Scanner Practice

[Cesare Berti]

A general question on scanner best practice.

When scanning 5x7 B&W negatives (with Epson scan) on my Epson 4870 I find myself doing very little manipulation in the scanning process other than looking at the histogram for clipping at the two extremes and then processing the image in photoshop.

I recently took delivery of and scanned my first two 120 negatives with the Nikon 9000 scanner using the Nikon Scan software. The software allows for fairly extensive manipulation of the image to be scanned, setting the back/white points, saturation, hue etc. It was while reading the manual that I questioned whether I should be using more of the scanner's capability to manipulate the image and process less in photoshop.

So to my question, is it better practice (for the final print) to manipulate the image during the scanning process or to keep it simple and scan to get maximum detail from the negative and then process the image in photoshop?. For example, using unsharp mask, is it your practice to do any sharpening during scanning or to do it later in photoshop?

May I just add that the digital world continues to be a challenge and I'm glad I can retreat to my darkroom when things get too frustrating.

[Peter De Smidt]

It depends quite a bit on the software in question, and whether the adjustments actually do something that can't be done in Photoshop. For example, if the scanning software can change the intensity or time of exposure of the lights used for scanning, than that adjustment cannot be made later in Photoshop. On the other hand, if the adjustment can be made in Photoshop, then there are advantages to doing so. For example, if you use unsharp masking in the scanning stage and you later want to change the settings,then you'll have to re-scan, or reconvert a Vuescan RAW file. On the other hand, if you sharpen non-destructively in Photoshop, then you can always change the settings. Unfortunately, it's often unclear what settings actually effect how the scan itself is done, i.e. what settings actually change what information is captured in the scan.

One workflow choice for color transparencies is to make an icc profile using a test target from Wolf Faust, or some other source. This will quickly allow you to produce a very accurate scan, which will allow you to interpret the image in a number of ways. On the other hand, if you do you artistic interpretation/manipulation at the scanning stage, then you'll have to re-scan to make another interpretation, or at the least reprocess a Vuescan "RAW" file.

All this said, you can do very good work either way. Try some various workflows and see what works best for you.

[Ken Lee]

Unsharp masking should be the last thing you do, just before printing.

I have found that with every digital "adjustment", there is an incremental departure from the smooth and natural "analog" quality of the original. If you've viewed a good large format color slide on a light box, perhaps you know what I mean.

So with regard to colors and tonality, I try to get the scan as close as possible to the finished product, and perform as little subsequent Photoshop adjustment as possible. This is analogous to getting the exposure and development of the original as close to perfect as possible: since it's a down-hill process, we had best start at the top.

This attitude will probably not rocket me to overnight celebrity on the Photoshop lecture circuit - but it makes sense, and seems to work in practice.

[Oren Grad]

I generate different products from a scan - for example, jpgs intended for different viewing contexts, or prints on different papers. Each of these outputs requires a distinctive post-processing recipe, so it doesn't make sense to "bake in" to the scan any more than is needed to give me a file that will be robust to many different possible manipulations.

So my primary objective in scanning is to produce a full-information file, which means no clipping at either end of the tonal scale and no disproportionate compression of any part of the tonal scale that has important information. For color, I'll also try to arrive at a mask correction that puts the overall color balance in the ballpark. I generally leave everything else for Photoshop.

[Ken Lee]

That makes good sense !

[Peter De Smidt]

It used to be the rule of thumb that sharpening should only be done right before printing, but this is no longer true. Bruce Frasier and others have shown that a workflow that involves capture sharpening, local sharpening, and output sharpening is more flexible than one stop sharpening at the end.

"So with regard to colors and tonality, I try to get the scan as close as possible to the finished product, and perform as little subsequent Photoshop adjustment as possible. This is analogous to getting the exposure and development of the original as close to perfect as possible: since it's a down-hill process, we had best start at the top."

Possibly. Setting exposure and development with film processing is analogous to "hardware adjustments" in the scanning software, since they effect the information that is captured. Adjustments in traditional printing would be analogous to "software adjustments."

One question, though, is whether there is a hardware adjustment provided by the scanning software that is analogous to exposure and film development. With drum scanners I believe there is, but it's very unclear whether there is with consumer flatbeds and film scanners. That is, it's an open question whether working with a curve in the scanner software is pertinently different than working with one in Photoshop. If it isn't, then working with a curve in the scanning software is just as much a down-hill process, to use your terminology, as doing the curve in Photoshop, with the disadvantage that working with a curve in the scanner software is destructive to the file data, whereas one can do this non-destructively in Photoshop.

[Cesare Berti]

Peter, your analogy is interesting. I do not know if the the Nikon Scan software can emulate adjustments in exposure and development but I will try to find out.

Oren, producing a full information file from a scan makes a lot of sense to me especially if one's ultimate goal is a "fine print".

You mention that for color work you scan and adjust in the scanning process so the overall color balance is in the ballpark. I just finished scanning and printing a old color negative so I could compare the resulting print to the Cibachrome print I made from the same negative way back when. The first thing I noticed is my scanned print is sharper than my size enlarged Ciba print, poor enlarging technique on my part though I've always liked this particular print.

The scanned print is also cooler (bluer) than the Ciba print. This is not much of a surprise when scanning color negatives. I take it that your practice would be to color correct at the scanning stage?

[Ken Lee]

"...working with a curve in the scanning software is just as much a down-hill process, to use your terminology, as doing the curve in Photoshop, with the disadvantage that working with a curve in the scanner software is destructive to the file data"

We agree, and this is really all that I am suggesting: the most important step is the first one.

[Peter De Smidt]

Cesare, I believe analogue gain, using the long exposure option, digital ice, and multiple pass are "hardware" adjustments. Be careful with multiple pass since while noise might be decrease, so might sharpness. If the negative moves at all during the passes, due to being heated up by the light for example, the combined scan will be softer.

What I'd really like in scanning software would be the ability to easily use analogue gain and exposure times to optimize the image capture. A good "live" preview or histogram, or the ability to switch been previews taken with different settings can't be that hard to make.

[Oren Grad]

You mention that for color work you scan and adjust in the scanning process so the overall color balance is in the ballpark. I just finished scanning and printing a old color negative so I could compare the resulting print to the Cibachrome print I made from the same negative way back when. The first thing I noticed is my scanned print is sharper than my size enlarged Ciba print, poor enlarging technique on my part though I've always liked this particular print.

The scanned print is also cooler (bluer) than the Ciba print. This is not much of a surprise when scanning color negatives. I take it that your practice would be to color correct at the scanning stage?

I'm not sure I entirely follow. Cibachrome is a direct positive medium intended for prints from transparencies. It's usually not worth the trouble to try to print negatives on it. Anyway, there's no reason a scanned negative, or print, should necessarily have a blue bias.

I do approximate color correction at the scanning stage. SilverFast Ai helps quite a bit in this - the NegaFix utility has a bunch of film-specific presets that in my experience tend to work pretty well for Kodak color negative films, though less so for Fuji. It also provides a function to autosense the mask on any frame, and alternatively allows independent manual adjustment of the brightness range recorded for each color channel, so that clipping can be avoided and the orange mask compensated to suit. Between these features, it's usually not too hard to bring the scan to a point where any additional modest adjustments to accommodate esthetic preferences can readily be handled in Photoshop. Nikon Scan is much less helpful in this respect.

My impression is that the defaults and the auto modes of scanning software are usually set to produce excessively contrasty scans with the ends of the tonal scale clipped. The different brands of software vary in how much of a nuisance it is to work around the default settings to get exactly what you want.

[phaedrus]

It's like Oren said: you shouldn't limit your possibilities of using that scanned file right in the scanning stage. So: no sharpening, no grain supressing (sometimes, the two together increase the appearance of grain) and no curves manipulation (the film already has a film curve). On the other hand, it makes sense to use the whole bit depth the scanner can produce by setting black and white points excactly.

[brad martin]

As a general rule Photoshop. And experiment, experiment, and experiment. I have found that failure teaches most. I am most familar with failure.

[aphexafx]

...it makes sense to use the whole bit depth the scanner can produce by setting black and white points excactly.

And just to add: produce your scans and save them at 16 bits/channel in order to obtain as much color/tonal resolution as possible before you make adjustments in PS - this will give you as much headroom as possible, which is important.

[sanking]

By coincidence this same topic was discussed recently on the ScanHi-End forum on Yahoo. There are some real experts on that forum who use a variety of drum and high end flatbed scanners. And guess what? They also disagree on the subject. Some, citing evidence that the proprietary software of certain scanners allows for more control, recommend correcting the image in the pre-scan as much as possible. Others recommend the raw scan approach and then do all of the corrections in Photoshop.

My own approach is to apply as many non-destructive corrections as possible in the pre-scan, but leave final corrections that may be destructive for Photoshop.I apply curves to adjust overall contrast and correct for color in the pre-scan, but take care to avoid clipping any of the shadow or highlight detail.

If possible always scan in high bit and keep the file in high bit through all of the corrections. The scanner I use for LF film, an EverSmart Pro, makes the analog to digital conversion in high bit, but then allows saves only in 8 bit. With this scanner it obviously makes sense to do everything possible in the pre-scan in order to take advantage of the high bit conversion of data from analog to digital. With this scanner my work flow is to correct as much as possible in the pre-scan, scan in RGB, and when the file is in Photoshop immediately covert from 8 bit to 16 bit. This work flow allows about the same amount of corrections in Photoshop without posterizing as with a 16 bit grayscale file.

Sandy King

[Cesare Berti]

Many thanks for your comments and thoughtful advice.

I now feel more comfortable that the raw scan approach with minimal processing is a good basis for further work and experimentation.

Oren, my mistake on the Ciba print. The print from the negative I mentioned was made on Kodak Professional Paper, according to the markings on the back not on Cibachrome paper which as you point out was for making prints from slides.

Cesare

[Tyler Boley]

It's important to note, as evidenced in Sandy's post, that this decision is often based on the particulars of one's hardware/software. The are no blanket rules that apply to every situation. I am not thrilled with Silverfast's negafix, so with negs I bring them in raw into Photoshop to work, in hi bit.
Also, for trannies, color management applies and a good input profile can make life much easier. However, it only applies to the setting with which it was made, so again it's best to make edits in PS after conversion to a perceptually uniform nice big working space, leaving the scanner settings locked down to those used for the profile.
Tyler