Is it better to scan B&W Negatives as Positives?

[jasonsmith]

I use a Nikon Coolscan 4000 film scanner. I have found when scanning

B & W Negatives the final scan sometimes seems to be lacking in

tonal range.

 

I read recently that to get around this you should scan Black and

White Negatives as colour positives which enables you to use the

full dynamic range of your scanner. This means that you need to

invert the positive image in Photoshop.

 

I have attached the Raw Scans from both methods here.

 

Does anyone use this method ?

 

Would the same hold true for colour negatives?<div></div>

[lwood]

I have found it to be better to scan color and b/w negatives as color positives and invert and correct in Photoshop. I have a Konica Dual Scan IV. I usually invert and desaturate the black and white stuff immediately anyway, but for some reason this seems to work best. Color correcting negatives has been a bear for me no matter how I do it. I suppose one of these days I'll break down and get a digicam,

Louie

[jasonsmith]

Here is a screenshot from Nikonscan.

 

When I scan as a positive the histogram for each of the Channels is initially quite thin looking.

 

I then bring the sliders in so that the information is stretched accross the range.<div></div>

[jasonsmith]

After bringing the sliders in it looks a little better.<div></div>

[xenotar28]

I did a quick copy of your color scan photo and expanded the tonal range in Photoshop levels. There's plenty of range, though I also scan my negs as slides as it makes for easier adjustments in PS.<div></div>

[jack paradise]

Hi Jason,

 

You can also try scanning your b&w negatives as "Grayscale Positive", instead of RGB Positive.

 

Here's a scanning workflow for it:

 

http://www.shutterbug.net/features/0902sb_bw/index.html

[helenbach]

I use a similar technique to the one outlined by Jason, except that I would normally avoid the clipping at the right hand side of the histogram (unless there was a reason, which Jason probably had) and I invert in Nikonscan, so that I'm previewing the positive image. The way to invert the image is to make the RGB 'curve' (or line) go down from left to right instead of up. I don't understand why everyone doesn't do this (only joking...).

 

I use greyscale positive if the negative hasn't been developed in a staing developer, RGB positive if it has. If the density range of the negative is low enough to avoid clipping when scanning as a negative, then I scan as a negative.

 

Best, Helen

[helenbach]

That should have been 'staining developer', not 'staing' developer.

[jtk]

I find the neg-as-pos workaround loses fine detail in the inversion step, even though in TIFF all the way, compared to one-step scanning with Vuescan. In other words, neg-as-pos in Nikonscan is less sharp than Vuescan.

[mendel_leisk]

With most scanners and their bundled software, if you scan as a color positive, you are saying: just give me what you've got, don't do things to it.

 

With the Minolta line, if you specify 16 bit linear output, it will be very close to the raw data, uninverted, and not even gamma adjusted. Don't know if there is something similar with the Nikons.

 

With Vuescan, if you output the raw file, it will be similar to the Minolta 16 bit linear, above.

 

With color negs, you might want to read my recent posting on the subject:

 

http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=00CWmS

 

Or maybe not, it's pretty convoluted :)

[kelly_flanigan1]

Re <i> Is it better to scan B&W Negatives as Positives? </i><BR><BR>This was a common trick over a decade ago. Then the canned scan software was alot klunkyier; scanners were alot more non linear; and had less range. With our old 300 and 600 dpi flatbeds that cost as much as a Hasselblad; this "trick" was a common way to get a better scan; with more shadow or highligh detail. Today scanners are radically lower in cost; with better software and way wider dynamic range. The ancient "trick" may not buy one as much today; but one should at least try it and still experiment.

[helenbach]

John Kelly wrote:

"I find the neg-as-pos workaround loses fine detail in the inversion step, even though in TIFF all the way, compared to one-step scanning with Vuescan. In other words, neg-as-pos in Nikonscan is less sharp than Vuescan."

 

At which stage or with which software did you invert the Nikonscan TIFF? I've found no difference in sharpness or detail between Vuescan and Nikonscan when scanning neg-as-pos in Nikonscan and inverting in Nikonscan.

 

Best,

Helen

[jasonsmith]

Jack, or Helen

 

Could you please elaborate a little on when you might scan as a Grayscale positive, as opposed to in RGB.

 

Helen - by clipping are you refering to bringing the highlight slider in.

 

Should I be bringing the sliders in on my scanning software to expand the tonal range - or is it better to leave them and let levels do the job in Photoshop.

[jack paradise]

Jason,

 

For "Media" choose "Slide film" and for "Bits per pixel" choose "16 bit Gray" instead of "48 bit RGB".

[erik scanhancer]

In addition to the Shutterbug method here's mine (it's a bit shorter reading and I keep

everything in 16 bit):

 

http://www.scanhancer.com/index.php?art=20&men=3

 

Read on from the Scanning Tips header and never mind about the Scanhancer mentioning.

Remember that scanning in RGB only lengthens your scanning time by a factor 3. Try to

scan in B&W Positive mode if possible! The final result will be the same (despite what the

RGB gurus claim).

[david_moore13]

One should never use the Photoshop "Invert" function to convert a negative to a positive. Although the result often "looks about right" it is a mathematical coincidence that causes this to be true. You are not performing the same mathematical operation that exposing photographic paper would cause. With a B&W negative, you can usually get by this way, because you can get the right contrast later using levels and/or curves. However, you are creating more work for yourself by using the Photoshop "Invert" function. Here's why:

 

Photoshop invert performs the mathematical operation 1 - x, where x is the pixel value, scaled to between 0 and 1. The original linear image data, captured by the scanner, has been gamma corrected with the operation y^0.45, where y is the linear data. Thus, after Photoshop inversion, the complete math from linear data to output positive is 1 - y^0.45, where y is the original data from the scanner (scaled between 0 and 1).

 

The "correct" mathematical operation, which occurs when you expose photographic paper, is c/x, where x is intensity of light shining through the negative onto the paper, and c is the chosen "exposure time" of the paper. A value of c=0.1 is about right for visualization purposes. (Dealing with film base color and other factors complicates this slightly, but that's the basic idea)

 

Coincidentally, the functions 0.1/x and 1-x^0.45 look rather similar, even though they are doing different things. This is why the Photoshop "Invert" function looks passable, but requires lots of cleanup with the curves or levels functions. This is also why this method fails utterly when you use the "16-bit linear" output of the scanning software -- because the x^0.45 factor is eliminated, and the function 1-x no longer is close to 0.1/x.

 

My practical advise is to use Vuescan, because it seems to use the correct function as described above. However, I don't particularly like the controls offered by Vuescan (things like paper gamma and exposure time are not available in the interface, leading to clumsy manual control). I'm working on a photoshop plugin that should do the "right thing" and I'll release that soon (for free).

[tomweis]

<p><i>One should never use the Photoshop "Invert" function to convert a negative to a

positive. </i></p>

<p>Never say never.</p>

<p>I scan my B&W negs as positives and invert them. It works for me and yes I have had

digital C-prints made from them with success. I find I get much more detail in highlights

and shadows this way. I assume it's because positive (transparency) film is contrasitier and

the scanner is making up for that fact.</p>

[marcos_rodriguez]

David, I found your equations most interesting.

 

I have a Minolta ScanElite 5400, and its scanning software simply stinks --it clips highlights and shadows like crazy. So, I've experimented extensively with a number of alternate methods (for color negs), like scanning as positive (16bit linear), which seemed to nearly do the trick.

 

However I find it nearly impossible to get an acceptable gamma curve, and color balancing is always near but never quite there --cyan and blue casts are the ever-present curse.

 

Now I have the mathematical explanation of why this is so. Very enlightening and interesting read, thanks!

 

Marcos Rodriguez / www.aukeramen.com

[mendel_leisk]

David, the plug-in would be great, I'll be watching for that!

 

I've tried Photoshop gamma adjustment and inversion of Vuescan raw file, scans of tri-x. In the reusults, the highlight areas in particular seemed flat and washed out, compared to using the same raw file with Vuescan scan-from-disk (with Tmax400 profile with D76ci:.55 contrast index).

 

I've seen another discussion here on the subject, here:

 

http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=00D21a

 

This stuff is over my head but I struggle to at least take advantage of anything that comes along :)

[joshua daniels]

I have two scanners - the Minolta Multi Pro (very similar to 5400, but max dpi 4800) and the Coolscan V.

 

After extensive experimenting with the Nikon and various of the driver setting, and with Vuescan, I got best BW results (low grain, long dynamic range), but scanning bw negs as color negs with GEM set to "2." It's a simple workflow, and based on my many hours of playing with the scanners, settings, and software, produces the best overall result from various types of bw film (Kodak, Fuji, and Ilford). Of course, your milage may vary. The results from the Minolta, without Scanhancer, are grainier, with or without their grain dissolver.

[jtk]

Helen asked about my observation:

 

"... neg-as-pos in Nikonscan is less sharp than Vuescan."

 

Helen, when scanning neg-as-pos I used Nikonscan in TIFF. I do not manipulate scans and in particular I do not allow sharpening. My steps were very simple: scan TIFF at film's default settings,

invert in Photoshop.

 

Mendel reports (TriX) muddy highlights with the Nikonscan inversion workaround: with Nikonscan inversion I observed (Neopan 400) loss of finest, most ephemeral details . Perhaps the fine detail loss had to do with our respective default settings, eliminating (or creating) Mendel's muddy highlights.

[mendel_leisk]

John, my converted Vuescan Raw File from Tri-X had highlights which seemed "flattened", not "muddied". Say on someone's cheek, facing towards the sun, the Vuescan version showed more tone gradation. With my inverted raw files, that gradation seemed lost, the highlights more washed out. Not blown, but just too flat.

 

Incidentally, I'm using Minolta Scanners, I'm nowhere near Nikonscan.

[jordan_w.]

I'm coming to this thread a little late, and hopefully people are still reading it, but a few thoughts on this --

 

The key is to not clip the histogram at all, and with basic scanning software like the one that came with my Scan Dual II, the best way to do that is to scan as a positive. Though I use Vuescan, I still scan all my B&W negs as positives. I do the inversion in Photoshop and adjust the black and white sliders as described at the top of this thread. Then I do a curves adjustment which usually involves "pulling down" the shadows. I find that this is almost always required to get good scans.

 

The math of the inversion process is very interesting -- discussed also in this thread, referenced above: http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=00D21a

I need to read through it in more detail and draw the curves to satisfy myself.

 

However, I fail to see how a straight inversion in PS can cause detail to be lost. I like to think of the B&W image as a three-dimensional graph, if you will -- x and y are the dimension axes of the image while the z-axis represents the tonal value at any particular (x,y) point/pixel. Seems to me that inversion should merely involve some function of z. Since x and y are not involved, how is detail lost in PS? Does PS secretly sharpen as it inverts? If it is manipulating data on the x and y axes during inversion, why is it doing so?

[david_moore13]

Yes, Jordan, you are completely correct that as long as you are not clipping, you aren't losing information. It would be possible to do inversion correctly by using the Photoshop Curves function, and drawing a custom curve that represents the inversion function. The only problem with this approach is your curve would depend on the film base color and the exact exposure, so it would be a user interface nightmare because it would have to be slightly different for each negative you scan.

 

But as you pointed out, with a B&W negative, it's pretty easy to do this process by hand to get an attractive result (even if it doesn't match the inversion function exactly). With color negatives, it's hard to eliminate color casts using this method.

[jordan_w.]

David -- I understand your point about the custom curve needing to be slightly different depending on base colour and exposure.

 

How about this: adjust the black and white point sliders to the edges of the histogram (or the equivalent -- this is mostly what Auto Levels or Auto Colours does, though I think these functions clip a bit too) BEFORE applying the custom inversion curve. Wouldn't doing so negate the effects of base colour and exposure by "normalizing" the image at the scanned Dmin and Dmax?