Dan Margulis Applied Color Theory
Resolution for the Future
Date: Sat, 12 Apr 2003 11:54:06 -0400
From: Denise Romeo
Subject: Resolution for the future
Hello everyone. I am a photoshop retoucher in Manhattan and the owner of my own shop. I have been retouching for over 10 years, mostly all fashion work. I have worked for Quad graphics, AGT, Conde Nast publications and most recently Box Ltd. Before opening my own shop.
Here is my question. My clients are famous fashion photographers. One of them asked me to start to archive his past work, in other words, scan his older negatives. We are going to do so many a month. We are trying to come to some agreement on resolution and file size so that these files can be used for many purposes in the future. Our right now we are considering future book projects, then large format output to Lightjet for exhibition, then I was going to do the grey scale at a much higher resolution for LVT output to neg should we want to make an edition of silverprints for exhibition. So you see , its quite a mixed bag. I have a Hell 3300, and am going to scan all the color in RGB and keep it that way, to convert later depending on the usage and output. That is my workflow for everything anyway. But the real question is resolution in the future. My photographer thinks we should make 600 meg files on everything, just so that in 10 years from now we will be safe. I think this is wrong. For some reason if I had to make an educated guess, I would say that in 10 years we are not going to need such big files anymore. I was thinking a nice 200 meg RGB file that can be sized down for reprint in magazines, or sized up for mural size light jet or whatever technology can come up with to make large format C-prints. We are not really thinking about inkjet, these guys like their c-prints. Anyway, I know its a long question.... But if you were a famous photographer, and you were going to digitize your library, What resolution and files size would you pick considering usage 10 20 years from now??
Ps... I scan very flat and open allowing for max detail in highlight and shadow and put an adjustment curve on top, and save my psd files this way so that I can readjust color per output device, so I am not concerned really about color. Working this way, I never commit myself to any contrast or color, I can always readjust without rescanning.
Thanks so much.. Enjoyed all your posts.. Have been a Dan fan for many many years. Never worked with him, almost once, at Cardinal!
Sincerely,
Denise Romeo
451imaging
____________________________________________________________________________
Date: Sat, 12 Apr 2003 10:23:34 -0600
From: Andrew Rodney
Subject: Re: Resolution for the future
on 4/12/03 9:54 AM, denise wrote:
But the real question is resolution in the future. My photographer
thinks we should make 600 meg files on everything, just so that in 10 years
from now we will be safe. I think this is wrong. For some reason if I had
to make an educated guess, I would say that in 10 years we are not going to
need such big files anymore.
You1ll never know for sure. I1d scan at the highest optical resolution of your scanner in high bit with an input profile for the scanner (don1t convert into a Working Space). And yes, scan flat! Give yourself headroom on either end (with the 16 bit file, you1ll have the option to set end points latter). Keep USM off or to a bare minimum.
Andrew Rodney
____________________________________________________________________________
Date: Sun, 13 Apr 2003 02:01:35 +0100
From: Martin Bailey
Subject: Re: Resolution for the future
At 16:54 12/04/2003, denise wrote:
Ps... I scan very flat and open allowing for max detail in highlight and
shadow and put an adjustment curve on top, and save my psd files this way so
that I can readjust color per output device, so I am not concerned really
about color. Working this way, I never commit myself to any contrast or
color, I can always readjust without rescanning.
Resolution isn't the only issue here. Digital information is amazingly fragile when you look at anything approaching long-term storage.
You're saving .PSD files? Will you have any tools to read the current version of those files in 10 years' time? I suspect that TIFF may remain more readable, at least if you stick to a 'simple' 16 bit-per-channel RGB.
What are you storing them on? Will you be able to read a CD in 10 years' time? Probably, but remember to build in a long-term plan for refreshing media - two copies of every disk and re-burn every so many years, reviewing the choice of media each time (I believe that the Guggenheim works on a 10 year cycle).
Oh yes, and ZIP and LZW compression don't damage the data at all. I'd suggest ZIP rather than LZW because the latter is going out of fashion real fast and ... you guessed it ... you may not have tools to read it in 10 years' time!
Regards
Martin Bailey
-------------------------------------------------------------
Senior Technical Consultant +44 1223 873800
Global Graphics Software http://www.globalgraphics.com
Harlequin - Jaws - MaxWorkFlow
martin.bailey@globalgraphics.com
-------------------------------------------------------------
If my views didn't usually coincide with those of my employer
I wouldn't want to work here, but I am not a spokesman for
Global Graphics Software
and the buck stops with me for what I say.
------------------------------------------------------------
____________________________________________________________________________
Date: Sun, 13 Apr 2003 10:55:59 EDT
From: Dan Margulis
Subject: Re: Resolution for the future
Denise writes,
if you were a famous photographer, and you were going to digitize your
library, What resolution and files size would you pick considering usage 10
20 years from now??
You'll go nuts if you try to cater to every eventuality. Suppose he has a shot of a model showing full body from head to toe. 15 years from now the model is very famous. Now he wants to crop close in on her face and output it to a 16 x 20 piece of film, if film even exists 15 years from now. In such a case, even the 600 mb scan your client advocates won't be nearly enough.
Furthermore, with a drum scanner, such extreme magnifications produce a quality hit if you later need to downsize them drastically. Now granted, fashion work can probably accept the overly soft rendition caused by excessive resolution better than almost any other type of photography. But still, one has to suppose that the *likeliest* use for each photo is going to require <20 mb of data. In that case, coming down from 600 mb is going to give distinctly inferior quality as opposed to scanning at approximately the right resolution in the first place. The 200 mb scan you propose might be acceptable but I wouldn't be that happy about it.
You're right when you say that required resolutions are likely to go down in the future. And of all the applications you mention, by far the most resolution-demanding is film recorder, which is going to be less and less of a factor as time goes on.
A lot depends on how many images there are and how likely huge outputs are. If the library is of any significant size scanning even at 200 mb is going to take a long, long time and a huge amount of storage space.
Plus, you have to look at re-archiving everything at least once in the next 15 years, just because what you write to today won't be readable then, either because the media has deteriorated or the format isn't valid any longer. If this project had been done 15 years ago, for example, the two methods of storage that vendors represented as ideal, archival quality were WORM disks and 400 mb opticals. Within 10 years it was hard to find anybody who could read them and today it would be nearly impossible. Or even 10 years ago, the storage method might have been 44 mb Syquests and larger Bernoulli disks. Who can read these today?
So, basically it's a crapshoot. Your 200 mb suggestion is reasonable. If there are >100 images involved I'd be inclined to scan them all to 50 mb, and consult with the client as to which ones he thought were the real prizes. I'd scan those at a higher resolution *in addition* to the smaller scan. But neither you nor I nor your client has a crystal ball.
Dan Margulis
____________________________________________________________________________
Date: Sun, 13 Apr 2003 16:43:46 +0100
From: Richard Kenward
Subject: Re: Resolution for the future
Dear Dan
I am beginning to think that film is not such a bad bet after all! It is not so difficult to keep archived in good condition given a modicum of care, Takes up very little space relative to it's potential Mb of information and at the moment at any rate it can be scanned to produce outstanding results.
What proportion of great digitally archived pictures of today will be available in say 100yrs time and if so at what cost? Also how many great pictures from today will be lost forever due to some digital calamity of some sort?
Sometimes simple IS beautiful!
Cheers
Richard
____________________________________________________________________________
Date: Sun, 13 Apr 2003 13:49:33 -0400
From: Jonathan Clymer
Subject: Re: Resolution for the futureyy
My photographer
thinks we should make 600 meg files on everything, just so that in 10 years
from now we will be safe. I think this is wrong. For some reason if I had
to make an educated guess, I would say that in 10 years we are not going to
need such big files anymore. I was thinking a nice 200 meg RGB file that
can be sized down for reprint in magazines, or sized up for mural size light
jet or whatever technology can come up with to make large format C-prints.
You didn't say what size the original negs are. I would think that a 600 meg file from a 35mm negative would be a complete joke. Even for medium format extremely high.
You are considering the resolution needed to meet the demands of output, but you also need to consider the quality of the input available to justify it. Your post implied that the images are fashion, which further implies an SLR camera which is the equipment with which most fashion images are created.
A Kodak lab tech showed me some film images at extremely high resolution which were shot with a variety of cameras and lenses under controlled conditions. Among the many interesting observations was one which surprised me: the way to get the sharpest image is to turn off all the ambient lights, open the shutter, let the camera stabilize, fire a flash, and then end the exposure. If you don't let the camera stabilize, the image will be softer due to mirror slap (for SLRs) and shutter acceleration, even if the camera is on a heavy tripod, and even if the camera is a calibrated professional camera. If you don't do this (and fashion photographers due to the nature of their work cannot do this) it doesn't mean that the image is bad, but it does mean that at some point further detail yields grain (or pixels), not image information.
At what point does higher resolution become useless? I don't know, but it would vary with the format and the conditions under which the photographs were made. Why don't you do a couple of tests and let us know what your results are?
Jonathan Clymer
____________________________________________________________________________
Date: Sun, 13 Apr 2003 14:13:48 EDT
From: Dan Margulis
Subject: Re: Resolution for the future
Richard Kenward writes,
I am beginning to think that film is not such a bad bet after all! It
is not so difficult to keep archived in good condition given a modicum
of care, Takes up very little space relative to it's potential Mb of
information and at the moment at any rate it can be scanned to produce
outstanding results.
Exactly right. If you want to have something available in 50 years, the best bet, by far, is film archived in controlled climate conditions. And there's no "at the moment" about it. The quantity of photographic material in this world that is *not* available in digital form is incomprehensibly large. So, film may die, but never scanning.
I've inherited box after box of family photographs that date from 1890-1930. Many of them are in quite good shape. I don't know if anybody will be interested in them in 200 years but if they are, there will be plenty of scanners available to digitize them.
Dan Margulis
___________________________________________________________________________
Date: Sun, 13 Apr 2003 10:26:11 -0600
From: Andrew Rodney
Subject: Re: Resolution for the future
on 4/13/03 9:43 AM, Richard Kenward wrote:
What proportion of great digitally archived pictures of today will be
available in say 100yrs time and if so at what cost? Also how many
great pictures from today will be lost forever due to some digital
calamity of some sort?
If handled correctly, the vast majority. Film for proper storage for the long term needs to be frozen with assurance that no humidity affects the film. Even then, it1s like to undergo some degradation.
A digital file is just 11s and zero1s and you can make as many identical clones as you wish. Storage is just getting cheaper. I remember when a CR-R cost $6. Now they can be had for cents. DVD is coming down and holds so much more data. Yes, you will need to copy that data in 10-15 years onto something that will likely hold way more data and cost less. But the transition causes no data loss and users can have as many copies as they feel safe having.
The bigger and more important question is how many of today1s 2D images will be viewed or even useful 100 years from now? There will of course be many that will be for historical reasons. But the vast majority of the billions of images captured a year (ebay alone accounts for nearly 1 billon images a year), will be unnecessary to archive.
Andrew Rodney
____________________________________________________________________________
Date: Sun, 13 Apr 2003 22:08:27 +0200
From: Claudio Corvino
Subject: Re: Resolution for the future
I remember that about 20 years ago I read about a Kodak's project for preserving photographs for the future: they were going to save for each picture four process lith films...
____________________________________________________________________________
Date: Sun, 13 Apr 2003 16:31:53 -0400
From: Dale Hoffman
Subject: Re: Resolution for the future
At 10:55 AM -0400 4/13/03, Dan Margulis wrote:
Furthermore, with a drum scanner, such extreme magnifications produce a
quality hit if you later need to downsize them drastically. Now granted,
fashion work can probably accept the overly soft rendition caused by
excessive resolution better than almost any other type of photography. But
still, one has to suppose that the *likeliest* use for each photo is going to
require <20 mb of data. In that case, coming down from 600 mb is going to
give distinctly inferior quality as opposed to scanning at approximately the
right resolution in the first place.
Please explain how this degradation works and what the limits of safe down sampling might depend on.
DaleH
____________________________________________________________________________
Date: Sun, 13 Apr 2003 15:41:02 -0600
From: Andrew Rodney
Subject: Re: Resolution for the future
on 4/13/03 2:31 PM, Dale Hoffman wrote:
Please explain how this degradation works and what the limits of safe down
sampling might depend on.
I think it1s hooy. In fact, scanning at a much higher resolution than you need is actually a benefit. Noise in shadows is usually a single pixel among good data. When you sample down 4 pixels to 1, that noise is removed! So doing a step resolution (up or down) produces superior results and in the case of sampling down, the end results are usually a removal of noise. I find that a compelling reason to scan at a higher resolution than ultimately desired!
Andrew Rodney
____________________________________________________________________________
Date: Sun, 13 Apr 2003 20:14:56 -0700
From: Dennis Dunbar
Subject: Re: Resolution for the future
One of the most important details to think of when scanning an image is how much information is there in the original image. Basically when it gets down to subsampling the film grain aren't you wasting your time and disc space?
If you have a very well exposed 4x5 tranny on fine grain film at what point are the pixels smaller than the film grain? If it's fashion images you're working on I'd bet most of the images are 2 1/4 trannies or negs. If so, then you reach that point of subsampling the grain even sooner.
In the time I've been doing digital, (since 1990), I've heard lots of myths propagated about the resolution necessary for a "good" file. Output size matters a lot, but so does the original. It used to be that "experts" said you needed at least 300Mb images to produce a decent 8x10 LVT. Turns out that was basically just a sales pitch and was not grounded in fact at all.
My suggestion would be for you to test scan a sampling of images from this photographer at various resolutions. Then examine them carefully to see at what resolution the grain is bigger than the pixels. This should give you your answer. What's more, if you include the photographer in the test he'll feel like he's learned something about this issue and thank you for great service. Besides cementing your relationship you'll also have insurance against having him tell you later that what ever choice you made was wrong.
Just my 2 cents.
Dennis Dunbar
____________________________________________________________________________
Date: Sat, 12 Apr 2003 10:23:34 -0600
From: Andrew Rodney
Subject: Re: Resolution for the future
denise wrote:
But the real question is resolution in the future. My photographer
thinks we should make 600 meg files on everything, just so that in 10 years
from now we will be safe. I think this is wrong. For some reason if I had
to make an educated guess, I would say that in 10 years we are not going to
need such big files anymore.
You1ll never know for sure. I1d scan at the highest optical resolution of your scanner in high bit with an input profile for the scanner (don1t convert into a Working Space). And yes, scan flat! Give yourself headroom on either end (with the 16 bit file, you1ll have the option to set end points latter). Keep USM off or to a bare minimum.
____________________________________________________________________________
Date: Mon, 14 Apr 2003 00:06:09 -0700
From: Kevin Connery
Subject: Re: Resolution for the future
This is drifting off-topic, but...
One question behind the 'how many will be useful' question (or answer) is when will we know whether something's important or not? A lot of 'important historical documents' from the US Civil War are still surfacing now, 140 years later. Many of them were NOT considered important then.
Film, for all it's flaws, degraded gracefully in most situations, and, even if kept in kinda-sorta-OK conditions and ignored for multiple decades, would still contain some--albeit nowhere near all--of the original information. Witness the snapshots of your parents' or grandparents' youth, ignored until after their deaths. Had that been on something which required any active level of handling, it'd be effectively lost. (Effectively, primarily due to the expense of even identifying what it was, more than the absolute inability to find something to read a 40+ year old format and media.
With a modicum of active care--as noted, all of the archiving bodies now have a regular update process--digital stuff will be identical to its original form. Without that activity...
--kdc
____________________________________________________________________________
Date: Mon, 14 Apr 2003 09:35:36 EDT
From: Dan Margulis
Subject: Re: Resolution for the future
Dale writes,
Please explain how this degradation works
and what the limits of safe down
sampling might depend on.
"Degradation" isn't the right word. When there's an area that's all of approximately the same color (like a man's face, or a lawn), the higher the resolution of the scan, the more uniform the color will appear, because there will be more samples averaged together to make the final halftone dot or whatever else the output is going to be. With less resolution, a pixel that's unusual in relation to its neighbors will be given much, much more weight. So, with higher resolution, the image looks smoother, with lower resolution irregularities are magnified and there seems to be more action.
This is a two-edged sword. Higher resolution cuts noise, which is good. Unfortunately, it also can cut detail--or at least the appearance of detail. In the picture of the man's face, if you perceive that he needs a shave, it's not because you can make out the individual whiskers. You're seeing some kind of action in his cheeks and your brain is whispering whiskers into your ear.
If you drum-scan this image at a significantly higher resolution, it will smooth that action out. You may want that and you may not. This is why a lot of high-end shops specializing in fashion work scanned women's faces at higher resolutions than those of men, knowing that the extra smoothness would be desired.
OTOH, in the image of grass, you can't make out individual blades and again you're relying on your brain to interpret the variation it sees. With an ultra-high resolution scan, the grass will become AstroTurf and the normal resolution will give a superior result. And similarly with anything textured where the viewer can't actually resolve the detail.
As for when the effect becomes undesirable, that's image-specific and hasn't been studied much, particularly because it is a high-end scanner issue only (cheap desktop models scan at the highest resolution and res it down, whether you like it or not). Also, to get into ultra-high resolutions, you have to have ultra-large amounts of storage space and computing time, which doesn't interest many people.
The conventional wisdom, to which I don't fully subscribe, is that you should scan at between 1.5-2 times the eventual screen ruling. I am completely comfortable with ratios of up to 3/1 or even 4/1: the extra smoothness will be there but it won't make enough difference to bother anyone. In my book, I show a sample (involving grass, natch) where one scan is at 2/1 and the other 6/1. There, the normal-res version is clearly superior and it isn't obvious that any form of sharpening the higher-res would equal it. And yet the higher-res version takes up nine times as much disk space.
Denise's client was talking about a 600 mb scan for output that might only normally require 10 mb. That's going to yield about a ratio of 16/1 in comparison to the normal 2/1, or approximately 256 scanning samples averaged to construct a halftone dot as opposed to the normal 4.
Now I grant you that I haven't actually tried this because I'm not excited about files that are 64 times as large as they need to be. However, I'd be flabbergasted if those ultra-high res drum scans, when taken down to a normal size, weren't soft and defocused in comparison to the same artwork scanned normally.
Many of the issues discussed in this thread are covered in the resolution chapter of my book, which is freely available at
http://www.ledet.com/margulis/PP7_Ch15_Resolution.pdf
Dan Margulis
____________________________________________________________________________
Date: Mon, 14 Apr 2003 10:15:05 -0400
From: Jim Rich
Subject: Re: Resolution for the future
Dan,
I just wanted to throw this into this discussion because I am not sure that general rules of thumb apply in this discussion about image quality and drum scanning.
Some of the points you make have merit, but it depends on the actual product you are producing with the drum scanner (and the brand of drum scanner) and the way the focus and the USM options are setup.
For example you have said that
In the picture of the man's face, if you perceive that he needs a shave, it's
not because you can make out the individual whiskers. You're seeing some kind
of action in his cheeks and your brain is whispering whiskers into your ear.
If you have the scanners USM set one way you can smooth out or sharpen the mans whiskers so it is hard to say this is going to happen even the your brain telling you to look at the whiskers.
Jim Rich
____________________________________________________________________________
Date: Mon, 14 Apr 2003 18:16:20 +0100
From: "Paul Graham"
Subject: re: resolution for the future
I was told that DVD-RAM was the preferred medium for long term storage, and ever since then have archived backup copies on that if you have a LOT of images to backup, then the other simple thing to do is to just dump everything onto a hard drive (once a year, or so) and take that out and put it away. as a backup of your backup. it's quick, easy and a 250Gb 7200rpm drive is now only about $230, so it's cheap too.
paul
____________________________________________________________________________
Date: Mon, 14 Apr 2003 14:06:21 EDT
From: Dan Margulis
Subject: Re: Resolution for the future
Jim Rich writes,
If you have the scanners USM set one way you can smooth out or sharpen the
mans whiskers so it is hard to say this is going to happen even the your
brain telling you to look at the whiskers.
True, but for the use Denise has in mind you really have to be ultra-conservative with the scanner's USM setting. A 600 mb scan downsized to 10 mb is going to need an enormous amount of sharpening, enough so to ruin it for any output that's significantly larger. Yet if you apply more than an absolute minimum USM to the 600 mb scan, it becomes almost impossible to resharpen it in Photoshop if it's necessary to downsize the file significantly.
In any situation where the same file has to be used to generate two different high-quality outputs, automated color management isn't going to be a complete answer but it usually gets one close. But the real killer is always the sharpening.
Dan Margulis
____________________________________________________________________________
Date: Mon, 14 Apr 2003 21:18:16 +0100
From: Richard Kenward
Subject: Re: Resolution for the future
DMargulis writes
In any situation where the same file has to be used to generate two different
high-quality outputs, automated color management isn't going to be a complete
answer but it usually gets one close. But the real killer is always the
sharpening.
Dear Dan
Sorry but I just don't see the problem......as applying sharpening will impact on the future use of the master file, why apply any at the scan stage. Yes I know some folk like to put some on when scanning, and yes drum scanner software allows many options, but on a decent drum scan it's already nice and sharp without sharpening. It just seems logical to me to apply exactly what is required when re-purposing the file for whatever size and application and keep the master pure.
Cheers
Richard
____________________________________________________________________________
Date: Mon, 14 Apr 2003 16:52:38 -0400
From: Jim Rich
Subject: Re: Resolution for the future
Richard,
Most all drum scanners need to apply some degree of USM. That might mean a digital value has to be input or an aperture needs to be set or a combination of both. It depends on the vendor of the scanner. And in some cases if you actually turn off all of the scanners USM then the image becomes blurry.
Since each scanner is different they will have different USM settings as well as each supplier will have different starting USM values which will ultimately affect the master scan. Once you come to the point where you have just enough USM then you begin your scan. I thought it was important to point that out because I think that is what you are referring to.
Jim Rich
____________________________________________________________________________
Date: Mon, 14 Apr 2003 17:51:06 EDT
From: Dan Margulis
Subject: Re: Resolution for the futureyy
Richard Kenward writes,
Sorry but I just don't see the problem......as applying sharpening will
impact on the future use of the master file, why apply any at the scan
stage. Yes I know some folk like to put some on when scanning, and
yes drum scanner software allows many options, but on a decent drum scan
it's already nice and sharp without sharpening. It just seems logical
to me to apply exactly what is required when re-purposing the file for
whatever size and application and keep the master pure.
I think you have misinterpreted my post because we appear to be saying the same thing. If you apply significant sharpening on the scanner, it severely limits alternate uses of the file.
That said, a lot of drum scanners *aren't* well focused without at least a small amount of USM. Remember, their design dates from a time when only a madman would try to sharpen outside of the scanner. So, how the scan looks with USM disabled altogether was not a big concern, and on a lot of drum scanners it looks pretty bad. So one often has to apply a tiny bit of sharpening at least.
But I agree, if you don't know the destination of the file, you shouldn't sharpen any more than absolutely necessary on the scan.
Dan Margulis
____________________________________________________________________________
Date: Tue, 15 Apr 2003 00:09:00 +0100
From: Richard Kenward
Subject: Re: Resolution for the future
Dear Dan and Jim
No I got you right but I am coming from a situation where we run our two drum scanners totally without any USM being used. In fact the USM module has been removed from the software just so that it's never applied in error!
I have to say that with no USM on at scan time these scans are sharp. This is not to say that the files do not benefit from careful use of sharpening techniques later on, they do. But it's nice to start with a scan in this condition <BG>
But I agree, if you don't know the destination of the file, you shouldn't
sharpen any more than absolutely necessary on the scan.
Totally agree. Good time for me to close down for the night!
Best regards
Richard
____________________________________________________________________________
Date: Mon, 14 Apr 2003 19:39:59 -0400
From: Lee Clawson
Subject: Re: Resolution for the future
Denise,
Here's a list of what turned out to be helpful archiving 35mm up to 4x5 transparencies.
1-- Plan for client needs. We approached this by assuming we might not be in business, media might not exist, software might not exist and/or may not recognize the (older) formats. The client should start with a duplicate of the entire archive off site. And my experience is that its easy to maintain an archive out to 5 yrs. after that its more time consuming.
2-- Make a way to identify the film stock/emulsion. Things like film notch codes and other "marks" are usually forgotten. And most scans don't show these outer areas anyway. We included on each CD a set of color targets for future reference.
3-- And I agree with Dan's desire for more than one resolution. With the very good optics of a drum scanner (we had a Crosfield) the ultra high resolution isn't necessarily better. Going from very large to very small is still very tricky. We scanned 4x5's at approx 150mgs and made another scan at 50mgs. I think 35mm was over just 200mg.
4-- Also scanned "flat" and used low USM. But flat was confusing as density range changed with emulsion and neg vs trans. In practice we calibrated for and gang scanned groups of similar emulsions/film types.
5-- Regarding file formats. We used TIFF or PSD with no compression, no layers. At the time, 5 years ago, we used 8-bit for the archive.
Other comments....
--I'd ask the client to keep the originals. Film is still quite an amazing recording media.
-- Ask the client to create (with you) a way to find images in the archive.
-- If your client hasn't worked "flat scans" explain early on what this is.
Lee
____________________________________________________________________________
Date: Tue, 15 Apr 2003 07:25:41 -0400
From: "john c."
Subject: Re: Resolution for the future
Folks,
I've been following this thread, and the remarkable thing that's come out of it is the "new" respect that film is getting as an archive medium. Whatever the scan resolution is, people are now saying to keep the original film as the archive. Of course this is counter to the last two decades' rush to archive film to digital formats, but I agree entirely. The question I keep coming back to is whether one shouldn't consider archiving digital photos on film! I think the original post in this thread involved 8x10 LVT output at some future time.
Comments please...
john castronovo
____________________________________________________________________________
Date: Tue, 15 Apr 2003 08:51:03 -0400
From: Lee Clawson
Subject: Re: Resolution for the future
Yes I know some folk like to put some on when scanning, and
yes drum scanner software allows many options, but on a decent drum scan
it's already nice and sharp without sharpening.
I can't remember anytime USM was "off".
Lee
____________________________________________________________________________
Date: Tue, 15 Apr 2003 12:36:24 -0500
From: "jsaale"
Subject: Re: Resolution for the future
Great Discussion! Timely too, as I'm scanning B&W wedding negs as a backup of the actual film.
I searched Google for "flat scans" and "scan flat" but can't get near the discussion point. So tell me early on what you mean by flat scans.
Jerry Saale
____________________________________________________________________________
Date: Tue, 15 Apr 2003 11:52:22 -0600
From: Andrew Rodney
Subject: Re: Re: Resolution for the future
on 4/15/03 11:36 AM, jsaale wrote:
I searched Google for "flat scans" and "scan flat" but can't get near the
discussion point. So tell me early on what you mean by flat scans.
Headroom on either end of the tone curve (Histogram). So you have some detail/tone in the brightest highlights and shadows. This allows you to set the precise end points LATTER in something like Photoshop (best way for me is using Levels, holding down the option key and sliding the black/white input sliders while viewing the high contrast preview. All the pixels show where you clip to either zero or 255 in each color channel).
Do this in high bit! Don1t scan flat in 24 bit. With a high bit scan, you can move the end points latter instead of at scan time without data loss. Frankly, the only way to scan when you don1t know the final output requirements. Also, USM can push highlights over the edge if they are close to the clipping point so having a bit of headroom is nice to have (and safe).
Andrew Rodney
____________________________________________________________________________
Date: Tue, 15 Apr 2003 23:20:12 -0400
From: "Preston Earle"
Subject: Re: Headroom and 16-bit; was Resolution for the future
Speaking of "flat" scans, "Andrew Rodney" wrote: "Headroom on either end of the tone curve (Histogram). So you have some detail/tone in the brightest highlights and shadows. This allows you to set the precise end points LATTER in something like Photoshop (best way for me is using Levels, holding down the option key and sliding the black/white input sliders while viewing the high contrast preview. All the pixels show where you clip to either zero or 255 in each color channel).
Do this in high bit! Don't scan flat in 24 bit. With a high bit scan, you can move the end points latter instead of at scan time without data loss."
-----------------------
With any scanner, shouldn't you set the end points in the scanner to 0-255 for the lightest/darkest (or is it darkest/lightest?) points the scanner sees in the image? Depending on the quality of the scanner, these points may be different for the same image scanned on different scanners, but for any one scanner, leaving "headroom" between the light/dark point and 0/255 doesn't seem to make sense. Am I missing something?
With the discussion on 8bpc vs. 16bpc scans, I tried a test. I scanned an image on my Minolta Scan Dual II scanner using Vuescan in 16bpc mode. This was an old 35mm Kodachrome slide scanned at 2810ppi. When looking at areas of this scan that should be of uniform color, such as adjacent pixels in a blue sky, or adjacent pixels in a white-painted building, I couldn't find two adjacent pixels with exactly the same values. Even checking pixels that looked identical on the screen, they generally measured at least 2 units difference in at least one of the three values (RGB) and frequently differed by 3-5 units in one color or another.
This readout (single-point color sampler) was 8-bit (0-255), so the 16-bit values were rounded (or truncated or whatever). Even so, I couldn't find *any* same-color adjacent pixels. Considering this, do the high-bit values have any real value? If adjacent-pixel values are different in 8-bit precision, does it make any sense to save them with 16-bit precision? Do other, more expensive (or just better) desktop scanners give adjacent-pixel values that are identical in 8-bit? Is this a valid test to see if 16-bit image files from a particular scanner are "better" than 8-bit image files?
Preston Earle
____________________________________________________________________________
Date: Tue, 15 Apr 2003 23:21:55 -0500
From: David Riecks
Subject: Re: Headroom and 16-bit; was Resolution for the futureyy
At 11:20 PM 4/15/2003 -0400, Preston Earle wrote:
With any scanner, shouldn't you set the end points in the scanner to
0-255 for the lightest/darkest (or is it darkest/lightest?) points the
scanner sees in the image? Depending on the quality of the scanner,
these points may be different for the same image scanned on different
scanners, but for any one scanner, leaving "headroom" between the
light/dark point and 0/255 doesn't seem to make sense. Am I missing
something?
Preston:
I've always advocated exactly what Andrew has said, and for pretty much the same reasons with the following provisions.
First, the assumption is that you are working towards a "scan once, use many" workflow. If you are scanning for a single use, with a single output in mind, then you should optimize the scan for that particular output. In other words, if I'm scanning a 35mm slide for use at 6 x 9 inches in a magazine for 150 lpi output, then I'm much more likely to choose the optical resolution of the scanner that is closest to that need (for a 4000ppi 35mm scanner, the 2000ppi setting would be the logical choice). I'm also going to leave at least the base "sharpening" of the scanner "on" and possibly go direct to CMYK if I'm satisfied with the scanners look up table for that output.
However, the resulting scan will not be of much value for anything except similar use in the future...so I rarely ever do this.
Having some "head room" on either side of the "goal posts" in either 16 or 8 bpc mode gives you some room for error in your selection of the maximum highlight or shadow value. Many scanner interfaces are not capable of the same "precision" in finding the brightest or darkest pixel. Photoshop allows you to "zoom" in and use several methods to locate the absolute darkest or lightest pixels.
Q: If you happen to choose pixels that are reading 20, 16, 12 (RGB) and set those to 0, 0, 0, what happens to "actual" pixels in the image that are at 3, 4, 4? A: They get "clipped" and you risk losing deep shadow detail.
If you are scanning your images to sell as stock, you need to realize that some of the image processing that clients will do to your file includes such things as Unsharp Masking. As anyone who has read Dan's books already knows, this is needed-- as the halftone screens used for printing break up the image into a series of dots, making the printed image appear less sharp.
The USM filter works by adding contrast near the edges of tonal values in scanned/digital images (an edge occurs when there is a specified difference between two adjacent pixels). The net effect is that this filter creates a "halo" that "pushes" the highlight and shadow values a few points up and down along these found "edges." The size of the halo and the amount that the pixel values are "pushed" up or down is controllable by the client via the Amount, Radius, and Threshold settings in the Unsharp Mask Dialog.
These halos will push the upper values of 247 to the point that they become 250, or 252 (depending on the Amount setting used in the USM filter). The darker edge of a shadow at 8 may become 4 or 5. [you can see an illustration of this effect at http: //www.controlledvocabulary.com/imagedatabases/downsampling.html]
That's why you don't want to scan an image to the point where you are smack up against the 0 to 255 boundaries. If you've allowed for this eventual expansion of the image values, then you are in great shape. However, if you've already got your shadows at 0 and your highlights at 255, where can they go? That's right, no-where except out of range (clipped). In a way it's like creating a "reproduction quality print" (one that is intended to be copied or re-photographed). If you have done this, you know that you have to keep the whites under "paper white" and the blacks a good bit above maximum black. Otherwise your eventual reproduction will have shadows that choke, and highlights that blow out.
Does that help?
David
David Riecks * david@riecks.com
701 W. Washington St * Midwest/Chicago ASMP
Champaign, IL 61820 *
ph/fax 877-646-5375 * http://www.riecks.com/
____________________________________________________________________________
Date: Wed, 16 Apr 2003 07:01:37 -0600
From: Andrew Rodney
Subject: Re: Headroom and 16-bit; was Resolution for the future
on 4/15/03 9:20 PM, Preston Earle wrote:
With any scanner, shouldn't you set the end points in the scanner to
0-255 for the lightest/darkest (or is it darkest/lightest?) points the
scanner sees in the image?
No, that1s the point of the high bit flat scan. You set the end points (which by the way may not be near zero or 255) in Photoshop latter based on the output needs and the image itself. If you set a shadow at 9/9/9/, you can latter set it to 4/4/4 but you can1t if the scanner captured 0/0/0. You've hosed any flexibility in the file from then on.
With the discussion on 8bpc vs. 16bpc scans, I tried a test. I scanned
an image on my Minolta Scan Dual II scanner using Vuescan in 16bpc mode.
This was an old 35mm Kodachrome slide scanned at 2810ppi. When looking
at areas of this scan that should be of uniform color, such as adjacent
pixels in a blue sky, or adjacent pixels in a white-painted building, I
couldn't find two adjacent pixels with exactly the same values.
I wouldn1t expect you would.
This readout (single-point color sampler) was 8-bit (0-255), so the
16-bit values were rounded (or truncated or whatever).
First, you1d want a 3x3 sample at the very least, especially on a file that size. As for truncated, well that1s up to your definition of the term. Photoshop shows 255 steps in high bit scans for a number of very logical reasons. For one, each line in the 8 bit histogram is one pixel. If you had a true 16 bit Histogram which showed the same data, (65,000 odd values), you1d need a display the size of a small wall to view the Histogram.
Even so, I couldn't find *any* same-color adjacent pixels. Considering this, do the
high-bit values have any real value? If adjacent-pixel values are
different in 8-bit precision, does it make any sense to save them with
16-bit precision? Do other, more expensive (or just better) desktop
scanners give adjacent-pixel values that are identical in 8-bit? Is this
a valid test to see if 16-bit image files from a particular scanner are
"better" than 8-bit image files?
When you pull these corrections on 8 bit scans, you only have 255 values to move around. The same scan in high bit (16 true buts which I doubt your Minolta provides but that1s immaterial), you have something like 64,000 values from white to black. So after you convert the data from high bit to 8 bit, you don1t lose any necessary data since you had far more than you needed in the first place.
Andrew Rodney
____________________________________________________________________________
Date: Wed, 16 Apr 2003 09:47:50 EDT
From: Dan Margulis
Subject: Re: Headroom and 16-bit; was Resolution for the future
Preston writes,
With any scanner, shouldn't you set the end points in the scanner to
0-255 for the lightest/darkest (or is it darkest/lightest?) points the
scanner sees in the image? Depending on the quality of the scanner,
these points may be different for the same image scanned on different
scanners, but for any one scanner, leaving "headroom" between the
light/dark point and 0/255 doesn't seem to make sense.
This is the old question about how much color correction, if any, one should try to do during the scan as opposed to in Photoshop later. The extreme views would be to either try to do *everything* on the scan or else to scan every piece of art to exactly the same settings and rely on a profile. Most people now modify these extremes. I myself examine the art quickly before scanning and if there is an obvious correction to be made I'll make moves in that direction, but not try to achieve perfection. People who aren't inclined to do this often scan to certain endpoints as opposed to leaving hopelessly flat images hopelessly flat.
If you're scanning to endpoints without taking any account of the character of the image, because you plan to fix whatever's wrong in Photoshop, then it makes sense to be slightly flat. That way, if there's critical detail in the highlights and/or shadows, it's easier to attack them, because those parts of the curve can be made steeper more efficiently than if the endpoints were set to normal values.
With the discussion on 8bpc vs. 16bpc scans, I tried a test.
And have come to the same conclusion as everyone else who has tested it. For this kind of semi-controlled workflow there's no gain in doing this, it just doubles file size.
Dan Margulis
____________________________________________________________________________
Date: Wed, 16 Apr 2003 10:03:18 -0400
From: Loring Palleske
Subject: Re: Headroom and 16-bit; was Resolution for the future
doubles? or quadruples?
On Wednesday, April 16, 2003, at 09:47 AM, Dan Margulis wrote:
And have come to the same conclusion as everyone else who has tested
it. For this kind of semi-controlled workflow there's no gain in doing this,
it just doubles file size.
Regards,
Loring Palleske
Creative Imaging
1.877.279.2441
905.666.6647
____________________________________________________________________________
Date: Wed, 16 Apr 2003 10:14:01 EDT
From: Dan Margulis
Subject: Re: Headroom and 16-bit; was Resolution for the future
Loring writes,
doubles? or quadruples?
Doubles. Doubling *resolution* quadruples file size; doubling bit depth merely doubles it.
Dan Marguli
____________________________________________________________________________
Date: Wed, 16 Apr 2003 10:31:08 -0400
From: Lee Clawson
Subject: Re: Re: Resolution for the future "Flat"
I wish I had a more descriptive word. Keep in mind the context of the discussion is based on unknown output.
The scan will look (and is) flat (contrast) compared to a scan made for a specific output. The captured tone (density) range is as wide as possible. The thing that you want to avoid is clipping (compressing) the tonal range. I like David and Andrew's explanation of all this.
One thing I'll add is a point about USM. On the drum scanner (Crosfield) while we kept USM "on" the set up is flexible. You can specify when and how USM starts and stops. While I never understood how to do this we had a set up with a gentle ramp up into the quarter tones. This kept skintone highlights soft. USM reached max effect after this. We'd input density numbers to describe this "ramp" and the scanner software smoothed the ramp.
Lee
____________________________________________________________________________
Date: Wed, 16 Apr 2003 09:05:41 -0600
From: Andrew Rodney
Subject: Re: Headroom and 16-bit; was Resolution for the future
on 4/16/03 7:47 AM, Dan Margulis wrote:
This is the old question about how much color correction, if any, one should
try to do during the scan as opposed to in Photoshop later. The extreme views
would be to either try to do *everything* on the scan or else to scan every
piece of art to exactly the same settings and rely on a profile.
Maybe with ancient scanners. I don1t have to lock anything down when scanning with my Imacon using a profile because the software is smart enough (like Photoshop) to allow a profile workflow AND use any control in the software I wish.
Andrew Rodney
____________________________________________________________________________
Date: Wed, 16 Apr 2003 11:29:38 -0400
From: Jim Rich
Subject: Re: Re: Resolution for the future "Flat"
One thing I'll add is a point about USM. On the drum scanner (Crosfield)
while we kept USM "on" the set up is flexible. You can specify when and how
USM starts and stops. While I never understood how to do this we had a set
up with a gentle ramp up into the quarter tones. This kept skintone
highlights soft. USM reached max effect after this. We'd input density
numbers to describe this "ramp" and the scanner software smoothed the ramp.
Lee,
That Crosfield/Fuji feature (called Cal 508) is very cool. I wish Photoshop had a filter with that level of sophistication. You can do that kind of thing in Photoshop but... you have to work hard at it by creating a gradient mask that you would load as a Selection and then apply USM through the mask.
Jim Rich
____________________________________________________________________________
Date: Wed, 16 Apr 2003 12:35:18 -0400
From: "Preston Earle"
Subject: Re: Headroom and 16-bit; was Resolution for the future
"Andrew Rodney" wrote:
Maybe with ancient scanners.
Or with modern ones <G>. The point I'm trying to make is that there *may* be some reason to scan high-bit files with high-end, professional scanners, but the current crop of desktop CCD scanners are different. If a scanner can't report adjacent-pixel values with 8-bit precision, why should *any* image need the 9-16-bit noise?
Is the only benefit of high-bit data to reduce banding/posterization in highly manipulated images, or are there other benefits?
With the proliferation of desktop CCD among quality-conscious amateur photographers, they hear that 16bpc is "better" and save all several thousand of their lifetime collection of images in files twice as large as they need to be. I'm not contending that "professionals" with "professional" scanners (PMT or high-end CCD models like the Eversmart) shouldn't scan in 16bpc, whether it presents any advantage or not. I *am* trying to determine if the same theories that apply to high-end scanners also apply to desktop CCD models.
Preston Earle
____________________________________________________________________________
Date: Wed, 16 Apr 2003 11:40:30 -0500
From: Nick Wheeler
Subject: Re: Re: Resolution for the future "Flat"
Jim Rich wrote:
That Crosfield/Fuji feature (called Cal 508) is very cool. I wish Photoshop
had a filter with that level of sophistication. You can do that kind of
thing in Photoshop but... you have to work hard at it by creating a
gradient mask that you would load as a Selection and then apply USM through
the mask.
Jim:
Actually using techniques described by Dan as well as others it is not too much trouble to build up a set of Photoshop sharpening actions which either mimic or exceed the capabilities of all the older scanner apps. About a half dozen saved actions would solve 90% of the sharpening problems that come across the desk every day. Dan did a really good job of outlining the basic problem when he described thinking about images as High Frequency or Low frequency etc.
I think the best approach is to pay attention to the steps used when doing sharpening and soon the user will notice a similarity between the steps taken for a given image. Save it as an action which can then be refined over time or made interactive by inserting dialog enabling.
One can then build shadow, highlight, edge or luminosity masks - save them as actions with different associated curve and level adjustments and blur radii and overlay modes and pretty much automate the whole process. The beauty of photoshop is you can take sharpening to a level never possible with scanning applications, and for really difficult situations the whole process is interactive.
Yes the USM filter by itself is fairly primitive, but when one thinks of all the tools available in Photoshop today the possibilities for really sophisticated sharpening are infinite.
Dan could write an entire book on just sharpening! Look what he's showed us is possible with just the curves dialog.
Nick Wheeler
____________________________________________________________________________
Date: 16 Apr 2003 17:49:32 -0000
From: Andrew Rodney
Subject: Resolution for the future
on 4/16/03 10:35 AM, Preston Earle wrote:
The point I'm trying to make is that there
*may* be some reason to scan high-bit files with high-end, professional
scanners, but the current crop of desktop CCD scanners are different. If
a scanner can't report adjacent-pixel values with 8-bit precision, why
should *any* image need the 9-16-bit noise?
Obviously having more bits from an inferior scanner isn1t going to do anyone any good. More bits is only helpful if it1s good data.
The original question and to do with archiving files for future use. Assuming a decent scanner, the need for future printing needs, I wouldn1t think of scanning in only 8 bits pre color, especially when one might want to scan flat and adjust latter.
Some have suggested that someday, having more than 8 bits pre color might be an advantage when it comes time to output a file. That time is today! Version 5.5 of ImagePrint from ColorByte will actually use high bit image data and produce superior output when feed a high bit file. With 8000 screens, it1s able to use the additional data. Who knows what printers in the future might need high bit data for superior output.
Is the only benefit of high-bit data to reduce banding/posterization in
highly manipulated images, or are there other benefits?
See above with regard to ImagePrint. That1s not the case anymore.
With the proliferation of desktop CCD among quality-conscious amateur
photographers, they hear that 16bpc is "better" and save all several
thousand of their lifetime collection of images in files twice as large
as they need to be.
With 4gig1s of storage at a few bucks per DVD (and who knows what1s coming at a lower price), storage is the last thing I1d be worrying about if the idea is to scan a piece of film and never need that film again.
I'm not contending that "professionals" with
"professional" scanners (PMT or high-end CCD models like the Eversmart)
shouldn't scan in 16bpc, whether it presents any advantage or not. I
*am* trying to determine if the same theories that apply to high-end
scanners also apply to desktop CCD models.
A lot has to do with the quality of the scanner and the software. IF you can only get 24 bits out the back end and the software isn1t all that good, then that1s a problem. You can use the same scanner and software but output high bit files and offload the corrections and so forth in something far superior; Photoshop. So having high bit data provides more options. But if you have a scanner that has really good software and you1re happy with producing the edits there, then with the exception of having more options (or having an output device that can use the high bit data), there1s little need. But why paint yourself into a corner? The only downside to high bit files is they are twice the size of their 8 bit cousins. Seems like a small price to pay for the flexibility.
Andrew Rodney
____________________________________________________________________________
Date: Wed, 16 Apr 2003 12:13:07 -0600
From: Jim Rich
Subject: Re: [colortheory] Headroom and 16-bit; was Resolution for the future
On 4/16/03 12:35 PM, "Preston Earle" wrote:
Or with modern ones <G>. The point I'm trying to make is that there
*may* be some reason to scan high-bit files with high-end, professional
scanners, but the current crop of desktop CCD scanners are different. If
a scanner can't report adjacent-pixel values with 8-bit precision, why
should *any* image need the 9-16-bit noise?
Is the only benefit of high-bit data to reduce banding/posterization in
highly manipulated images, or are there other benefits?
Not from what I understand.
Another reason or two I am told is that high-bit images are supposed to make you feel better about protecting your image data because there is so much information in the file and there is supposed to be more flexibility when doing image processing. There is more information in the file, but as for having more flexibility with image processing...??
While there is some theoretical evidence to suggest that high-bit images provide more data and flexibility (by inspecting Histograms), the hard evidence indicates you cant see the difference between 8 and 16 bit images that have dramatic color edits and that are printed and compared side by side. And if you do compare 8 and 16 bit images side by side, you are only guessing which image is 8 and 16 bits 40% to 60% of the time. And that is if you are using experts to judge the images. If you use novices or inexperienced judges then they don1t have a clue.
However, in the case of how this thread started, high-bit scanning is a good thing. Some feel more comfortable with this as a back up strategy and I agree with it.
As for banding, in todays market, higher quality capture devices and refined image processing in programs like Photoshop allows you to capture images using high-bit strategies, such as using 16 bits in the scanner and then sampling it down to the best 8 bits. When the image is delivered to the desktop for a working session, it has been proven by many people that 8 bit data is just fine so there is no need to worry of degrading an image.
Some 16 bit advocates will argue a theoretical point but they don1t offer any evidence that the benefits of using high-bit images is real. Dan made that real clear here last year. There has been lots of testing and ultimately hard facts that have proven that if you capture an image correctly you can work safely and flexibly in the Photoshop 8 bit mode.
If you have banding from your scanner or camera, the technology most likely is not integrated well and you should consider it broken. One solution is to fix it. Or you can continue to use it and in that case you would want to work with high-bit images.
As for the 16 vs 8 bit argument and working in Photoshop, I thought that one was put to bed. That is unless there is some new hard evidence that shows you can actually see real (and not theoretical) differences in the final output between 16 and 8 bit images. And if there is, I am willing to reconsider my position.
I am sure some will get offended by my comments, but hey, you or I are not bad people for choosing to work in 8 or 16 bit Photoshop modes. It is a matter of choice. The 16 bit mode is limited and the 8 bit mode offers a full set of features. And it has been proven that both modes will work equally as well to create the same looking output even with dramatic color edits. Based on the evidence I have seen, it is really just a logical choice of which mode you choose.
Jim Rich
____________________________________________________________________________
Date: Sat, 19 Apr 2003 23:55:27 -0500
From: "Howard Smith"
Subject: Re: Resolution for the future
From: Dan Margulis
In any situation where the same file has to be used to generate two different
high-quality outputs, automated color management isn't going to be a complete
answer but it usually gets one close. But the real killer is always the
sharpening.
What is the advantage to sharpening at the time of scanning? I seem to recall that you recommended waiting to do it in Photoshop. For that matter, is there any real advantage to do anything at the time of scanning if you're doing the work yourself from start to finish? My own experience has been that service bureaus can do some pretty awful things with a simple scan even without using any of the adjustment tools.
Or could it be that the scans are awful because they don't know how to use them? For example some of the scans done by others have had everything squashed to the left or the right of the histogram, making it necessary for me to do quite a bit of data stretching (no, that's not intended to be technical terminology).
But then maybe that's the answer. Maybe the scanner does all this correcting with no data loss. So far not one of my reference books has explained why adjustments at the time of scanning are beneficial, if they are at all. Makes me wonder if more time spent with my Nikon film scanner would save some headaches in subsequent color correcting or if it would just be a duplication of effort?
Any light you could shed on this will be very much appreciated.
____________________________________________________________________________
Date: Sun, 20 Apr 2003 10:34:33 -0400
From: Denise Romeo
Subject: Re: Resolution for the future
On 4/13/03 10:55 AM, "Dan Margulis" wrote:
Furthermore, with a drum scanner, such extreme magnifications produce a
quality hit if you later need to downsize them drastically. Now granted,
fashion work can probably accept the overly soft rendition caused by
excessive resolution better than almost any other type of photography. But
still, one has to suppose that the *likeliest* use for each photo is going to
require <20 mb of data. In that case, coming down from 600 mb is going to
give distinctly inferior quality as opposed to scanning at approximately the
right resolution in the first place. The 200 mb scan you propose might be
acceptable but I wouldn't be that happy about it.
Hello Dan,
Thank you so much for your insight. The photographer is a celebrity photographer and everyone he shoots is already famous. His compositions are more like Annie Leibovitz, so there is all this composition work that has been extensively done, so nothing is going to be cropped. I was involved in the rearchiving of Annies negs over the last few years as I worked on the prints for her American Express exhibition. Everything had to be rescanned and redone for that show due to poor scanning by god knows who over the years. This current archive is going to last well into my retirement....so his children can hire someone to get it off its current media which has yet to be determined. I am just trying to come up with a valid argument so that I can talk him into something more workable than a 600 meg scan. Somehow he got this number in his head... And I would guess that most of these images will be used for book publication...so 600 is too big. As you state, we will take a qualitiy hit by downsizing, which is what I thought, and want to avoid. I would rather rescan for an exhibition... But at 200-300 megs each I don1t think I would have to.
you say..
In that case, coming down from 600 mb is going to
give distinctly inferior quality as opposed to scanning at approximately the
right resolution in the first place. The 200 mb scan you propose might be
acceptable but I wouldn't be that happy about it.
why not happy with 200?
Denise Romeo
____________________________________________________________________________
Date: Mon, 21 Apr 2003 11:42:49 EDT
From: Dan Margulis
Subject: Re: Resolution for the future
Denise writes,
why not happy with 200?
I don't like scanning to a *much* larger size than the probable use. It's inefficient and may actually harm quality. I'd still prefer to scan at a more likely resolution--maybe 100 mb scans--with ultra-high resolution second scans on certain pieces.
However, what you're suggesting is reasonable. What your client was suggesting--600 mb for everything--was not.
Dan Margulis
P.S. There's another factor lurking here, one that shows its face frequently in many different contexts. That is, how much flexibility do we want to give some hypothetical future user who's going to pick up our files in some unknown context? In your case, you're relying on this hypothetical person to be good enough with USM to figure out how to downsize a high-quality file to 1/20th of its original size. Do you really trust that person to do it, or should you do some of the work yourself, understanding that you don't have all the information you'd like?
Similarly: if the likely use of the file is CMYK, do you give an RGB file, or do your own conversion? Do you embed a profile? Do you work in a wide, or narrow-gamut RGB? Do you apply USM first, or do you rely on the next person? Are you going to set highlight and shadow yourself, or leave it alone?
Loosely speaking these are all choices of whether to trust the next user, who may know more than you do about final conditions but less than you--sometimes a lot less--about how to handle an image. Also loosely speaking, the politically-correct side often comes down in favor of trusting the next user, whereas people with lots of production experience like myself tend to favor taking the bull by the horns immediately.
But, it's a hard choice and much depends on individual circumstances. If you're absolutely positive that the next user is me or somebody who's taken one of my classes, then the best approach is not to do anything at all--just give a raw, untouched scan. That's totally extremist and nobody in their right mind would do it.
OTOH, if you're producing screen grabs of Photoshop dialog boxes that will appear in print, no matter how committed you are to calibration and color management and trusting the next person, you give that file to the next person in CMYK, with no profile embedded, and with lines in black only. Any other approach is totally extremist and nobody in their right mind would do it.
____________________________________________________________________________
Date: Thu, 24 Apr 2003 17:54:35 -0300
From: Cícero Rodrigues"