Scanning with 8- or 16 bits per channel?

I am using Nikon Coolscan 8000 for my 6x7 negs. I find

difficulties in printing the dark shadows in my prints. (most picts

taken inside by exisitng light). Dark areas start looking dirty, also

when I donot use any Colour- or lightningadjustments. The

neg's look fine - no underexposure. Should I consider scanning

with 16bits dept. I hesistate a bit because of the large files I

would get and the long time the scanning will take (large

file+dig.ice+16 bits). Has any one experience in this field.

Thanks, Marrigje

yes you want 16 bit. 8 bit means that you have a range of 256 shades per

color channel. 16 bit has a range of about 64,000 shades per color channel.

<P>You also do not want to be in the sRGB colorspace it is just too narrow.

Better is the Adobe (1998) RGB colorspace as it is much larger. sRGB has it's

uses, but only as a final colorspace to switch to when you are getting ready to

publish on the internet. 8 bit is You may even been to do multipass

scanningto get everything out of the film.<P>Don't do any color or other

adjustments in your scanner software, wait till you open the image in Adobe

Photoshop.<P>Also : do you calibrate and profile your monitor? If not, How do

you know what your scanned images actually look like?<P>As for storage

space, how much does a CD-R cost? $0.18 to $0.20 USD?

Thank you Ellis, I will try it. It is not because of the storage space

that I hesistate, but because of the time. Also I need a lot of RAM

and patience to work at these files. I work at a school and

scanningtime is limited (much students want to use the

scanner). I do work in Adobe RGB 1998 and I never work on my

scanns. Thanks again for the very quick response. Marrigje

With marginal negatives and slides; I tend to scan in the higher bit modes of my scanners. Also one can use curves in ones scanning software before the actual scan. This sometimes gives better/more smooth shadow detail.. You will be spreading the detail over more bits in the underexposed toe area of the film. I did this last night with a scan of a 400TMAX negative that was slightly underexposed; and also overdevloped a tad. The usage of a inverted s curve boosted the shadows; and dropped the highlights; which yielded a better "image". When done scanning; I only had to do a slight adjust in Photoshop. <BR><BR>Then I reduced the file back to a 8 bit file; to save space. For valuable images; I still keep a 16/14 etc big file; which is usefull if one wants to do more Photoshop work. Some Photoshop tools are greyed out in hi-bit modes; and many older printers will not accept more than 8 bit files.

For minor color adjustments; I do the adjustments after scanning; like Ellis mentioned. <BR><BR>For radical color adjustments; sometimes one must do this before scanning to get a better image. This is ususally when I am scanning ancient color negatives or slides with little colors; and radical dye fading that varys with each different color. Doing this will yield a much better image; but this requires alot of patience to get the best prescan. Typically I will have to use a different radical steep curve for each R G B layer; because the dyes are all shot after 50 years (1940's Kodacolor negatives and Ektachrome, Anscochrome, Fuji ). Better scanner software will allow one to save this prescan profile; so it maybe used on the next image/slide/negative. Typically a group of slides or negatives stored together will fade sort of in a similar manner. Sometimes the outer ones also got light; and will fade across the frame. These require major work; sometimes spicing two different scans of the same image. This is a major league sink hole for time; an one should charge for it if it is for a scan for others.

I found this site useful in understanding Histograms and trade offs when look at shadow areas. http://www.marginalsoftware.com/HowtoScan/tutorial_page_1.htm

I've also started to use Vuescan which I'm finding easier to use than the given Nikon scanning software.

Good luck, Les.

"I find difficulties in printing the dark shadows" + "The neg's look fine". - Scanning is certainly one possible culprit, but just on the information provided I'd also suspect your printer profile. Getting a good profile of the heel of the tonal curve is very hard to do even with expensive profiling equipment.

 

Try this. Take a crop from a problem shadow area of one of your image files to use as a guinea pig. Add a Curves adjustment layer. Lock down the initial "curve" (diagonal line) with several dots along its length, then give it a slight hump at the bottom quarter of the curve (the heel). Print this, then make a few variations on the height and extent of the hump and print them.

Using 16 bits does not only give you finer gradation in luminance, but also in gamut. In fact, this is the primary reason for using 16 bits when working in wide color spaces.

<p>

<i>Analogy</i>: If your picture is likened to text, the color space is the language and the gamut is the vocabulary. Pretend you want to write a discourse on "ice" (that's our small gamut image). You could write it in English (wide color space), or you could write in Eskimo language (narrow color space). The eskimo language has a poor vocabulary (requires fewer bits per channel); they don't need to express a wide range of idea. However, they do have twenty subtly different words for "ice". How fitting for a discourse on ice! So you decide to use Eskimo language, which can be represented using fewer symbols (bits per channel). Had you decided use English, you would have been stuck with a vast vocabulary, most of which is useless to your discourse!

<p><b>

The conclusion is that for best results you should match your color space to your image</b>; i.e., sRGB is sometimes more appropriate than Adobe RGB! If your image does not have a wide gamut, as is typically the case, then you can use a narrow color space and 8 bits per channel, but if you do use a wide color space, then you <i>must</i> use 16 bits per channel.

<p>

<a href="http://www.creativepro.com/printerfriendly/story/7627.html">Here is another explanation, with examples</a>.

Thank you all for your most helpfull answers. Plenty of work for

more weekends to come. I will certainly tru the 16 bits scanning,

but I will not forget the "eskimo-route" through sRGB. See what it

brings me. You ever read the book "Miss Smila"s feeling for

snow". Only for that reason! Regards, Marrigje

This is a problem I've been confused on. Why EVER use 16 bit when 1) 1/2 the items in Photoshop 7, the most important ones, don't work, 2) most printers are only 8 bit?

<a href="http://www.creativepro.com/story/feature/16098.html">Why do high-bit editing?</a>

<ol>

<li>I have no difficulty editing in 16 bits. <a href="http://www.creativepro.com/story/feature/16097.html">Here are some pointers</a>.

<li>How do you know that printers in the future won't be able to use all 16 bits?

</ol>

Mike ; Scanning at higher than 8 bits allows one to get a better scan with marginal images. It is true that alot of Photoshop fucnctions and printers wont work with greater than 8 bits; but one can always convert to 8 bit after scanning. For a properly exposed negative; the net gain is slight to nil. For an underexposed negative with shadow detail; scanning with a higher bit mode will yield smoother shadow details. <BR><BR>An analogy is balancing ones checkbook; do you use cents; or just round to the nearest dollar? Balancing using cents will yield a balance within a penny. Round each transaction to within the nearest dollar; and then balancing will yield a balance that may vary several dollars; after many hundreds of transactions. A proper Accountant may work to cents; and then display the results to the nearest dollar; thousands of dollars etc. Enron and Worldcom listen up! :)

Kelly,

 

You cleared it up a little but you're making one big assumptions. This is that Photoshop does a better job in converting to 8 bit than my scanner software. I use Viewscan and it always scans in 16 bit mode regardless of the options set. It will then save the final image in whatever mode and space you choose. My native software did the same thing.

 

I would be really suprised if most scanners' software does anything different. It wouldn't make sense.

 

Mike,

If you adjust an 8-bit scan in Photoshop, the histogram develops gaps. It gets worse if you make additional adjustments. This can appear as posterization, especially around highlights. Faces are particularly affected.

 

If you adjust a 16-bit scan, you are just picking the 8-bit slice you display. The histogram stays smooth (if you exit Histogram then re-enter.) Adjustments must be pretty extreme before you get posterization. You can then change to 8-bit mode, and the adjustments hold.

 

Scan times are mostly affected by file size and post-scan processing, such as Digital Ice. At 2900 dpi with Digital Ice, the Nikon 4000ED takes 4 min/frame at 8-bit, and 6 min/frame at 14-bit depth. Without Digital Ice, the difference is negligible.

After all the good advice I tried to make a 16 bit scann, but no

luck. Our schoolcomputer refused to do it because of

"scratchdisks are full". I tried several sizes but noway.

8-bitscanning + dig. ice works okay. So I tried another solution. I

know I have a very simple printer (Epson 750) so I took one of my

files to a proshop. The print was better, but not completely okay.

According to them my files were to big. I should make scanns for

A4, A3, A2 etc. instead of downsampling. Can you react on this.

Here is what I do. 6x7 neg. scanned 65x55cm 350 dpi + dig.ice,

no sharpening gives a 130MB file. I do some slight work on the

histogram and the midgray-point before scanning. All within the

range of the scanner (nikon coolscan 8000) so no extrapolation.

I work in photoshop 6. First changing the colorprofile to the

workingprofile Adobe RGB98. Than in an adjustment layer I

make some coloradjustments in curves and if necessary make

a slight S-curve to change the contrast. Than add unsharp mask

and rescale in the printing program to make an A4 print.

You find some of my pictues under

http://www.photo.net/bboard/user?user_id=582864. The name of

my presentation is Silent wittnes. Hope you can give me some

more clues. Thanks, Marrigje

"Scratch disks are full" means your hard disk ran out of space! Remove Windows' temporary files, if you know how.

 

As for the file size: ask them what file size or resolution they want. If they give you the pixel dimensions than you can go to "Edit>Image Size" in Photoshop, select "Resample Image" AND "Constrain Proportions", and enter in "Pixel Dimensions" the numbers they gave you. Of course the aspect ratio has to correct, so if the height and width don't agree you also have to crop. The Crop tool allows you to set the aspect ratio.

 

On the other hand if they give you the resolution again use "Edit>Image Size" but this time enter the information into the "Document Size" section. You might need to change the units from inches to cm/mm. They should recommend a resolution between 200-400 pixels/inch = 80-160 pixels/cm. Do not accept a wildy different number.