unprocessed scanning

[ian_dunross]

<p>I plan to do all my post processing work at a later date with Lr or Ps. Right now, I just want to gather the all the available data from my scanned photos.</p>

<p>I am using an Epson V700 to scan 20-30 year old photos.</p>

<p>In this context, is there anything I am missed out by using Epson Scan instead of SilverFast AI or VueScan? In other words, do they offer anything other than just post processing? Do they so anything "special" that I won't be able to do later in Photoshop?</p>

[Wouter Willemse]

<p>I do not know Epson Scan, but one thing that I like in Vuescan is the ability to save raw files (DNG format), which helps avoiding making changes to the captured data (which would happen with TIFF/JPEG - tone curves, some white balancing). Nothing you cannot do in PS or LR after, or that either of those programs would do better than the scanning software - more a matter of keeping as much data for as long as possible.<br>

Either way, make sure to not save directly to JPEG, or 8-bits TIFFs. Your scanner can capture more, so accept the large filesize of 16-bits files (TIFF or raw).</p>

[ian_dunross]

<p>Wouter, thanks for the response.</p>

<p>I thought tiff is already a raw format file, same as DNG, PSD or NEF?</p>

[Wouter Willemse]

<p>No, most raw formats are based on TIFF, as is PSD (which is also not a raw format, but Photoshop's file format), but they're not the same thing. TIFF (and PSD) are "processed", in the same way as a JPEG is. However, TIFF (and PSD) is a much more flexible file format, allowing for different colour channels, 16-bits colour, layers and optionally compression (so also no compression at all is an option). This makes it a much better choice for files that need editing. But raw files are even less "cooked" still.</p>

[ellis_vener_photography]

I do not scan much anymore but by strategy was to do a good enough job that I would not have to scan the film again,

unless I had to have a much better drum scan made (see http://www.eigerstudios.com or http://www.nancyscans.com)

 

To fulfill that mission, my tactics were:

 

Scan at 16 bits per channel. I like the EktaSpace5 color space for scanning, details are at

http://www.josephholmes.com/profiles.html

 

Use the maximum optical (not interpolated ) resolution of the V700, 2400dpi.

 

I like SilverFast but because it was built for pre-press houses the interface can be daunting. Using Silverfast's profiling

targets for transparencies and prints, and SilverFast's built in profiling software does help get color right.

 

The only thing I do in any scanning software is to bring in the clipping point at the right end (highlightsin a print, slide or

transparency; film base in a negative) to about 5 points above the end of the histogram. This allows the dynamic range of

the image to efficiently fill the available bit depth. The higher registers of bit depth can contain more tonal and color data

than the lower registers.

This last link is my review of the V750M from January 2007: http://www.ppmag.com/reviews/200701_epsonperfv750m.pdf

[digitaldog]

<p>Scanners do not create raw data, not as we understand raw data from a single sensor capture device. This is what raw data looks like:<br>

http://www.digitaldog.net/files/raw.jpg<br>

A scanner can create an unfinished, not properly or fully rendered appearance. I don't know what the point is if you're the scanning operator, you have good scanning software (super important) and you just want to end up with a TIFF or similar document that has the global tone and color you desire. You either do it at the scan stage (fast) or later in Photoshop or some other app. IF the scanning software is poor, there's something to be said for scanning high bit, wide gamut and getting at least as close as you can with one default scan setting, then '<em>correct</em>' (that's what you're doing) later with another app to get the color and tone as you wish. You can pay now or later, what makes more sense in the workflow based on the tools and your time? </p>

<p>You can save a fully rendered RGB document as a DNG. <strong>That does not make that data raw</strong>, it isn't any more than saving a TIFF as a PSD or vise versa changes the basic image structure. There is really no reason to save a fully rendered image as DNG instead of TIFF and neither are true raw files as you see above in the illustration. A Viewscan DNG is not raw data! It is a 'raw' uncorrected scan and that's true if it's a DNG a PSD or a TIFF. The editing capabilities are not anything like the way we deal with camera raw files. </p>

<p>You have to '<em>cook</em>' the pixels from a scanner. Do it at the scan stage or later, the data isn't raw and the end results could be identical but the main factors are the control you have over rendering the image as you wish it to appear. </p>

[joe_c5]

<blockquote>

<p>This is what raw data looks like:</p>

</blockquote>

<p>That is what raw data <em>from a typical digital camera </em>looks like. Data from a scanner probably does not use a Bayer filter format, but it might have some other arrangement of color filters. For a flatbed scanner there would probably also be some duplicated data as the scanner head moves (the Epson V700 seems to use a 6 line CCD).</p>

<p>Since there would probably be significantly more data than TIFF format, I doubt there is much demand for scanning to an actual raw format so I would not expect many scanners, if any, to support it.</p>

[mendel_leisk]

<p>I've always saved Vuescan Raw File first, in tiff format, red/green/blue, 16 bit. If I'm using infrared scanner I apply the cleaning to the raw file, don't save it as a separate channel.</p>

<p>I'm not sure that there's an advantage to DNG, out-of-my-depth on that. I do know that in earlier releases TIF was the <em>only</em> option.</p>

<p>The most direct and simple method to use that Vuescan Raw File is again via Vuescan: you direct Vuescan to read the file instead of doing an actual scan, and produce whatever finished image file you want. It can be a relatively flat, finished gamma file, for post processing with another program. Or the finished product.</p>

[digitaldog]

<blockquote>

<p>That is what raw data <em>from a typical digital camera </em>looks like.</p>

</blockquote>

<p>Yes, exactly why I posted it!</p>

<blockquote>

<p>Data from a scanner probably does not use a Bayer filter format.<br /></p>

</blockquote>

<p>It absolutely doesn't (unless you can find a single CCD scanner), otherwise it's true color coming from a trilinear CCD/sensor. Nothing like raw data in a bayer format! </p>

<blockquote>

<p>I'm not sure that there's an advantage to DNG, out-of-my-depth on that.<br /></p>

</blockquote>

<p>DNG like TIFF it's cousin is just a container. You can put scanned data in there. That data isn't raw, it's cooked. You can of course embed raw sensor data from a camera in a DNG. You can embed a JPEG in a DNG. And lots of metadata. The only thing that makes DNG a Digital Neg is when you put the actual digital neg into the container. That's the case when you convert a proprietary camera raw to DNG and not the case when you embed a scanner created data into that container. </p>

<blockquote>

<p>It can be a relatively flat, finished gamma file, for post processing with another program. Or the finished product.<br /></p>

</blockquote>

<p>Which is no different than saving that data in a TIFF and futzing with it in Photoshop. That TIFF can be finished or need more processing. Nether state is raw like that from a DSLR as an example. And gamma has nothing to do with it. When I built scanner profiles for my Imacon, testing showed that a TRC Gamma of 3.0 produced the best data. So that's how I set the scanner and profile, in a color managed app, it looks just fine. As treated untagged, it would look awful even though it's not (just like 1.0 gamma images can look fine or very dark depending on if you treat the data correctly or not). </p>

[ant_nio_marques]

<p>PLEASE, HEADS UP, unless Joe made a mistake he's talking about scanning PRINTS, and the talk of what is or isn't 'raw' is neither here nor there.</p>

<p>As far as I know, all that VS or SF offer (besides obviously being different programs) is the possibility to save infrared data to a separate channel, instead of 'flattening' the image with dust and scratch correction (correction can thus be experimented with later on). But that only matters for FILM scanning. You aren't supposed to be scanning prints (or B/W film that is not C41) with infrared. So I'd say that you are not losing anything in terms of the actual files produced.</p>

<p>As to the format of the files produced, you are advised to use 16-bit TIFF with a large gamut colour space (ProPhoto?). I'm not sure if sRGB isn't enough, given we're talking about prints. Also, probably 300 dpi is enough, it's unlikely that your prints have more than that in them (but possible, I suppose, it can't hurt to go to 600 - if you use VS, try setting 2400 or whatever the second-highest value available is, and set 'reduction' to 4, and see if you like it better).</p>

[digitaldog]

<blockquote>

<p>As to the format of the files produced, you are advised to use 16-bit TIFF with a large gamut colour space (ProPhoto?).</p>

</blockquote>

<p>The input scanner profile should tag that data. It's the biggest gamut the device can produce and it's based on measurements of said scanner and behavior (that's key) of data used to build the profile. No reason to convert to ProPhoto, the fewer conversions the better. You don't necessarily get R=G=B behavior which is useful unless you set that correction at the scan stage. </p>

[joe_c5]

<p>The Epson V700 seems to be a flatbed scanner, so yes, I assumed the OP was scanning prints.</p>

<p>Some form of raw data does exist inside the scanner, just few if any scanners bother to save it.</p>

[JDMvW]

<p>Whatever software you use, just set it as neutral as you can, and you will be able to play with it later on, especially if you save large enough files up to the limits of your scanner, even large jpegs.</p>

<p>Actual RAW, as beaten into the ground above, out of a camera is better, but for copies of either prints or slides, the film and processing of it has already made that possibility remote to impossible (analogy: blood from turnip).</p>

<h1>"Scan large, scan once,"</h1>

<p>is my empirically-derived motto when it comes to scanning. It's my corollary to the carpenter's "measure twice, saw once"</p>

[q.g._de_bakker]

I have never done so (don't know why i would want to), and it doesn't help when using an Epson V700 (a rather capable film scanner as well as a flatbed print scanner. Not that it makes any difference), but on the topic of raw files and scanners: Nikon Scan can safe files in NEF format. Useless outside Nikon software. But as raw (i.e. unspoiled/unspoilt by any adjustment settings used in the software) as you can get.

[digitaldog]

<blockquote>

<p>But as raw (i.e. unspoiled by any adjustment settings used in the software) as you can get.</p>

</blockquote>

<p>Why would anyone want an image, unspoiled by adjustments unless it produced the best quality image? <br>

How is an ugly 16-bit image a better start than a better looking 16-bit image? </p>

[ian_dunross]

<p>Thanks for all your responses! I haven't had time to read every one of them yet, but will this afternoon.</p>

<p>I have one question though: What's a good resolution to scan at? Using tiff and 48bit color depth, some of the file sizes can be a bit overwhelming.</p>

<p>800dpi - 372mb<br>

1200dpi - 828 mb<br>

2400dpi - 3.21gb</p>

[q.g._de_bakker]

<i>"Why would anyone want an image, unspoiled by adjustments unless it produced the best quality image?"</i><br><br>To have a chance later to spoil it again by applying other adjustments, without being hindered by having the scope for adjusment limited by earlier attempts to adjust the image?

[q.g._de_bakker]

<i>"What's a good resolution to scan at? Using tiff and 48bit color depth, some of the file sizes can be a bit overwhelming."</i><br><br>What JDM said: <b>"Scan large, scan once"</b><br>Files may be large, but hard drives are larger.

[digitaldog]

<blockquote>

<p>To have a chance later to spoil it again by applying other adjustments, without being hindered by having the scope for adjusment limited by earlier attempts to adjust the image?</p>

</blockquote>

<p>How are they limtied? In each case you have baked RGB pixels you're just altering values at different times. This isn't parametric editing ala ACR or Lightroom on unrendered (raw) data. </p>

[ant_nio_marques]

<p>Andrew, can you give John practical tips on how to create a scanner reflective profile?</p>

<p>John, in an earlier comment I had some musings on DPI options. (Short version: I think 600 is already higher than the resolution of your photos.)</p>

[digitaldog]

<p>To make a custom profile one needs an it8 target (film or print) and some software that can build a scanner profile. Kind of expensive. <br>

Scan at the highest optical resolution, high bit, widest gamut which again would be a good input profile.</p>

[ant_nio_marques]

<p>Come on, do you all really suggest scanning a print at 2400dpi? I'd be interested to know why. What is the maximum dpi we can expect of a print, and what is the average? Or lpcm, if you like.</p>

<p>16-bit is essential, yes. (Also called 48 because it's 3 channels.)</p>

<p>As to the colour space, Andrew is right, but I doubt you need to lose a lot of sleep over it.</p>

<p>(Or 4800 dpi, if we are to trust Epson: it claims native 4800 dpi for platen scanning and 6400 with the lens that is engaged when using the holders, my tests suggest between a third and a half of that.)</p>

[digitaldog]

<blockquote>

<p>I'd be interested to know why.</p>

</blockquote>

<p>If the idea is scan once, use many and the use isn't defined, get all the data you can capture. <br>

If you <strong>know</strong> you'll <strong>only</strong> output an 8x10 print, scan accordingly.</p>

[q.g._de_bakker]

<i>"How are they limtied? In each case you have baked RGB pixels you're just altering values at different times."</i><br><br>What cases would that be?<br>Well, what Nikon's scanner NEF offers is indeed parametric editing ala ACR and Lightroom. It does store the image scanned with whatever setting you have set at the time. And then it allows to return to that by discarding anything you may have done to it afterwards (in a program capable of dealing with Nikon's NEF). So you get the chance to spoil the image again later by applying other adjustments without being hindered [etc.].<br>You are presumed to know know how changing values limits the scope for further changes, including the possibility to change back to the original values. So no need to explain.

[digitaldog]

<blockquote>

<p>Well, what Nikon's scanner NEF offers is indeed parametric editing ala ACR and Lightroom</p>

</blockquote>

<p>The NEF from the scanner is still baked. It's not raw. LR and ACR can produce parametric edits with raw <strong>or</strong> rendered data but there's a big difference in the original and subsequent data. A JPEG and a raw in LR are treated the same in terms of the instructions but the source data sure isn't and that's why there are '<em>edits</em>' to produce resulting pixels that can't be done on existing pixels. White balance of JPEG vs. raw as one such example. </p>