View Full Version : Expert opinions on digital workflow

Ed Pierce
14-Jan-2004, 12:36
I'm a digital newbie trying to scan 4x5 b&w negatives. After quite a bit of research and experimentation I've found the following procedure to work ok, but I suspect they could look better. I'd appreciate any input from people who know about ways to make this work better.

Main purpose is publication on the web. Secondary purpose is to have scans useable for newspapers, magazines, small posters etc. I am not anticipating making fine prints with these scans.

Materials: 4x5 b&w negatives.

Equipment: Epson 2450 scanner, 256mb memory, Photoshop 7.

Here's how I'm doing it now:

1 Using Silverfast SE, set up as a color positive transparency. Using the histogram, push the black and white points all the way out. Set 48>24 bit RGB mode. Scan at 1200dpi, results in a file of about 70mb, which is about as big as my system seems to want to handle.

2 Save a master copy as raw.psd and move to a CD.

3 Save a working copy as working.psd, crop.

4 Convert to lab mode, delete b channel and alpha 2 channel.

5 Convert to greyscale

6 Invert

7 Unsharp Mask

8 Create a new adjustment layer for Curves; adjust curves.

9 Burn/dodge as needed.

10 Save working file.

11 Save as web.psd.

12 Change image size to 500 pixels wide for large web image.

13 Save as large_web.jpg.

14 Open web.psd, change image size to 125 pixels wide and save as small_web.jpg.

Any comments or suggestions? Thanks very much.

Paul Kierstead
14-Jan-2004, 12:41
Well, workflow varies wildly. For example, I scan in 16 bit grayscale mode in Silverfast.

My only comment is that most images require additional sharpening after resizing; not a lot, but a light touch really improves things. I use Photokit Sharpener (which I find truly wonderful) but you can certainly do it be hand as well.

Mark Houtzager
14-Jan-2004, 13:21
Hi - looks fine. I hope that you would first clone dust etc out before you save on to CD! I would do unsharp mask at the last moment, and you may want to experiment with other methods of sharpening, including using a sharpening layer that does not affect the basic file.

My flow is:

1. scan in 48 bit negative RGB

2. raw levels

3. channel mixer monochrome

4. mode grayscale

5. clone dust etc

6. save onto cd

7. mode 8 bit

8. levels / curves / whatever in layers

9. save

9. resize to print size

10. sharpen in layer

11. print

12. put in box and never see the light of day again ;-)

Ralph Barker
14-Jan-2004, 13:56
As noted, workflows vary widely - particularly the methods used for scanning and color-to-B&W conversion. Some variations depend on personal tastes in the look of the B&W version, others on the image itself.

I, too, scan B&W negs as RGB positives, and invert in Photoshop, as that seems to give the cleanest results using SilverFast with my Epson 3200. I'll often make minor adjustments to the Levels and Curves on the negative image prior to inversion, and then final adjustments after that.

I've found the best way to preserve image quality when down-sizing for the web is to do so in steps of less than 50%, using a light touch of unsharp masking at each step. The unsharp mask settings I use are Amount=50%, a Radius of 0.4, and a threshold of 0.

In theory, the radius should be adjusted based on the size of the image in pixels. A large image of 3K-4K pixels on one dimension, for example, might use a radius of 2.4 or so. However, a radius of 0.4 works well on most images of around 550 pixels on the long side - a good size for web display. Plus, Photoshop stores the values, and keeps the last filter used at the top of the Filter menu drop-down - very handy when trying to speed up the work flow. So, for larger files, I simply apply the filter multiple times, depending on the characteristics of the specific image. Thus, my work flow for image-size reduction looks like this:

1. edit full-size scan (e.g. levels, color balance, spotting, etc.), let's assume 3,000 pixels on the long dimension. Save as .PSD file.

2. apply unsharp mask 2-3 times, depending on image,

3. reduce image to 2200 pixels, apply unsharp mask 1-2 times,

4. reduce image to 1200 pixels, apply unsharp mask once,

5. reduce image to 900 pixels, apply unsharp mask once,

6. reduce image to 550 pixels (display size), apply unsharp mask once.

7. add copyright notice, flatten layers, save as .JPG, typically using something around level 8 in Photoshop's compression scale.

Michael Chmilar
14-Jan-2004, 15:34
I have to say that I find the advice to reduce image size in steps to be baffling. Ralph presents this advice above, but it is also in Martin Evening's Photoshop for Photographers book.

When you resize using many small steps, you achieve the same effect as resizing in one step, and then blurring the image with a very small blur radius.

If you prefer the "look" of the multi-step process, that is fine. However, from a technical viewpoint, you are just producing a slightly "softer" image than the one-step resizing produces. This same softness can be produced, by one-step resizing, followed by applying a blur. You gain more control, too, since you can adjust the radius of the blur, whereas in the multi-step method the blur radius is implicit.

By applying an unsharp mask at each step, you are attempting to undo the slight softening that is introduced in each resizing operation. This seems counter-intuitive to me: if you don't want the softening, why do the multi-step resizing? Why not get a sharper result with one-step resizing, and avoid sharpening?

Henry Ambrose
14-Jan-2004, 19:21
Too many steps IMO.

The short version: 1) scan with Epson driver in 48 bit gray with proper adjustments in the driver (yes, the Epson driver) 2) do overall adjustments with curves and spotting in 16 bit, convert to 8 bit and continue 3) sharpen via USM a bit maybe a little less than you think is enough - save the real sharpening for when you know the final output needs. For now just get rid of the softness. 4) archive the file to your big external hard drive which you will migrate to the best new storage media at some point in the future 5) when you need the file, go to your archive and then sample down and USM as required

Longer version: Scan in 48 bit -grayscale- after setting tones in Epson driver. Set the highlight and shadow a little outside of what the neg seems to indicate to make sure you don't clip anything. Maybe a little curve to save you having to do it all in PS. (you'll soon enough develop a routine for known films and conditions but for one-off scanning just skip this). When you get the scan in your machine do overall adjustments in 16 bit including spotting. Convert to 8 bit and tune more as needed. After you get it toned and adjusted just right, USM just enough to make it look pretty good on screen at 100%. When viewed at 66.7% it should look almost too sharp. At 100% you should be thinking its not quite sharp enough. and you are there. USM settings somewhere around 200-300% at .3-.6 pixels and 4-6 threshold. Save it - this is the file you will archive. I suggest a big external firewire hard drive that you fill with your archives rather than CDs. CDs could be a useful back-up archive. Do not count on them for long life. I have a few (out of several hundred) CDs from 1994 and 1995 on up to about 1997 that are dead. Trust me on this. But you do have the film so if you lose one you can scan again. Next (or later when you know what size you want to repro) size it as you please for its intended use. USM again. You'll need different settings depending on how you will use it. Offset printing needs more and perhaps coarser than you might think. Use the same test as above except for little web thumbnails which you will "oversharpen" to make snappy but just short of ugly.