one single method that works best all of the time, and the operator needs to determine which will look best for the specific image he/she is working on at the moment, subjective as that determination may be.

BRIAN PYLANT

Manager, Electronic Prepress

:::::::::::::::::::::::::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::

Disc Makers

7905 North Route 130, Pennsauken, NJ 08110