epson scanner software (curves, w/b points etc.)

[octofish]

Hi dpug folks,

Is anyone able to shed some light on the way the Epson scanner software works? I am trying to get the best possible results out of it for colour negs (using a 4180 scanner and windows 7) but I'm getting a bit confused about levels/curves etc.

When you do an auto-exposure, it sets the black and white points a little too aggressively, even with the soft-clipping that the output levels slider give you. What I am doing is trying to move the white point to include more of the information. I also find that auto-exposure has also set different gammas for each channel, so I normally set them all to 1.0 for all channels and then tweak the white points a bit more to get a reasonable colour balance.

What Im not getting is that sometimes the auto-exposure actually looks reasonable with these weird values. My method gets scans that tend to be a bit on the flat side, and I have to boost contrast in post. I assume that the 'soft-clipping' is trying to simulate the shoulder you would get by printing on paper so you get reasonable contrast. Are the weird gammas compensating for the curve thats introduced as a result of the highlight points being in different spots for different colours? How does this even work? Sometimes I struggle to get colour as accurate using my workflow to what auto-exp gives me even with its weird gamma values, so I want to understand what its doing and why, so I can do it myself, but better.

What I really want is to just get the most accurate results as far as whats on the film. Is my assumption that 1.0 gamma is accurate incorrect? While Im getting less cross casts than the aggressively clipped curves from auto-exposure sometimes give me, they are still present, so Im suspecting I may need to tweak the gammas a little bit at least. Im also finding that if I dont allow at least some soft-clipping of highlights, apart from it looking flat (which I can fix), I start to run into posterization issues and noise as the middle values are compressed too much.

Im really loathe to use much of the softclipped shoulder and non 1.0 gammas cause they are really tough to get right based on just eyeballing the preview without introducing all sorts of cross casts which are impossible to fix in lightroom (no proper curves tool). But I really dont want to go to doing a full linear scan followed by curve correction in photoshop as its much more labour and data intensive. What I really want is to make sensible choices about contrast and dynamic range at the scan step and have the colour be basically right, so I only need to make minor tweaks in post. Is this reasonable or am I being a big baby?

Anyone been down this road?

Sorry for the long post!

 

[tkamiya]

I use EPSON V500 with EPSON's scanner software.

What I usually do is to do a pre-scan first, then set the white point and black point away from the active portion of the histogram. That sometimes mean WP and BP are all the way to left and right respectively. I usually don't play with the curves at the time of scan. What I observed is that auto-exposure clips on both sides, as you have also seen. So I manually change that to scan the full range of tones. If the image is "clipped" already at the time of scan, it's lost forever. That's not good....

I am not sure if there is such thing as "accurate scan" when it comes to scanning films. In my darkroom processes, I "interpret" the negative by using all sorts of techniques. My goal isn't to produce a print that is "accurate." My goal is to produce a print that is pleasing and convey my intent and emotions. As such, when I scan, my goal is to get the maximum range of tones. Then when I print (ie. in post), I interpret the scanned images. It will have to be this way because the range of tones available on film is far wider than what can possibly be printed on paper. So some compression, omission, manipulation, etc, etc, etc, would be necessary.

Anyway, right or wrong, that's how I do mine.

 

[octofish]

Cool, thanks for the reponse!

Do you do anything to the middle slider in the historgram that does gamma, either for the whole image, or for individual channels? Do you manually set it to 1.0 or just leave it as whatever auto-exposure sets it to? This is the bit that's confusing me. I'm starting to suspect I need to mess with it to get the best results, but it seems hard to get it right without introducing weird colour casts that are hard to fix.

Also, am I right in thinking you're getting pretty flat scans by moving the wp and bp all the way out? What's your process for getting reasonable contrast back in during post? Do you use a curve either for everything or specific colors, or just brightness/contrast/color balance type controls?

Curious to know how people are doing this...

 

[chuck94022]

I generally scan it flat (no adjustments at scan time, no auto anything). I adjust in post, using PhotoShop. My normal process is: 1) straighten, 2) for B&W, invert (because I scan negative as positive), 3) Levels as needed, but usually only the left and right, not the mid arrow, 4) visit Curves, but usually try to leave it alone here, if I got a good negative (which I try hard to do) - but if I do Curves, it usually means I had a problem negative, 5) Brightness/Contrast (which I think is just simplified curves, but a simple tool for simple tweaks) - or, perhaps if needed I'll go into Exposure at that point, maybe adjust that, and maybe even Gamma if I think I need to, then finally, 7) I'll sharpen as needed to fix scanning deficiencies. Since I've gotten a www.betterscanning.com mounting station to replace my Epson EOM film holders, and I have done the necessary adjustments to get sharp focus, I don't need much sharpening anymore.

Hope this helps! (And others may suggest improvements in my flow, everything helps.)

 

[chuck94022]

Oh, and I should mention I have an Epson v750 and usually use VueScan scanning software, but the standard Epson software is similar. VueScan has a few more features but you can get along without it.

 

[tkamiya]

Yes, my out-of-the-scanner images are pretty flat.

Here's my process:
1) Scan as negative. Set the white point and the black point wider than what the histogram indicates. I do not touch the middle slider. Change the exposure setting if necessary.
2) In post, using Element, adjust the following in this order:
. dust busting
. contrast
. exposure
. white point and black point
. middle slider or the curve ONLY if necessary
. selective adjustment (dodge/burn)
. review image and go back to prior adjustment if necessary

I do these in this order to minimize going back-and-forth. I almost never sharpen as I find the edge contrast is usually sufficient. Finally, I typically save the original image intact. Then once dusted, I save it as imagefile.dusted. Then once adjusted, imagefile.a. If I make different versions .b, .c, and so on.