From: "Dave Badger", INTERNET:dbadge@worldnet.att.net Date: Thu, Jan 18, 2001, 9:12 PM RE: [colortheory] Enlarging Digital Files

I get frequent calls from customers wanting to "res up" small files rather then get a rescan (for printing). One told me today of a program (Genuine Fractals?) that claims to enlarge up to 800% and still retain acceptable detail.

Does anyone on the list have experience with such a program and how well it works? Are there any tricks for interpolating images up in Photoshop and retaining or restoring detail? Thanks

Dave Badger

-------------------------------------------------------------------------------

From: "Rob Outlaw", INTERNET:routlaw@IMT.NET Date: Fri, Jan 19, 2001, 12:02 AM RE: Re: [colortheory] Enlarging Digital Files

Dave Badger writes:

> I get frequent calls from customers wanting to "res up" small files rather > then get a rescan (for printing). One told me today of a program (Genuine > Fractals?) that claims to enlarge up to 800% and still retain acceptable > detail.

It is my understanding the worlds largest digital print is on display in Times Square of a shot taken with the Nikon Coolpix 990 and resized to somewhere in the neighborhood of 40x50 feet. Thats probably not the exact dimension but close enough.

> Does anyone on the list have experience with such a program and how well it > works? Are there any tricks for interpolating images up in Photoshop and > retaining or restoring detail?

I do not think it possible to restore detail that was never there to begin with. You can not make something out of nothing in othe words. However there are some tricks for interpolating in PS.

I have used the program quite a bit as well as a number of other photographers that I know of. For the most part I prefer to use PS bicubic interpolation. However you will find others who rant and rave about GF until the cows come home thinking it is the best thing since sliced bread. For me I have spent probably way too many hours in front of the computer comparing the difference with these two methods of resizing with the following findings, which are based upon digital capture rather than scanned film.

First and this is the most important part. When resizing with PS I have found that I get much better results when running my USM before interpolating. Granted this goes against conventional wisdom but don't take my word for it, try it yourself. I have compared the difference way too many times to ever go back to upsizing first.

At any rate what I do not like about GF at least for display size prints in the neighborhood of 20x24 or so is that the program due to its apparent de-pixeling of the image gives an image a very watery or liquid type of texture. On some images this is not too bad but it can make others look odd to say the least, similar to one of the PS artistic filters. I have seen GF literally take the minute texture right out of some clothes shot at a distance while the PS version did not do this. Furthermore GF enhances some edge detail so much so that it starts to look like elements within an image are cut out and pasted on the page. Again not so with the PS verson of interpolation.

On numerous occasions I have upsized images using GF and with a duplicate of the same image upsized it using my non conventional PS method to 20x24. On screen the GF versions does just fine but never has it shown itself to be vastly superior to the PS method. At times with the GF version there is a slightly hightened sense of edge detail compared to the PS version viewing at 100% pixels. More often as not though, myself and pretty much anyone who has looked at prints from my Epson 1270 in the 11x14 to 13x19 range prefers the PS method of upsizing because it just looks a bit more natural.

Finally while on a call to tech support with Altimira Group, the devolopers of GF, I was told the following. The program was really intended for extreme enlargements, billboard size stuff in other words, and that it worked best with files that were at least 18MB to begin with. At the time I think that they were dubious of its benefits with some of the digital cameras that create small files. If this is what they were really thinking makes you wonder why a trial version was included with some of the Epson printers. In conclusion I would say that the type of image will dictate whether or not GF is the best method of interpolation, but for me I almost never use it anymore.

Hope that helps

Rob Outlaw Rob Outlaw Photography www.roboutlawphotography.com

-------------------------------------------------------------------------------

From: Andrew Rodney, INTERNET:andrew@digitaldog.net Date: Fri, Jan 19, 2001, 8:46 AM RE: Re: [colortheory] Enlarging Digital Files

on 1/18/01 10:00 PM, Rob Outlaw at routlaw@IMT.NET wrote:

> It is my understanding the worlds largest digital print is on display > in Times Square > of a shot taken with the Nikon Coolpix 990 and resized to somewhere in > the neighborhood > of 40x50 feet. Thats probably not the exact dimension but close > enough. >

True but that doesn1t say anything about the quality of the upsizing nor how much better or worse it would look just using Photoshop. Photoshop could easily make a file that output larger. Since GF Pro is a plug in for Photoshop, it has the same pixel limitations as Photoshop.

Andrew Rodney

-------------------------------------------------------------------------------

From: Andrew Rodney, INTERNET:andrew@digitaldog.net Date: Fri, Jan 19, 2001, 8:45 AM RE: Re: [colortheory] Enlarging Digital Files

on 1/18/01 6:45 PM, Dave Badger at dbadge@worldnet.att.net wrote:

> Does anyone on the list have experience with such a program and how well it > works? Are there any tricks for interpolating images up in Photoshop and > retaining or restoring detail?

On some images it works well and on others not as well as just using Bicubic interpolation inside of Photoshop. It1s not a free lunch. You need to ideally start with a file that1s at least 15-20mb in size! It1s VERY slow. Most pepole who rave about GF Pro never make the identical up rez in Photoshop to compare so when the prints come out looking half way decent, they are amazed. GF seems to work better on digital camera files in the tests I did. I1d say that IF your back is up against the wall and you just have to upsize a few hundred percent (and you have time), go for GF.

Andrew Rodney

-------------------------------------------------------------------------------

From: INTERNET:DMargulis@aol.com, Date: Fri, Jan 19, 2001, 8:46 AM RE: Re: [colortheory] Enlarging Digital Files

Dave writes,

> I get frequent calls from customers wanting to "res up" small files rather > then get a rescan (for printing). One told me today of a program (Genuine > Fractals?) that claims to enlarge up to 800% and still retain acceptable > detail.

Yes. Their advertising claim is, "Why scan and work on high-resolution images in Photoshop when you can get the same results from a low to medium-resolution original?...Genuine Fractals can make the enlargement look as sharp as the original."

Unfortunately, as is so often the case, there's little to this claim. I've tested the program extensively, on all kinds of images from all kinds of sources, against resampling up in Photoshop, and also, where available, against a higher-resolution scan of the same image.

1) For the majority of images there is absolutely no comparison between a high-resolution original and something rezzed up by either method.

2) The claim that GF rezzes up better than Photoshop is somewhat tenuous. I agree with the analysis made by Rob Outlaw. It's better than Photoshop at identifying edges (that's why Rob's method of sharpening first, then rezzing up in Photoshop, works for him) but it's not as good at noise reduction. Its handling of fleshtones is particularly poor and often results in what appears to be an impressionist-filter look in the skin.

3) GF consistently did better than PS when the magnification was 500% or more, or when the image featured a lot of fine detail with edges. PS often did better than GF when the original file size was small, and when there were artifacts of compression. This is probably why Rob isn't satisfied with his digital-camera upsizes.

In short, for certain images there's a marginal improvement, occasionally the program actually does worse than Photoshop. So, if you're doing a *lot* of these enlargements, it can pay to have the program, but the advertising claim quoted above is absurd.

Dan Margulis

-------------------------------------------------------------------------------

From: "Rob Outlaw", INTERNET:routlaw@IMT.NET Date: Fri, Jan 19, 2001, 11:13 AM RE: Re: [colortheory] Enlarging Digital Files Dan Margulis writes:

> 3) GF consistently did better than PS when the magnification was 500% or more, or when the >image featured a lot of fine detail with edges. PS often did better than GF when the original file size >was small, and when there were artifacts of compression. This is probably why Rob isn't satisfied >with his digital-camera upsizes.

Actually Dan I failed to mention this, but have tried comparing both methods not only with my D1 which provides a 7.5MB file but also with my Phase One scanning back with an initial capture of around 20-30MB. With few exceptions as stated before I still preferred the look of the PS method of upsizing. I certainly agree with you on flesh tone look with the GF method.

While we are on this subject I would be interested in hearing what others have to say about the following statement. Many including myself believe that you can resize an image captured digitally with much more leeway and extremity than with scanned film. Although I have never made a comparison with drum scanned film I have from the Kodak Pro and Master CD scans. Yes some amount of upsizing can be done succesfully but not near to the degree of a good digital capture be it from a single shot, or scan back. Adding to this I seem to remember hearing a report last year about a trade show in Germany where they compared a drum scanned 4x5 film with a rezed up Lightphase shot to somewhere in the neighborhood of 100MB if my memory serves me correct. The Lightphase image won in the resolution contest at least.

I should also ad that I have taken some of my D1 images from their initial 7.5MB size and interpolated them upto 128MB and output to the Ilfochrome equivalent of the Lightjet Sciences prints. The results were spectacular to say the least, but these were subjects shot from only a few feet away, not half way across a football field. My experience is that a camera like the D1 simply does not have initial res to make a succesful print under those infinity focus type of conditions, either from GF or PS interpolation but can be very succesful at images say from half way across a normal size room.

Rob Outlaw

-------------------------------------------------------------------------------

From: Joey Benton, INTERNET:jwbenton@interactivenet.net Date: Fri, Jan 19, 2001, 9:25 AM RE: Re: [colortheory] Enlarging Digital Files

on 1/18/01 8:45 PM, Dave Badger at dbadge@worldnet.att.net wrote:

> I get frequent calls from customers wanting to "res up" small files rather > then get a rescan (for printing). One told me today of a program (Genuine > Fractals?) that claims to enlarge up to 800% and still retain acceptable > detail. > > Does anyone on the list have experience with such a program and how well it > works? Are there any tricks for interpolating images up in Photoshop and > retaining or restoring detail? Thanks >

I use GF Pro and would say it works pretty well, but it is not a panacea. To me 800% demands a rescan. There is only so much data there and as you pull it apart something has to give. However, GF uses a better interpolation algorithm than Photoshop, so sampling up does seem to work better.

Joey Benton

-------------------------------------------------------------------------------

From: Andrew Rodney, INTERNET:andrew@digitaldog.net Date: Fri, Jan 19, 2001, 11:24 AM RE: Re: [colortheory] Enlarging Digital Files

on 1/19/01 9:06 AM, Rob Outlaw at routlaw@IMT.NET wrote:

> While we are on this subject I would be interested in hearing what > others have to say about the following statement. Many including > myself believe that you can resize an image captured digitally with > much more leeway and extremity than with scanned film. >

No question in my mind. I1ve done it many times. There was an article I wrote for PEI mag (Film verses digital) where I took a 2x2 chrome and actually scanned it to over 100mb and then took a file off a PhaseOne (2000x3000 18mb) and sized it the same size using Photoshop. The capture from the PhaseOne was far better in nearly every respect! There1s no film grain in digital capture which makes a huge difference when sizing up a file. That along with the vastly wider dynamic range makes film look pretty sick in comparison! In print (to the mag) as well as to digital output devices, the sized up file from the PhaseOne looked a lot better than the non sized file scanned on the drum scanner!

I think there may be a small section of each residing on the tips and tricks page at http://www.digitaldog.net.

Andrew Rodney

-------------------------------------------------------------------------------

From: INTERNET:DMargulis@aol.com, INTERNET:DMargulis@aol.com Date: Fri, Jan 19, 2001, 1:00 PM RE: Re: [colortheory] Enlarging Digital Files

Rob writes,

>> Many includingnmyself believe that you can resize an image captured digitally with much more leeway and extremity than with scanned film. Although I havenever made a comparison with drum scanned film I have from the Kodak Pro and Master CD scans. >>

I *have* compared it with drum scans. Yes, digital captures can be rezzed up with better results than any film-scanning method, which inevitably captures some of the film grain that isn't there in digital shots.

>>Adding to this I seem to remember hearing a report last year about a trade show in Germany where they compared a drum scanned 4x5 film with a rezed up Lightphase shot to somewhere in the neighborhood of 100MB if my memory serves me correct. The Lightphase image won in the resolution contest at least.>>

So the drum scan was originally double the resolution, maybe a bit more, of the Lightphase capture. It's certainly possible to come up with images for which this would be true; you need an image without fine detail and in which film grain is somewhat pronounced. Or, of course, if one is unintentionally or otherwise not producing the optimal drum scan.

Good digital cameras don't require as much resolution as the conventional wisdom suggests. However, rezzing them up to match a drum scan at double the resolution or higher is usually not going to produce a winner.

Dan Margulis

-------------------------------------------------------------------------------

From: Terry Wyse, INTERNET:terry@mail.allsystems.com Date: Fri, Jan 19, 2001, 4:52 PM RE: Re: [colortheory] Enlarging Digital Files

on 1/19/01 11:19 AM, Andrew Rodney wrote: > That along with the vastly wider dynamic range [of digital cameras] makes film > look pretty sick in comparison!

The statement about "vastly wider dynamic range" has me a bit perplexed. I thought digital cameras (good ones) had only about 8 stops of dynamic range where a decent drum scanner should have in excess of 10 stops (3.00 density range or better). And I believe transparency film should be on the same order of +8 stops as well. Enlighten me!

Terry _____________________________ Terence (Terry) Wyse PrePress Specialist All Systems Integration, Inc. 781.935.3322 voice 781.935.6622 fax http://www.allsystems.com terry@allsystems.com _____________________________

-------------------------------------------------------------------------------

From: INTERNET:DMargulis@aol.com, INTERNET:DMargulis@aol.com Date: Fri, Jan 19, 2001, 6:18 PM RE: Re: [colortheory] Enlarging Digital Files

Mac writes,

>>If I understand the gist of this correctly, using Pshop's bicubic would yield roughly the result as using the various "resolution enhancers" out there for the following type of usage:>>

I understand Rob and Andrew and I to all be saying this, yes. You might find certain images that would do better with GF and you might find others that would be better without it.

>>Presented to use for this: a jpeg of about 2MB (if I'm lucky, some could be smaller). In addition, some cropping will be needed. I have no control over any of the settings on the digital camera; do not even know who the user might be.>>

If the JPEG itself is 2mb, that probably is enough resolution for the full page, especially if the file is in RGB, which seems likely. It all depends on the level of compression, but typical ranges on digital cameras are 10-12/1. And, as we have been saying, you don't need quite as much resolution with digital captures as with conventional scans.

Dan Margulis

-------------------------------------------------------------------------------

From: Bob Smith, INTERNET:rmsmith@calpha.com Date: Fri, Jan 19, 2001, 6:22 PM RE: Re: [colortheory] Enlarging Digital Files

Terry Wyse wrote:

> I thought digital cameras (good ones) had only about 8 stops of dynamic range > where a decent drum scanner should have in excess of 10 stops (3.00 density > range or better). And I believe transparency film should be on the same order > of +8 stops as well. Enlighten me!

You're confusing the range of brightness that a transparency presents as seen by the scanner with the range of brightness of the original scene that the film recorded. A transparency laying on a light table usually presents greater contrast range than the same film can record. That's why you can't simply photograph a full range transparency in a copy situation with another transparency film and expect to record all of the information that was in the original. A scanner that can capture all of the info in that original piece of film usually has to be MORE capable than the film itself. The technology that's driving that scanner is very similar to what's driving the digital cameras.

The real beauty of digital cameras is that the info that they capture remains much more linear over a broader range than film. Film will always record color and contrast shifts as you move further into highlight and shadow areas. A digital camera can more easily maintain gray balance and relative contrast throughout the range so you wind up with more truly useable data.

Bob Smith

-------------------------------------------------------------------------------

From: Andrew Rodney, INTERNET:andrew@digitaldog.net Date: Fri, Jan 19, 2001, 6:57 PM RE: Re: [colortheory] Enlarging Digital Files

on 1/19/01 2:52 PM, Terry Wyse at terry@mail.allsystems.com wrote:

> on 1/19/01 11:19 AM, Andrew Rodney wrote: >> > That along with the vastly wider dynamic range [of digital cameras] makes >> film >> > look pretty sick in comparison! > > The statement about "vastly wider dynamic range" has me a bit perplexed. I > thought digital cameras (good ones) had only about 8 stops of dynamic range > where a decent drum scanner should have in excess of 10 stops (3.00 density > range or better). And I believe transparency film should be on the same > order of +8 stops as well. Enlighten me! > With some of the scanning cameras, the range has been as high as 13 stops! Again, I have files shot with a BetterLight scanning back and a PhaseOne Lightphase side by side shot with file (bracketed to boot) and the degree of tonal range simply puts film to shame. Have you ever seen any of Stephen Johnson1s work with the older Dicomed digital scanning back of the national forests? The quality is amazing.

Andrew Rodney

-------------------------------------------------------------------------------

From: "Jim Domnitz", INTERNET:jdomnitz@studiod-online.com Date: Sat, Jan 20, 2001, 11:04 AM RE: [colortheory] Film vs Digital

We have been doing digital capture since the end of '96. One thing that we have found is that in capturing images of products on film, lighting is so critical if there are extremes in the image. Imagine photographing some tennis shoes where the majority of the image is white and the rest may be a dark blue or black. With film we have to compromise the values.. hold the detail in the dark areas and blow out the highlights, or hold the highlights and loose the details in the dark areas... Now with digital capture, we shoot for the highlights, then for the shadows and combine the images.. This can be done with film, but perfect alignment is an issue. The color is great and my customers have been very happy,

Jim Domnitz

-------------------------------------------------------------------------------

From: Bob Smith, INTERNET:rmsmith@calpha.com Date: Sat, Jan 20, 2001, 10:48 AM RE: Re: [colortheory] Enlarging Digital Files

Terry Wyse wrote:

> Guess it's time to drag out the 4x5 and shoot a few tests...

Go for it. You're more motivated than I. I process so little film now that I've got to be really moved to get into the darkroom. My gut feeling is that a zone system style testing of a typical transparency film might show recorded detail across a fairly large range but the usefulness of the data on those last couple of stops on each end is questionable. Neg film will record over a much larger range.

The rule of thumb that photographers have been taught for ages is to light the set to keep reproducible detail within a four stop range when shooting a transparency that's destined for typical 4 color press reproduction. Clearly the film is capable of recording data over a greater range; and a good scanner operator has the ability to compress that data into a useable range for the press. However I think the reason the 4 stop rule has been so widely stressed is that when you venture outside that 4 stop area you're far more likely to have an image that shows some pretty dramatic gray balance shifts and contrast changes. It becomes a challenge for the scanner operator to try and hold that detail in those deep shadows or nearly burned out highlights while not toasting something else in the image. This is where the digital captures really shine. Data at those extreme tones is going to be far more useable. A photographer should still strive to control lighting for a more realistic printable range on the subject, but for those instances where that's not possible or practical a direct digital capture can really save the day.

Several months ago I shot a small table top set that was destined to be reproduced as a large poster. I didn't feel that the digital camera that I was using could carry off an image that large as well as film so we arranged the set using the digital camera sort of like an instant Polaroid and then shot the final image on large format transparency film. The art director loved the last digital image and took that to start building the layout while I processed and scanned the film. I placed the digital capture on the screen while I was doing the scan in an attempt to get the high res scan to closely match the digital capture. It was futile. I fought that scan for an hour or more and wound up with an image that created a very nice poster but it never had as smooth a tonality as the digital capture had. Such an exercise is a great way to hammer home the differences between film and digital. It sure enlightened me.

Bob Smith

-------------------------------------------------------------------------------

From: Terry Wyse, INTERNET:terry@mail.allsystems.com Date: Sat, Jan 20, 2001, 9:26 AM RE: Re: [colortheory] Enlarging Digital Files

on 1/19/01 6:21 PM, Bob Smith wrote:

> You're confusing the range of brightness that a transparency presents as > seen by the scanner with the range of brightness of the original scene that > the film recorded.

I should've clarified. I understand that trans film has a DENSITY range in excess of 10 stops (about .15-.20 Dmin to 3.2-3.4 Dmax?) which of course isn't the same thing as the scene brightness range it can capture, but even the "scene" I thought would be in the 8-10 stop range. If not, I'll guess we'll have to revise the Zone System to accomodate trans film. Ansel, where are you when we need you? :-)

And Andrew, I had no idea that digital was out there in the +10 stop range. amazing!

Guess it's time to drag out the 4x5 and shoot a few tests...

Terry "I feel like about a Zone 3 right now" Wyse _____________________________ Terence (Terry) Wyse PrePress Specialist All Systems Integration, Inc. 781.935.3322 voice 781.935.6622 fax http://www.allsystems.com terry@allsystems.com _____________________________

-------------------------------------------------------------------------------

From: INTERNET:DMargulis@aol.com, INTERNET:DMargulis@aol.com Date: Sat, Jan 20, 2001, 11:28 AM RE: Re: [colortheory] Enlarging Digital Files Al Ferreira writes,

>>Dan, If I understand your statement, are you saying that digital captures can be outputted with less than the standard 2X line screen? Could you clarify?>>

2X line screen is a simplistic rule. Optimal resolution depends not just on the screen ruling (the higher the screen ruling, the less of a multiple is needed) but on the character of the image.

I endorse scanning at 2x screen ruling, even though you don't need that much, because you never know when you are unexpectedly going to have to enlarge the image. But at 133 line screens and over, 1.5x is ordinarily sufficient.

The danger of proceeding with insufficient resolution is that the picture can look coarse or jagged in print. Digital captures are less subject to this effect than film is, because their original data is smoother. Film is a very sensitive medium but it is also somewhat inconsistent, and one picks up grain and other artifacts, particularly on fast exposures. Plus, it then has to be processed, which introduces further artifacts, and then scanned, which introduces more.

This is all manageable but it can become a problem at low resolution. The much smoother data of a digital is less likely to cause this. Nobody I know has yet documented this effect, but my guess is that, with a reasonable digital capture, about 1.2x is all you need at 133 or 150 lines.

Dan Margulis

-------------------------------------------------------------------------------

From: INTERNET:gowens01@coin.org, INTERNET:gowens01@coin.org Date: Sun, Jan 21, 2001, 11:49 AM RE: [colortheory] Re: Film vs Digital

--- In colortheory@egroups.com, "Jim Domnitz" wrote: > We have been doing digital capture since the end of '96. One thing that we have found is that in capturing images of products on film, lighting is so critical if there are extremes in the image. Imagine photographing some tennis shoes where the majority of the image is white and the rest may be a dark blue or black. With film we have to compromise the values.. hold the detail in the dark areas and blow out the highlights, or hold the highlights and loose the details in the dark areas... Now with digital capture, we shoot for the highlights, then for the shadows and combine the images.. This can be done with film, but perfect alignment is an issue.

The alignment issue can be solved by putting a registration mark in the picture. Shoot the highlight image; then shoot the shadow image. When you select the shoe include the regristration mark in the selection. then when you combine the shodaw and detail you can use the arrow keys to align the regristration mark. After that the regristration mark can be erased or masked.

This has worked well for me when I have combined images of horses jumping over fences. I select part of the fence as a regristration mark.

Gary Owens

The color is great and my customers have been very happy, > > Jim Domnitz