PDA

View Full Version : what is your scanning procedure for epson 4990?



dede95064
1-Oct-2008, 21:40
hello,

i'm having trouble figuring out the settings for the epson software that comes with this scanner.

specifically, in the pro mode while using the histograms adjustment. i am confused about the input and output levels. now i presume i'm supposed to have the black/shadows and white/highlights triangles at the end of the curve for the input. the output defaults at 10 and 200. why? does that not mean that it cuts out the shadow values from 0-9 and the highlight values from 201-255? should it not be closer to 10 and say 245 instead? the output preview has a different curve and photoshop has the curve looking different as well.

the help tab is anything but helpful.

any insight appreciated!

thanks!

Ken Lee
2-Oct-2008, 05:55
Writing is slow. If you would like to arrange a time to chat over the phone, PM me and I would be happy to walk you through it. I have a 4990 and use the Epson software. In fact, I like it better than the alternatives. I'm on the other coast.

Jim Ewins
2-Oct-2008, 20:02
I'd suggest you save the histogram adjustment for Photoshop camera raw as well as most cleanup. Go with a high resolution and a size that will be closer to the intended output. contact me off forum if you wish, Jim Ewins, Seattle.

Mick Noordewier
2-Oct-2008, 21:06
Well, dede95064 has some questions that have puzzled me as well.

I like the Epson software because it is easy to crop the initial scan. I turn off automatic image adjustment, set the white and black points outside of the histogram, capture, then do further fussing in photoshop.

However:

1. Why isn't the Epson output 0-255?

2. I figure that setting closer white and black points results in a scan that requires less image interpolation. But am I just making work for myself: What if I capture the entire range in 16-bit and adjust in photoshop?

I may just call Ken to get some answers myself...

-Mick

Ron Marshall
2-Oct-2008, 21:39
I use Silverfast, scan in RBG, and strip out the green chanel in Photoshop (it is slightly sharper than the others). I make all corrections in PS, I only set the endpoint sliders in Silverfast.

Ken Lee
3-Oct-2008, 05:04
I don't know why the Epson software doesn't default to 0-255. You have to stretch those values every time you scan a new image, and that is annoying. They are probably trying to help the average consumer from making a harsh image.

There is no need to make sure that your output fills the entire histogram, especially if your subject doesn't have anything in it which is pure black, or pure white.

When scanning, I try to get the overall image tonality as close as possible to the way I would want to print it - as if there were no Photoshop adjustments to follow. Even in 16-bit, the fewer changes we make, the better.

This approach is something like what we do in the Zone System and BTZS, where we try for a negative which will print nicely on a #2 paper.

Getting images to print nicely on #2 paper may sound like a silly or compulsive goal, but we admire the richly nuanced tonality and the intangible natural feeling that results. We know that if we don't get the magic of the subject onto the ideal range of the film in the first place, it will be harder to get it afterwards, and what we do get, will probably look stilted, or amateur.

It may not help sell a lot of Photoshop instructional videos, but I try to stay as close as I can to what's on the film, and make as few "downstream" adjustments as possible.

When I look back at images I made via scanning when it was new to me, I want to re-scan them and do them again. I now recognize a certain flatness in the image that comes from too many digital adjustments.

Ken Lee
28-Oct-2008, 05:00
Here are some example screen captures to illustrate one approach to handle the Epson 4990 scanning parameters: the Histogram Adjustment tool.

http://www.kenleegallery.com/images/scanning/Preview.jpg

Initial Preview - Here is a 5x7 negative, taken on Ilford FP4+, developed in Pyrocat HD. The subject is some pink roses, in overcast light, and the lens was a vintage design, chosen for its nice blur, especially at close magnification. The actual negative is rather soft in appearance: as the histogram shows, the tonal scale takes up only 40% of the possible range of the scanner. However, the default Preview settings have made the image appear quite harsh. It would be nice to make a "proof" of the image, to see what it really looks like.


http://www.kenleegallery.com/images/scanning/AdjustedHistogram.jpg

After Histogram Adjustments - We have changed the Output settings so that they go from 0 to 255. We have adjusted the Input settings so that the middle setting is 1.0, the left-hand is enough to render the black film edge as black, and the right-hand setting renders the high values naturally, without any clipping or un-natural loss of texture. (These roses were pink. We don't want them to look otherwise). The middle setting of 1.00 is a good rule of thumb: it tends to give us a linear or literal interpretation of the negative. In other words, no contrast adjustment is applied.


http://www.kenleegallery.com/images/scanning/Photoshop.jpg

Photoshop - In Photoshop, we have rotated the image to make it right-side-up. We have applied a non-destructive color adjustment Fill Layer (http://www.kenleegallery.com/bronze.html) according to taste.

We have managed to get a natural image that feels like light, right out of the scanner. Since the image was scanned in 16-bit monochrome, we have plenty of tonality available for any further adjustments.

Note that against a grey background, the image feels quite different than it looked against the black background provided by the scanner software. Depending on our choice of matting and frames, we may decide to adjust these roses further.

Note that the image has been cropped. That's a separate conversation, for another time :rolleyes:

dede95064
28-Oct-2008, 07:58
wow! thanks a bunch ken!

i'm surprised that the initial default value you were given for the middle value (gamma i believe) was 1.00. i always get 1.50. if i change it to 1.00 it becomes too dark so i tend to leave it on 1.50.

what about the tone curve viewer? i set both to normal but i never see any difference between that and the default settings. :confused:

thanks for your hard work ken!

Ken Lee
28-Oct-2008, 08:48
You're most welcome !

I plan to show a better example soon, using a negative whose scale is longer and more centered. Close-ups of flowers, using vintage lenses, are not exactly the standard test subjects we should be evaluating.

Even so, this negative (which was inadvertently over-developed), provides a good tutorial on how to avoid the 4990's default tendency to blow out the high values.

I myself, have abandoned the Tone Curve tool in favor of the Histogram Adjustment tool, since it can give us the digital equivalent of a contact print (or cold light head enlargement) on Number 2 Paper. I like the results better than what I can get with VueScan. One would expect that the native drivers would do best, after all.

I don't know if the 4990 drivers give us a gamma of 1.0 every time, but I set mine to that every time. I will look more closely, when I do the next "tutorial". I'm no expert, and I certainly welcome any insights and assistance.

mrladewig
28-Oct-2008, 15:22
I'm using Silverfast with the 4990. I've been using SE, but recently switched to Ai so that I could gain better control over color negative scans plus 16 bit for color positives. SE was OK for the color positives, but not so great for color negatives. I do like the control interface of Silverfast much better than EpsonScan though, so I've chosen to upgrade.

Like Ken found with Epson's Scan software, I've found that Silverfast tends to do the same thing to negative film in auto mode, blowing out useful information for no apparent reason. I use the expert mode and make my adjustments to the histogram for color and B&W, but I also use curve adjustments to get the scan about where I want it. Then in photoshop I can use luminosity masks to further tune the image.

Ken Lee
29-Oct-2008, 03:54
I stand corrected about the Gamma setting. Using 1.0 all the time, is a wrong notion.

You have to adjust to taste.

Daniel_Buck
29-Oct-2008, 09:12
I've found an easy way to find the white and black points for each scan. First I find the white points, I move the midpoint (gamma) over to the right, real close to the white point. Then I move the white point until the bright white highlights become darker (medium grey or so). For the black point, I move the midpoint (gamma) way over to the left, then move the black point until no pure black is seen.

Give it a try, it's probably harder to explain than it is to do. I'm a little disappointed by the Epson software (and silverfast software that I tried!) to correctly find the darkest dark point and the lightest light point. For me, thats really all I care about in the scan. No pure black pixels, and no pure white pixels, and a gamma that looks about right. If it's a very important scan, I'll scan like normal, then check my image for black and white point, and then adjust the settings and scan again if I feel it needs adjusting. I like a flat scan, and use no additional contrast curve while scanning, I leave the curve linear. Oh, and have my output levels set to 0 and 255 as shown on the previous page.

g.lancia
30-Oct-2008, 14:45
Crop histogram in RGB, then R, G, B. Output value 0-255. If you crop the histogram, you will optimize scanner's dynamic range. Not a lot of people realize that.

Ken Lee
30-Oct-2008, 16:25
Sorry, could you expand on that a little ? Your posting was so concise, I was not able to follow it.

john borrelli
10-Nov-2008, 09:25
I am something of a novice with my Epson 4990 and Epson scanner software.

I did recently come to the conclusion that I have been shooting my Velvia 50 landscapes a little too underexposed. The slides look great on the lightbox but for scanning, I found very slightly overexposed slides, to my eye, were easier to scan and resulting prints looked better.

I'm guessing my adjusted exposures were closer to 32 for Velvia 50 whereas I had been closer to 50 with my Pentax spot meter.

One helpful step for me was when I began to try to understand what the initial scanner histogram was trying to tell me.

For these slightly underexposed slides, the histograms tended to show a curve weighted toward the low values side. There also were not the "spikes" on the curves they tended to be rounded off.

Again, I'm just a novice at this but I would encourage any other novices reading this thread to take a close look at that initial scanner histogram(before any manipulation) and try to interpret what it is telling you because it will effect your print one way or the other.

I also wanted to take a moment to really thank Ken for his excellent contributions here.

Ken, great info and the way you have taken the time to post this information with the scans and actual image is incredible.

Ken Lee
10-Nov-2008, 11:44
John - You are most welcome ! It's fun to share.

This forum really is the best !

FYI, I have gathered up my info about the 4990 and placed it here (http://www.kenleegallery.com/html/tech/scanning.html)

venchka
13-Nov-2008, 08:17
Thanks Ken! I tried your fill layer technique last night. Early results are in. I like it!

Ken Lee
13-Nov-2008, 08:57
Super. Let's see a sample ! :)

venchka
13-Nov-2008, 09:03
Here you are. A work in progress.

Guadalupe River above Gruene, Texas

http://gallery.leica-users.org/d/142481-2/River-Warm-Tone.jpg

Thanks for all of your help. I'm struggling with Photoshop 6. Doing the best I can.

Ken Lee
13-Nov-2008, 09:17
Lovely.

We're all struggling with Photoshop :rolleyes:

Looks like 1 part Palladium, 2 parts Selenium, and... a little Coffee just for effect.

What I like is that the adjustment is merely a layer, and can be tossed, or ignored, or its opacity percentage adjusted according to taste. There's no limit to the kinds of tones you can get, and you get to keep your monochrome image too, with no loss of quality or digital artifacts. No special inks, no special nuttin'.

I also like the format, 5:8 I suspect. Golden Section. It lends a graceful feeling, all by itself.

venchka
13-Nov-2008, 09:42
Thank you Ken. You're very kind.

Believe it or not, I printed your pages on using the fill layer and typed in the color number from your example. However, my ancient version of Photoshop has different numbers for the Toyo Color Picker. I'm also hampered by Photoshop 6 only offering layers in 8-bit mode.

I'm going to a friend's house tonight and use his Epson 4990 and scanning the whole negative. I'll start over and see what happens.

Thanks again for freely sharing your knowledge with all of us.

AJSJones
13-Nov-2008, 11:12
Thank you all, Ken in particular, for the help and examples.

I have a question - do the adjustments recommended for the scan acquisition affect the hardware (and therefore the actual raw data captured by changing detector signal amplification or exposure times) or are they just transforming the same data in a "pre-processing" step before on gets to e.g. Photoshop? Another way of asking the question is "Can you achieve the same effect post scan if you just get the same raw 16bit info from the scanner?"

venchka
13-Nov-2008, 11:48
Huh? Que? The scanner output is in 16 bit mode. Assuming the scanner+software has 16 bit greyscale or, in my case, 16 bit RGB (a.k.a. 48 bit) output.

Ken Lee
13-Nov-2008, 11:52
<speculation>
I don't know where the adjustment(s) is/are made. The best results would be obtained if Epson controlled both exposure and post-processing by the scanner software.

Were there no way to adjust the behavior of the scanner before the actual scan (IE exposure), then the "Preview" step would do little for us, other than let us to crop the image.

What we see with the Epson software, is that when you adjust the cropping (for example, to include or exclude the blank film edge), a definite change is seen in the preview image. Depending on the dynamic range of the cropped image, the scanner seems to contract or expand its rendition, to best accommodate the target - and that is performed before we specify any corrections with the histogram or curves tool.

The negative is "analog"*: it has an unlimited number of tones. The scanner, even at 16-bit, is digital: it has a limited number of tonal steps available. So it makes sense to map the analog onto the digital, as skilfully as possible, so that no clipping is seen at either end of the scale, and so that the central portion is as linear as possible. (Sounds a lot like film, doesn't it ?)

By analogy, the scanner is like a lens with a wide circle of coverage: Even though we could use the edges of the image circle, we generally try to shoot through the center of the lens, where performance is best and distortion is least.
<end speculation>

*Grain is either on or off, but it is so fine, that film is effectively analog.

AJSJones
13-Nov-2008, 16:04
Thanks Ken,
I've always wondered how the preview step worked - and I'm still not sure.

<speculation>
Let's think about two examples (using transparencies to keep it simple). 1) A typical well exposed image with "average" range of well controlled highlights, midtones and no really dark shadows and 2) an underexposed image where the brightest part is only at the level of the midtones on the "good" one. (I expect we all have a couple like this!)

So in a "standard scan" (hardware fixed amplification and exposure time per pixel) We can imagine the scanner taking the range of analog signals between "noise" (let's say 0.1mV) and detector saturation (let's say 5V) and assigning the 16bit digital data to the range (we have a certain height difference between the darkest tone in the cellar to the brightest on the 5th floor roof and the A/D converter makes a staircase with 2^16 steps out of it)

In image 1) the scanner raw data range from e.g. 1 mV to 4V (12 stop dynamic range) and
in 2) the output may be from 0.1 mV to 128 mV (also let's say 12 stop ) - accuracy isn't critical here - just the concept.

Scanning possibilities
A)We could accommodate both these images in one scan and the good one would get the top 12 bits and the poor one the bottom 12 bits just fitting in the 16bit range. To bring the poor image to the "proper" exposure, Photoshop would use levels/curves etc to "correct" the black pt, white pt, tonalities etc
Or
B) we could speculate that the scanner software controls the A/D converter so each range is given its own 16 bits (from the cellar to the 3rd floor or the ground floor to the roof are separate staircases each with 2^16 steps) In each case, the image analysis program receiving the file will have black and white defined the same way but the analog noise will have been encoded differently and the tonalities will be very different when compared to the originals. This seems to be your "skillful mapping" done at the A/D conversion stage.
Or
C) The scanner acquires the same data all the time and the software preview just tells the post A/D conversion software to transform the file on the fly so it is saved during the scan with the desired ranges, curves etc and doesn't need to be manipulated (as much) when it gets to Photoshop.

In A we just use the default scan parameters for everything and adjust later.
In B we have the A/D conversion incorporate our processing wishes
In C we have the scanner software do on the fly what we would otherwise do ourselves in Photoshop to A after the scan

Given that this is all happening in 16 bit space (or 48 bit depending on your convention) and that we will drop down (usually to 8 bit) eventually for output, the only possible benefit the scanner controls can offer us is option B. A and C are not really different and work because the data have enough precision at 16 bit to yield satisfactory tonal gradients by the time they're in 8 bits, wherever they are converted.

If we could control the exposure that would be great at extracting detail from shadows while averaging out noise, but the scans would take longer - I'd be happy with that but I don't think it happens, because we'd notice darker images tasking longer to scan! Some software seems to be able to do multiple samples and average them - also suggesting a "fixed" scan mechanism.
<end speculation>

The question then becomes Does the scanner software intervene before or after the A/D converter? And the follow-up is (with 16 bit raw capture) is post scanner optimization going to give much worse 8bit images compared to having the scanner do it (however that happens)?

My scanner is down right now otherwise I'd do the comparison myself - set the "input to 0-64 and output to 0-255" and see how that compares to "input 0-255/output 0-255" and convert 0-64 into 0-255 in Photoshop, all in 16bit mode. Which is noisier after conversion to 8bit? Is it obvious?

BTW, I do it in PS after the fact because I recall trying it when I first got the scanner and not seeing much difference in that proposed comparison (but I didn't have as much experience back then :) )

Thanks for any thoughts

Andy

venchka
14-Nov-2008, 15:05
I posted a sample earlier and said that it was "a work in progress." LFPF member wclavey kindly allowed me to scan some of my recent negatives on his Epson 4990. I am finally beginning to see the light. The scans are getting better. Westley installed the EpsonScan software for me. With Ken's tutorial in hand, I was able to make what I consider to be decent scans. Now it's up to me to get better negatives to scan. And figure out this digital toning thing. :rolleyes: :eek: :cool:

One more time. The whole picture. Guadalupe River above Gruene, Texas.

http://gallery.leica-users.org/d/142610-2/Gruene+Hall-7.jpg

Ken Lee
14-Nov-2008, 17:34
Dy no mite !

venchka
15-Nov-2008, 09:46
Thanks Ken! I couldn't have done it without all the help I've gotten from you and Westley (wclavey of this forum) and all the folks on this and other forums. Everyone has shared their knowledge freely. Hopefully I can return the favor to others in the future.

Thank you Large Format Photography Forum!