rgb vs. greyscale scanning

[Dave Aharonian]

I have about 35 4x5 sheets and 25 6x12 negs that I want to have high resolution drum scans made from. Film is mostly tri-x and fp4 all processed in PMK. I have read that it is best to have them scanned as rgb files. I'm curious about the reasons for this as if I scan them as greyscale files the cost is about 50% less than rgb scans. At $45 per scan I'm happy to save some $$ but not if it really affects the image quality.

Thanks for any info!

[chris jordan]

Dave, if the original film is B&W, then there is no benefit to scanning in RGB. If you scan a B&W negative in RGB, the resulting three color channels will all be identical, so that every pixel will have the same three values (128, 128, 128 for example). So in RGB your files will be three times as big as necessary and the scans will be much more expensive.

If you were scanning from color film, to be converted into B&W, then it is a good idea to scan in RGB because then you have three different color channels that you can selectively mix when converting into greyscale. But when starting with B&W film, scan in greyscale and save your money.

~cj

[julian_4860]

What Chris says is right. But I've had drum scan of BW done where one ch. is sharper than the others, so you can choose which ch. to use. However, I've also had scans done where all 3 chnls were equal, so it depends whether you see see anything on their scanner. These kinds of differences are quite common on film and flatbed scanners

[Paul Butzi]

Remember, a negative developed in PMK is anything but monochrome. If I developed negatives in a staining developer, I would be exploring how the quality of the scan could be improved by exploiting the color of the scan.

In fact, now that it's been mentioned, I'll be evaluating my scanner to see if I can get better scans that way, even WITHOUT using staining developer.

[julian_4860]

another thing to try Paul is using NegPos as a conversion tool (PC only). It allows you access to all the default patameters. In colour I'm getting perfectly instep histos from scanning as a linear scan and then inverting using NegPos - never had it before. For pyro BW you could precisely take out (or not) the stain and match the gammas of all three chnls.

[Paul Butzi]

julian- Thanks, I'll check it out.

[Ted Harris]

It's a tossup. IF it is a good well aligned drum scanner then what Paul says has merit. The down side is that there is a good deal of variance in the color fringing from scanner to scanner, although the drum scanners we have tested all had the requisite almsot dead on aligned square wave. Not true of Imacon's necessarily and I mention this because there are labs out there that will tell you an Imacon is a drum scanner when it is not. The color fringing, if it were severe enough (which is highly unlikely on a well calibrated properly operated drum scanner) could offest the gains from an RGB scan.

[Paul Butzi]

Ted-

What I had in mind more than the advantages of scanning, say, pyro negatives, was the idea that for some relatively inexpensive scanners, the alignment and chromatic abberations might be better for, say, green, and ok for red, and then really bad for blue, or some other combination.

If so, then you might get better results from scanning a normal B&W negative in RGB mode, and throwing away the red and blue channels entirely.

Just one more thing to check out. I just finished the LAST development run of film development calibrations I'm doing. Eight film developer/dilution combinations.

Talk about sheer, brutal drudgery. On the bright side, as soon as the film is dry and I read the densities off the films, and entire them in the mondo spreadsheet, I will have accurate development times for TMX and TMY in D-76, D-76 1+1, Microdol-X, Microdol-X 1+3, XTOL 1+0, 1+1, 1+2, 1+3. Whew.

Good news, I guess, for anyone who can use my times. I'll put them up on the web. Maybe I'll take donations to defray the cost. That works out to a big pile of film through the Jobo.

[Brian Ellis]

I've seen conflicting recommendations on this question. David Brooks wrote a lengthy article about scanning black and white negatives in the September, 2002 issue of Shutterbug magazine. He advocates scanning them as slides. His rationale is that scanners are designed to scan transparencies and have a dynamic range of about 3.3, which corresponds to the approximate dynamic range of slides, whereas b&w negatives have a dynamic range of only a half to a third of that. So he says that when you scan a b&w negative as a negative the information from the negative fills only a small part of the scanner's gamut. I'm not saying this is right or wrong, I really don't know, I'm just summarizing his reasoning. Then I've read other articles other places that say there is no benefit to scanning a black and white negative as a transparency. Most notably, IIRC John Paul Caponigro has an article on his web site in which he says that.

So take your pick. One problem for many of us is that if we're scanning 4x5 film in rgb at a scanner's maximum ppi, we can end up with a file either too big for our computers to handle or so slow that the whole process becomes painful. A 4x5 b&w negative scanned at 4800 ppi (the theoretical maximum of the Epson 4990 scanner) as an rgb image will result in a file of more than 2 gigs in size, which isn't practical for most of us. Of course you could scale down to say 300 dpi at your anticipated maximum print size but then you're involved in another area as to which there is some disagreement, i.e. whether it's best to always scan at the scanner's maximum ppi and then scale down in Photoshop or whether you can scale down in the scanner without affecting the quality of the scan.

[Kirk Gittings]

Depending on your scanner there may be an advantage either way in terms of noise. My 4990 has less noise in grayscale than RGB (Michael Mutansky told me in a conversation at the VC conference that certain channels of RGB have less noise. I haven't had a chance to test it).

I actually scan 16 bit grayscale 8x to lessen noise with Silverfast AI Studio. Then I convert it to RGB so that I can use the capture sharpen feature in PK Sharpener (only works RGB) then I convert it back to grayscale. I have tested every option available (except Michaels above ) and that work flow gives me the best starting file on my equipment. It is time consuming but worth it.

[Brian Ellis]

Kirk - What do you mean by "scan 16bit grayscale 8x?"

[Kirk Gittings]

8x refers to the multipass feature 2x,4x,8x,16x in Silverfast AI Studio for noise reduction. It seems with my machine that 8x is optimum.

With the "studio" upgrade they seem to have solved some of the registration issues with multipassing on a flatbed, though it always works better on the second scan. I have incorporated this into my workflow by doing a "batch" scan on each negative. That is I do two scans of each negative, the first is 1x which does nothing more that heat up and expand the negative (which I deleat later) the second is 8x which is the "real" scan. No problems with fuzzy multipass scans since I figured this out.

[Janko Belaj]

Dave, I would recommend you to have one negative (the best, i.e. "the sharpest") scanned as RGB and as grayscale on your service recommended settings so you could compare results. If you could came to Croatia, my scanners (hmm, how do you call a man who scan with scanner?) will made 4 scans at no cost - you will pay for 60 scans at the end and most of us are willing to show you most of possibilities.



Depending on drum scanner your service have, and depending on software they use, and depending on b&w film you have - they will use the best option. Almost every b&w film have some "color" in it or on it's base. Maybe just a small and almost no visible percent, but there it is. So every film needs a slightly different scanning settings.



While working with old Dainippon Screen scanner the best results where those I'w got with red channel or with red and green channel. Latter, with "newer" ScanView's ScanMate scanner (the one with PMT, not CCD), the best results where those I got from RGB scan after extracting the "L" channel - RGB to L*a*b (all in 16 bit mode), then extracting lightens only and after correcting levels and applying some additional unsharpmask saving as final, still large, 8-bit tiff.



As for resolution - there is no "optimal" resolution on drum scanner. It is not CCD with fixed numbers of cells, it is one photo-reader with more or less "frequent reading" of your film structure. But continuous reading. If that would be CCD scanner, than the best, or the optimum setting would be exact physical numbers of cells in CCD (in combination with scanner's lens). Drum scanner can not interpolate, they can scan larger area or smaller area (on some models the PMT goes closer for finer reading, or farther for coarse reading, and some models have a kind of zoom lens).



And finally - that myth for different price of RGB and grayscale scan... the amount of time for mounting color or b&w media is the same, the preview time is the same, time needed to apply settings is usually the same, although some time you need to experiment more for fine b&w scan, final scanning time is the same... only final size of document is different. Of-course, today no-one will be able to put the price down, price is calculated from the time when you had to scan color with 3 passes... mentioned here just for record. My opinion only.

[Dave Aharonian]

Thanks for the info everyone. It looks like I'm going to do a little test and have a PMK neg. scanned as both RGB and greyscale. If I find anything noteworthy, I'll report back.

Thanks!

[phil sweeney]

From the few scans I have had done at my service bureau, I was lead to believe that the unmanipulated scan is RGB. I requested grayscale because of final file size and transfer costs.

So are not all scans inherently (or should I say initially) RGB?

In my case the scanned negatives were RGB converted to grayscale and I understood that the scanner actually did not have a grayscale mode.

[Paul Butzi]

I took a few minutes today and played with scanning a handy test target (my Stouffer 4x5 31 step density step wedge, which happened to be in the scanner when I started).

Some interesting results. Bottom line - in the future, I'll be scanning my B&W negatives in color, throwing out the red and green channels using the channel mixer to switch to grayscale in Photoshop, and then my usual workflow.

Check out my results at www.butzi.net/articles/colorscan.htm (http://www.butzi.net/articles/colorscan.htm).

I'd be very interested to see the results if other people try this with their scanners...

[Struan Gray]

On my Epson 3200 the blue channel is a tad sharper, but it is also noisier and more contrasty. I get best results scanning as a B+W negative, but then I'm not interested in squeezing the absolute last line pair of sharpness out of the scan.

[Kirk Gittings]

Paul that is very interesting and definitive. I will let you know what my similar tests reveal. What about noise in the different channels?

[Paul Butzi]

Kirk-

The big issue regarding noise in different channels is that I have no quantitative way to assess the noise.

It turns out this is an issue in determining the optimum contrast of B&W negatives, as well.

So, I'm thinking hard about noise measurements. I can easily generate various densities on film, but it's quite difficult to separate the film noise from the scanner noise, and the easy path to figuring all this out would be to treat them separately.

What would help would be a set of noise free neutral density. Lee offers a set of polyester ND filters, 75mmx75mm. That would do the trick but it's a bit pricey given what I've already blown on film and developer for this little project. And, of course, I don't know if they're noise free (or nearly so).

If anyone has clever ideas, I'd sure love to hear from them.

[Kirk Gittings]

Paul,

Short of science, do you have a sense from your tests about the noise in different channels? Michael Mutmansky told me that he thought the red channel was the worst for noise. In my opinion I would rather deal with some sharpness issues than noise problems. Since I do not print very large I can deal with the minor lack of sharpness.

A friend of mine has used Noise Ninja to create a profile for flatbed scanner noise that he believes lessens the noise considerably. I have played with the idea and haven't figured it out yet.

[Paul Butzi]

Kirk-

Ok, here's what I've done.

First, I tried comparing scans done on my Microtek 1800f with the Microtek software, and with Silverfast 6 ai

To my astonishment, the scans with Silverfast are profoundly sharper. There's another web article on the way for THAT particular little discovery.

Even with Silverfast, using one channel of a color scan is sharper than a grayscale scan.

Rough experiment - scan your handy-dandy Stouffer step wedge. No sharpening, 8 sample.

Now, for each of the colors, go and select a portion of each step. When you select a section, try to not select a section that has bits of dust, etc. Try to select the same region in each version. In particular, you don't want to compare, say, step 18 on one version to step 12 of another version. You'll also need to carefully adjust the tonality of each version so that they match as closely as possible.

Note on the histogram panel in Photoshop, there are statistics for the selection: mean, median, standard deviation.

For our purposes, we can read 'standard deviation' as noise. That's not quite strictly true, but it's good enough to advance our understanding. You'll never get a standard deviation of zero because of the noise in the step wedge, but if that std. dev. is lower, you can figure you've got lower noise in the scan. (that is, you're closer to the noise floor of the scanned target)

For my 1800f, I got the lowest noise from the blue channel. It was substantially lower than the noise in the grey scale scan.

Again, I'd point out that the results from my scanner almost surely can't be generalized to other scanners. About all we can hope for is a methodology for assessing scanners to get the best results possible. Heck, I suspect that the unit to unit variation might mean that my results can't even be extended to other Microtek 1800f's.

[Kirk Gittings]

Paul,

My tests (all 8x) show the blue channel as the sharpest also, considerably sharper even than an 8x greyscale scan. The blue channel also shows some minor improvement in deep shadow separation also and d-max (very slight). There may be a slight increase in noise over the grayscale scan but much less than the full RGB scan. Overall though enough of an improvement to do all my scans this way from now on. It actually saves me some time as I was converting from grayscale to RGB to run the PK Sharpener capture sharpening anyway. It saves me a step.

[Paul Butzi]

Somewhat more along these lines - beyond the color scanning, it can make a hell of a difference what software you drive the scanner with.

Check it out at www.butzi.net/articles/scannersoft.htm (http://www.butzi.net/articles/scannersoft.htm)

Those results were a heck of a surprise to me. Silverfast whips the stuffing out of the Microtek software for the ArtixScan 1800f.

Kirk, could you do a more indepth review of the Silverfast upgrade?

[Kirk Gittings]

In the upgrade there are some aditional image adjustment tools which are useful but not necessary. I bought it because there are some additional improvements in the multipass software particularly in the alignment aspect. However from your posts you do not seem to have an issue with alignment of the passes. Maybe the Microteck is more precision in that regard than the Epson.