Workflow for Color Negative scanning with Elite 5400 - comments?

Looking at my log I see I started on this June 05. Here it is June 12

and I'm still experimenting. This is my first real attempt at

scanning color negatives. Had a few heady false-starts, then crashed.

Back to experimenting. I do think I'm making progress, but who knows

what I'll be thinking tomorrow. Anyway:

 

I'm attempting to wrestle a decent color negative scan (Fuji Reala

100) out of my Minolta Scan Elite 5400. I've been reading and

experimenting.

 

I would prefer to use Vuescan, for various reasons. However, with a

pure Vuescan workflow, I have one major stumbling block: the quality

and completeness of the infrared cleaning. I believe I've come up

with a solution, involving the production of files through Minolta

Scan Utility that are suitable to be used as Vuescan Raw Files. First

though, some preamble:

 

I've tried Erik de Goederen's (Scanhancer) pdf tutorial on the

subject of scanning color negative film with unconventional use of

Minolta Scan Utility, which you can find here:

 

http://www.scanhancer.com/index.php?art=17&men=17

 

I've also researched Vuescan's advanced workflow suggestions, in

greater depth than the cursory try I gave it before. One thing I

learned, searching the archives of the Usenet scanner forum, locking

film base color is a color balance function. Vuescan's locking

exposure IS a good thing to do at the initial scanning stage. I think

it essential for subsequent locking of film base color. However,

setting film base color at this stage does not effect the Vuescan Raw

File color balance.

 

What can effect the raw file balance is hardware level adjustment of

the red, green and blue exposure. In Vuescan, setting your media

to "color negative" will do this. By design, it raises the exposure

level of the green and blue channels significantly, to rough

compensate for color negative film's orange mask. In MInolta Scan

Utility, moving the red, green and blue Exposure control sliders by

different amounts, will also affect color balance.

 

As evidence that these adjustments to individual color exposure are

hardware level, the scan times are affected, both in Vuescan and MSU.

 

One thing I had never understood till now was how to use Vuescan's

advanced workflow locking of film base color when following a scan-

from-disk workflow. My understanding through research on the net,

borne out by experiment:

 

1. When scanning Vuescan Raw Files (or any file to be used as a

Vuescan Raw File)from one roll, with whatever software, lock your

exposure, at the outset.

 

2. While doing these scans, with the exposure locked, also scan a

blank frame from the leader.

 

3. When doing subsequent Vuescan scan-from-disk, preview the blank

frame, crop to exclude the film holder portion around the edge,

preview again, tick lock film base color, in the Input tab. This will

add 3 red, green and blue film base values to the Color tab.

 

Finally, getting back to my problem with Vuescan cleaning. I got to

thinking, why don't I:

 

1. Scan a film frame with Vuescan, in a traditional color negative

workflow. Go through the complete advanced workflow steps first, to

be on the safe side, and output a 64 bit Vuescan Raw File.

 

2. Pre-scan the same film frame with Minolta Scan Utility, as a color

slide, with ICE and Grain Dissolver on, and auto exposure off. Set

output to be 16 bit linear.

 

3. Switch over to the MSU Exposure control tab.

 

4. Open the Vuescan Raw File from the previous step, in Photoshop

(CS, preferably, for it's realtime histogram).

 

5. Meanwhile, in the MSU Exposure control tab, move the master, red,

green and blue sliders to get the histogram displays roughly per the

Vuescan Raw File's histogram display in Photoshop. (Note, what you

see is NOT what you get at this stage. You are seeing what 16bit

output would be. But since the MSU output will be 16 bit linear, the

result will be much more squished to the left.) Anyway, follow

through, do the scan.

 

6. Open the MSU output in Photoshop, and compare combined, red, green

and blue histogram statistics. The MSU 16 bit linear file will likely

be a lot darker. Note also the porportion of red to green to blue.

 

7. In MSU, adjust the Exposure sliders (invariably to the right),

aiming to match the Vuescan Raw File histogram levels and

proportions, and rescan.

 

I had to repeat step 7 about 4 times to get a MSU 16 bit linear file

that was extremely close to the Vuescan Raw file. Along the way, the

scan times increased, though not quite as much as Vuescan color

negative scan.

 

Following are some of my Minota Scan Utility settings, for your

reference:

 

- Scan as a Color Slide

 

- 16 bit linear tiff output

 

- Exposure: Manual

 

- Auto Focus (Or manual? Haven't made my mind up. Manual focus will

not work in Custom Wizard.)

 

- No crop or rotation

 

- ICE/GD on

 

Exposure Control Settings

 

master: +0.4

 

red: -0.4

 

green: +0.9

 

blue: +1.4

 

This yields a file very similar to Vuescan Raw File, and useable

accordingly, with the benefit of improved ICE cleaning, incorporated.

 

I would much appreciate if any other Elite 5400 users have give this

a try, and let me know how it works for you.

Or if you're already doing something like this. Or you've tried it, it doesn't work, and why.

Mendel, like you, I've been struggling to get decent negative scans out of my Elite 5400, given the terrible highlight and shadow clipping its proprietary software produces.

 

After countless hours of online research, trial and error and frustration, I've come up with a method that I believe gives the best results --and using only the Minolta Scan Utility and Photoshop CS, as VueScan simply lacks a good implementation of ICE.

 

 

AT THE MINOLTA SCANNING UTILITY:

 

1. Scan as color positive, 16bit (NOT linear), autofocus at scan

 

2. At the Exposure Control tab, set:

 

Master=0

 

R=0

 

G=1.2

 

B=2.0

 

-This helps to balance the histogram and eliminate most of the negative orange mask. Some difficult images may require different settings, but these cover most situations.

 

 

IN PHOTOSHOP:

 

1. Load TIF file

 

2. Invert

 

3. Assign profile "Minolta DiMAGE Scan Elite5400" (installed along with the Minolta software)

 

4. Convert to either ProPhoto or Adobe1998

-I personally prefer to work on 16bit all the time, so ProPhoto is my choice for colorspace given its fantastic wide gamut; but most people prefer to work in Adobe1998.

 

5. Get rid of the remaining colorcast: Image>Adjustments>MatchColor>Neutralize

-I've found that with night shots or very dark images this step is not necessary, and in fact it can be counterproductive. However for most daylight photos it works like a charm at correcting the histogram.

 

6. Create a Curves Adjustment layer. Open its Options, under Algorithms select "Enhance Per Channel Contrast", check "Snap Neutral Midtones", and on "Target Colors & Clipping" select 0.01 both for shadows and highlights.

 

7. At this point the image still looks too bright, so create a second Curves Adjustment layer on top of the previous one, and use it to correct the overall gamma --a simple pull down on RGB (say Input around 200, Output around 180) is usually enough.

 

 

Having done all this, you should now have a decent file to begin working on, with no clipping and neutrally balanced colors. At this point what I do is create a Hue/Saturation Adjustment layer and use it to increase saturation, and create a Levels Adjustment layer and use it to clip the shadows and highlights as necessary (for contrast.)

 

Hope this helps.

 

Best,

 

Marcos Rodriguez / www.aukeramen.com

Nice one Marcos. I have an Epson but i have saved your steps as an action and is very useful. Thank you!

Thanks for the coments and workflow, Marcos. I have dabbled with MSU to Photoshop workflow, but I'm still going to try work atleast the finishing steps in Vuescan. I agree, Vuescan cleaning is not good enough, that is why I'm attempting to make Vuescan Raw Files thru MSU, to incorporate true ICE.

 

The way MSU exposure controls seem to be working at hardware level (as evidenced by variations to the scanning time) can allow you to optimize the color balance at time of scan, before any software deals with it. At least, I think so. Also, I'm still just getting my head around what can be done by scanning a leader, how it can be used to calibrate things. Some more notes:

 

Last night I tried Vuescan scan-from-disk with the dozen or so "quasi" Vuescan Raw Files I had produced through MSU, per my workflow above. Using the color negative setting, I first locked film base color, after previewing the leader frame I had scanned.

 

Based on the finished gamma tiffs I got, I would say these MSU produced raw files are close, but do not appear to be what Vuescan would normally expect to see. If I set color balance to "white balance", there is a nasty cyan cast. D.O. for neutral, or manual with 1/1/1 values.

 

Setting color balance to manual, and right-clicking on supposed neutral grey, I was able to get a half-decent color balance, but it required a heavy adjustment of the red/green/blue neutral settings, and a lot of back forth between various frames, hunting for true neutral greys.

 

This morning, out of curiosity, I tried another Vuescan scan-from-disk as "image", of the MSU generated leader frame raw file. To my suprise it looked exactly the same tone as the leader itself. Then I realized I still had the heavily adjusted manual color balance set. When I set the 3 manual settings back to 1, the color got a lot closer to clear, but there was a definite cast still.

 

My sleep-deprived brain cells could not sort this out, but I suspect it is not coincidence. Flipping from negative back to postive, while retaining the manual adjustment set in manual scan-from-disk, seems to have brought me full circle. Hmmm.

 

Anyway, still in Vuescan, with the manual balance set back to 1/1/1, I moved the mouse around the preview of the leader raw scan, observing the (quite uniform) r/g/b values. I fed perhaps 1/2 dozen of these readings into a spreadsheet, and averaged them, just for an exercise. In the un-reversed scan-from-disk with this leader frame raw file, the red was still too high, and the green a little high. Taking measurements like this, I think I can further optimize the MSU exposure control settings, and look at raising all 3 channels, to be just under 255.

 

Having fired up the scanner at least 20 minutes earlier, I restarted MSU and and adjusted the exposure control sliders in a rough attempt at this, to decrease the red exposure a moderate amount, and a slight decrease in green as well. I did a quick low res. scan of a regular frame, a daylight scene, outputting another 16 bit linear, and tried Vuescan scan-from-disk with this. Much improved color balance. Using Vuescan's white balance, or neutral, the result is almost cast free. Still, I think I have a bit of tuning to do.

 

Tonight, I'll scan the leader with MSU a few times, finetuning the exposure settings, and assess by doing scan-from-disk as "image" on the result. My object is to get the raw file as neutral as possible, at the hardware stage. It may be impossible to dial it in exactly, since the MSU exposure sliders only allow 20 increments (from -1 to +1 in increments of one tenth). I think this scan of the leader is the key to producing good Vuescan Raw Files through MSU, and for minor subsequent tweaking, within Vuescan.

Marcos, after playing around with MSU's exposure levels some more, I finally tried your workflow, step by step. The result is very good! The balance looks very natural, and the histograms look smooth and unclipped. Thanks for sharing.

My pleasure, Mendel. Glad to know you found my method useful. As I said, it took me endless hours of tweaking, trials and research on the Internet to finally arrive to this approach, and so far I have not found anything that can top it. Best of all, it does not involve spending even more money on third party software like VueScan or SilverFast.

 

Too bad Minolta bundled a fine piece of equipment like the 5400 with such lousy software, otherwise the Elite 5400 would have been a real killer. Now I hear that the new 5400 II is way faster than the 5400 and comes with a much improved algorithm to process negatives; I haven't tested it myself, but if true it only makes me madder: Just months after having provided Minolta with a bunch of my hard-earned dollars for a scanner that I might not have bought had I known of its software flaws, they come up with a much improved version and I'm stuck with this "beta", so to speak.

 

Unless Minolta comes up with a major software update for the 5400 --and pretty soon-- I am going to boycott them for good. In fact, one reason I didn't even consider the Konica line of digital cameras was precisely this. Now I'm shopping for the Canon 350D, and saying goodbye to film --of course, I still have my negative archive, and it needs its scanner...

 

Anyway, glad to have been of help.

 

Regards,

 

Marcos / www.aukeramen.com

Excellent thread, with scan acquiring workflow by Bart van der Wolf here (usenet):

 

http://groups.google.ca/groups?hl=en&lr=&ie=UTF-8&group=comp.periphs.scanners

 

and search for:

 

Scan workflow for negatives using Minolta 5400

Marcos, regarding MSU wish list, a few things I'd like:

 

1. Ability specify off-center autofocus focus point.

 

2. Readout of focus result, and the ability to key in a focus setting.

 

3. Ability to decouple Grain Dissolver from ICE. (Though, I think cleaning quality would take a hit. Still would be nice to have the option, to see for oneself.)

 

Here's my (rough) take on the thread I linked, with a little tweaking. The main tweak, I DO think you can scan full frame and not have the auto exposure screwed up by the frame itself, using the Auto Exposure Crop function (not sure if I've got the language exact, it's described on page 39 of software manual). Tried it this a.m., very promising.

 

The object is to optimize the scan itself, as much as possible. To extract max info, and negate the mask. The approach is very objective; I can't fault it.

 

Workflow for scanning color negative film with Elite 5400

 

In MSU:

 

Settings:

 

* Scan as color slide type

 

* Auto exposure on (to get base exposure, adjustment will be made)

 

* 16 bit linear output (or 16 bit?)

 

* Manual focus

 

Steps:

 

1. Scan a frame containing just leader.

 

2. In PreScan tab, set crop to maximum. (assuming you want a full capture, safest)

 

3. Set Auto Exposure crop to be a little inside the actual image. (hold shift key down while dragging crop edge) (to ensure exposure isn't influence by holder)

 

4. Depress the Auto Exposure crop button. (to use the AE crop from previous step). Do another Prescan. (This, coupled with step 3, will for sure exclude the holder from the exposure calc)

 

5. In Exposure tab, set master slider to 2. (max. brightness)

 

6. Move mouse around the image and note the red, green and blue values. Should look just like the leader: orange in color.

 

Move the red, green and blue sliders off of the default 0 values, to bring the values as close as possible to being equal, and also to bring them all as close to, but under, value of 255. After each slider(s) adjustment, hit the exposure refresh button. It does not do a second prescan, just recalculates with the software. Probably 245 is a safe highpoint target, since there will be some errant readings you won't notice, which may otherwise blow-out to 255.

 

If you still have clipping to 255 on any of the channels regardless of pulling the individual color sliders, move the master slider down a bit. It really doesn't matter, the effect is the same. The object is just get all 3 colors as equal as possible, and comfortably close to 255. Move the mouse around, watch the values carefully.

 

When sastisfied, you can save these settings (with a name), and also apply them to all frames (with the buttons across the top), if you like.

 

You have now neutralized the film's orange mask, and the image (just leader) should look plain white, if you've done things right.

 

7. Switch back to Prescan tab and do a careful manual focus. Choose a light, smooth tone area, somewhere near middle or a bit off center. Click ok, and a prescan will start.

 

8. Switch back to Exposure tab to verify the red, green and blue values are still near balanced, and not clipping to 255.

 

9. Scan. Note, this is a scan of a blank leader, for reference. Perhaps make it's name's numeric suffix "-00".

 

10. To scan subsequent "real" frames, repeat the above steps, but omit step 6, the exposure adjustment process. The master/red/green/blue auto exposure adjustment values set on the leader, which apply an adjustment to MSU's autoexposure, will eliminate the orange mask effectively, on all frames of the same film, and will probably be very close for any other rolls of the same film type, though you should check. Also, verify these values have indeed been carried over to all frames. To repeat, in the Exposure tab, the "apply to all frames" button should accomplish this.

 

Also, I believe the auto exposure crop window will be remembered once you've used it once on a frame, so the process of setting that and doing another prescan can probably be avoided. Just press shift in the preview tab, to verify AE crop is still a bit inside the actual image area, and verify the Auto Exposure crop button was depressed, PRIOR to prescan.

 

 

Ok, I've tried to tidy up my method, and added the use of Photoshop's level dialogue, to increase accuracy. Hopefully someone's still reading. Any comments greatly appreciated.

 

METHOD OF CALIBRATING ELITE 5400 FOR COLOR NEGATIVE SCANNING

 

OBJECTIVE:

To optimize the scanner's response to a roll of color negatives. To this end, the goal is to set the scanner's exposure settings so that each of the three color channels is recorded with maximum brightness values as close as practically possible, but not at, pure white (255), when scanning the film's leader.

 

Accomplishing this objective will neutralize cast caused by the film base color, and ensure maximum extraction of detail from the darker areas of the film (the highlights in the inverted, normal image).

 

SUMMARY OF THE METHOD:

 

In Minolta Scan Utility, prescan a cropped frame of the film's leader, as a color slide, with auto exposure on. Then, adjust the exposure controls to equalize the red, green, and blue channels as closely as possible, and to bring the average reading of all 3 channels fairly close to pure white, say a starting value of 225.

 

Make a scan with these settings.

 

Open a levels dialogue on the resulting file in Photoshop. Set clipping to be 0.0 per channel, and note what Photoshop intends to clip from the white end of each color channel. Ideally, each channel should be clipping an equal value, very slightly below the highest value of 255. Significantly lower white clipping numbers indicate exposure was too low for that channel. Clipping value of 255 (no clipping), indicates the exposure was too high for that channel.

 

If low or no clipping of white point is the case, repeat the scanning process in MSU, adjusting exposure to suit, and re-evaluate in Photoshop.

 

After setting the exposure bias satisfactorily, and for the remainder of the roll, continue to scan with Auto Exposure on, with these bias values. If you want to scan full frame, use AE exposure cropping window, as described on page 39 of software manual, to exclude the frame from the autoexposure calculation.

 

DETAILED METHOD:

 

Minolta Scan Utility operations:

 

Settings:

 

media: Color film

 

auto expose for slides: yes (will be adjusted, through adjustment in Exposure Control tab)

 

focus: manual (preferred, for greater accuracy and control)

 

output: 16bit tiff (not linear, for now. Once the exposure values are established though, it is ok to output 16 bit linear if you like.)

 

dpi: 5400 (For maximum accuracy of calculation. With ICE/GD off, it's still quite quick.)

 

ICE/Grain Dissolver: off (just for efficiency, when you have gone through all the steps and determined optimum exposure, repeat the process, with ICE and GD. Some further adjustment may be needed)

 

crop: crop to completely exclude the frame (For now. Full frame can be captured, without skewing the autoexposure calculation, with the use of MSU's AE crop area, per page 39 of software manual)

 

Steps:

 

1. In MSU, Prescan leader. Crop to exclude frame. Manually focus (forces second prescan)

 

2. Go to Exposure tab. Move mouse around the image and note the red, green and blue values.

 

3. Move the sliders to get the values as close to equal as possible, and around a value of 225. Be sure to press the exposure refresh button after each adjustment.

 

(The object is to equalize the scanner's response to the three channels, and to get the values as high as possible without clipping. Since moving the mouse over the image will likely not find the highest values, 225 is a reasonable, somewhat arbitrary, starting target)

 

4. Scan.

 

5. Open the resulting file in Photoshop, and open a levels dialogue. Click the options button, set white and black point clip both to 0.0, per channel (no "snap neutral midtones") and click ok.

 

6. Back in the main window of the levels dialogue, review what Photoshop proposes to clip from each of the individual red, green and blue channels:

 

If any of the channels shows a white point clipping value of 255, the scanner exposure you set for that channel, in step 3, was too high.

 

If any of the channels shows a white point clipping value quite a bit below 255, the scanner exposure you set for that channel, in step 3, was too low.

 

Ideally, each of the channels would show white point clipping of 254.

 

Exit the command without completing the process.

 

7. Repeat the process from the beginning, adjusting the exposure settings you make in step 3 in an attempt to achieve the ideal white clipping values from step 6.

 

(Of course it is impossible to get exactly 254/254/254, but try to get as close as possible without going over)

Marcos, it's not clear in your your first posting, are you using manual or auto exposure in MSU? The exposure controls will work with both.

Hi Mendel, sorry for not replying sooner, I hadn't checked this thread.

 

Regarding your question. I've tried both auto and manual exposure, and find the results vary little. Having said that, perhaps on the overall auto gives better results.

 

Also, I think a lot depends on the original image and the emulsion type --this as some film brands seem to have a heavier orange mask than others. For instance, if you have a well-exposed frame to begin with, the results are not likely to vary much with or without auto exposure; however, with heavily under- or over-exposed frames, auto-exposure seems to yield somewhat better results.

 

As for 16bit linear, I never use it for negatives as it crams all the histogram on one corner, thus greatly reducing the file's color information and quality. Normal 16bit seems to be the way to go with negatives.

 

Best,

 

Marcos / www.aukeramen.com

Thanks for responding, Marco. I'm continuing to experiment. One of your suggestions, to use adjustment layers wherever possible, I have really taken to heart. You did not mention adjustment layer for the inversion step, I see you can do that in an adj. layer as well. Doesn't hurt.

 

Did you read the Usenet thread? Here is the direct link:

 

http://groups-beta.google.com/group/comp.periphs.scanners/browse_thread/thread/42c3b2fa6068fc62/

 

Very interesting. Bart van der Wolf recommends autoexposure, with exposure control overrides. I followed workflow to the letter, with fairly good results.

 

Yesterday though, I started experimenting with complete manual exposure. Per Bart's workflow, my first step was to scan the leader, as a slide. I adjust the sliders to show r/g/b values all around target of 235 in msu, then output 16bit scan. Open in PS, see I have any pixels at 255, or none above 245. If either is the case: back to msu, adjust exposure again. Then I lock on these settings for scan of rest of roll.

 

Note, this is esentially the same as Vuescan advanced workflow for raw file output.

 

Ideally, you should do this dial-in for each roll, but perhaps if you're shooting all the same film, and taking to the same processor, once is enough. Something to verify.

 

So that is where I stand now. Regarding output format, I still wonder if 16 bit linear is not better. You can convert thru use of the supplied Minolta posi-linear profile. One benefit, it works quite well as a Vuescan raw file, and I really would like to have that option.

 

Last night I output both a 16 bit, and a 16 bit linear of the same frame. The 16 bit linear, converted thru the posi-linear profile. Then I ran the same levels/curve on both. The 16bit linear had better saturation, and cleaner whites, to my eye. The one thing I didn't do though, was convert the 16 bit version thru the regular minolta profile (non-linear), so that could explain the difference.

 

One thing I found with the pure manual exposure, my times per scan dropped dramatically, from 23~25 minutes per, to 14 minutes (consistant) per. This is with all 3 channels in 16 bit output having pixels above 245, but none by 255. Of course, when I then ouptut 16 bit linear, these numbers drop, but I do not alter the settings.

 

I'm wondering, how are your times?

Hi again, Mendel,

 

I've been wondering too if 16bit linear is the way to go. Essentially, I think it boils down to which software can be trusted the most to make the best conversion from the scanner's raw data: Minolta's Dimage or Photoshop in combination with the posi-linear profile.

 

After reading what you've written so far I've done some extra experimenting, and indeed, in most exposures, colors look more saturated using 16bit linear (with its corresponding posi-linear profile applied in Photoshop.)

 

One definite advantage of using linear is the markedly shorter scanning times; I was able to reduce a 2700 dpi scan (with ICE on) to a "mere" 5 minutes, down from the 18-20 minutes it used to take.

 

About adjustment layers: I did not suggest using one for the inversion step as it would have messed with the applying of the profile, as this has to be done after inverting (otherwise you get a gamma that is way off.)

 

However with 16bit linear this problem is solved --as the profile is applied as a first step-- so one can also use an adjustment layer for inverting the image.

 

Another thing I do is convert the background to a layer (simply double click on the background at the layers palette), that way I can use the "hide" option while cropping, thus keeping the entire original image.

 

I'll keep experimenting for a couple of days and will get back to you.

 

Best regards,

 

Marcos

Carlos, a few more comments:

 

1. You've mentioned a few times that you are scanning at 2700 dpi. I've found I save little-or-nothing in scan time doing that. If your prime motivation is to save disk space, why not scan at 5400 dpi, and downsample later, when the disk space crunch comes, bicubic downsample in Photoshop? Also, this is one of those irreversable decisions. Perhaps stay at 5400, and hope storage technology will catch up with you?

 

2. Not sure if you're doing it this way, or not, but I would recommend you save the as-scanned file, and do your Photoshop work on a copy. The steps you take in Photoshop are easily documented, and reproducable. If, again, space is a consideration, convert your Photoshop output to jpeg. Pls excuse if this is an obvious measure.

 

3. I do believe I've got the (totally manual) Elite 5400 exposure dialed-in now, atleast for Reala. The workflow of setting the exposure in prescan of leader to have all 3 channels just under 255, is working very well. 16 bit linears work up nicely in photoshop.

 

4. Regarding the first curve adjustment that you describe in your workflow, I think you can get the same result with levels. At any rate, with the levels adjustment, I've found (following your levels guideline, as a base) that increasing the white point clip from .01 to .02 does a significant amount of beneficial "aligning" of the 3 channels. Also, simply setting the mid slider in rgb to (say) 0.9 negates the need for a second, curve adjustment of overall gamma.

 

5. Now that you brought it to my attention, I think I'm getting faster scans outputting 16 bit linear, vs 16 bit. I'd never noticed or thought about it, up to now. Will have to do some testing. At any rate, I'm partial to 16 bit linear, even though I don't have a clear understanding of it.

 

Anyway, I'll be away on vacation for 2 weeks commencing sunday, so will be sort-of out of the loop. Will check in on Photo.net once in while, if and when I get the chance, but I don't want to push it.

Hi Mendel,

 

Thanks for your continuing update. I spent all afternoon doing tests --mainly batch-scanning as 16bit linear at 1350dpi and doing bulk, quick-fixing in Photoshop, to see if the workflow holds for most images. So far it does, although I want to re-check the results tomorrow and see how everything looks after a good night's sleep.

 

Regarding your comments:

 

1. I also work on copies, leaving the original "raw" scan as a backup should the need arise to re-process again. When I'm done scanning, I also compress the final image to high-quality JPEG, to save space --as I set Photoshop's preferences to store a detailed history log as metadata on the file, I can always re-trace the steps that led to a good image.

 

Also, I'm currently scanning the entire roll at 1350dpi (which only takes a little over 2 minutes per frame using 16Bit Linear), just to see what images came out fine and make a contact sheet; then I re-scan the best frames at higher resolution for printing and archiving.

 

2. I've also noticed that scanning either at 2700 or 5400 takes practically the same amount of time, so for the "keeper" images (that is, those going into my portfolio) I will probably scan at top resolution, even though I'll only be able to fit some three files on a CD.

 

3. About the first curve adjustment, since it is an automatic step (using Photoshop's auto-color) it is indeed the same to use curves or levels. However I prefer curves for the second adjustment layer (whose blending, by the way, I set to "luminosity" in order to avoid introducing any color cast) as I think it does a finer job than levels --it is my impression that levels tends to darken the shadows too much, whereas curves does a finer job at preserving detail there.

 

And you are quite right that setting the white clipping point to 0.02 (and sometimes even higher) instead of 0.01 does a good job at aligning the three color channels when these differ in size or placement significantly --on images taken under generally neutral lightning conditions and good exposure, however, the effect is not as dramatic.

 

I set a clipping point of only 0.01 for both shadows and highlights merely as a starting point, with the intention of increasing it further along the way if need be.

 

I currently have the second curves adjustment on an action (which I call gamma), with the following settings: Input 217, Output 197, blending as "Luminosity". I think it provides a good starting point from which to work on later.

 

4. About 16bit Linear versus 16bit. I find the two take practically the same time to scan, providing that you don't make any manual corrections on the Exposure Control; if you do that, then there is no telling how long the latter will take to scan --as I've said on other threads, I usually get scanning times of up to 20 minutes per frame when fiddling with the Exposure Control. (Right now, scanning 16bit Linear at full resolution, I'm getting a little over 5 minutes per frame, including auto-focus time.)

 

I'll probably spend much of the weekend scanning and doing yet more tests. I really need to nail the workflow, as I want to scan my entire negative archive --this in preparation for my planned move out of film and into the dSLRs. I'm saving to get my hands on the new Canon EOS 350D (a.k.a. the Digital Rebel XT), which seems to be a good choice for my budget range.

 

In any event, I'll post any new developments here, and in the meantime have a great time in your upcoming vacation.

 

Best,

 

Marcos Rodriguez

Hi Marcos,

 

Regarding increasing the clip beyond .01, I'm seeing too it depends on the image. If there's a persistant cast, the extra .01 or .02 will make a difference. On other images, cast is minimal/gone by .01.

 

Thanks the tip regarding luminosity mode with curves. I really am a novice at Photoshop. I'll experiment with this. Well, probably 2 weeks from now, with pending vacation.

 

It's seems there is always something to bug you when scanning. By that I mean, if you resolve one issue, something else will come along that seems to need improvement. For me, having dialed in the exposure issue (I think/hope), I turned my attention to focus. Some points:

 

1. I suspect the scanner is not able to "return" to a previous focus setting, with 100% accuracy. For example, if you gang focus frames 1 through 6, in prescan, and then tell the scan all the frames, it has to adjust the focus from frame to frame. This adjustment is a little sloppy, I think. Also, all the extra back-and-forth travel of the holder can't help matters. Accordingly, I'm elected to do careful manual focus, frame by frame, followed immediately by scan of the that frame. Followed in turn by verifying the focus of the output in Photoshop.

 

2. I think I prefer manual focus with the case front knob. There is a feeling of direct connection that seems absent when clicking the slider arrows.

 

3. A good focus point seems to be about 1/3 distance from center to either upper corner, in a smooth, light tone. This gives me the most consitant corner to corner sharpness, with little or no apparent softening of the center.

 

4. It's kind of interesting to observe, after you've carefully dialed in the focus: tapping the holder gently will make the black bar "shiver" away from the max. extent white bar, then return to full length.

Hi again Marcos,

 

A digital Single Lens Reflex camera is something I'm contemplating also. It's a little depressing looking at the massive grain on 5400 dpi scans of Reala. And scanning is getting WAY too time consuming and finicky.

 

The Canon 20D seems to be the "Honda Accord" choice, and likely this is what I will (would) get. It's not TOO much higher in price than the dig. rebel, and I think worth the extra change. Will need to read up more.

 

It's more expensive counterpart the 1D MkII is tempting, with the 1.3 crop which opens up your lens choices somewhat, more solid build, more exposure controls (spot metering, which the 20D doesn't have, I think). But then I come down to earth, look at the sticker price, and consider that it's the same pixel dimensions as the 20D.

 

The 1Ds MKII? Would love to have one. With the some L series lenses... Oh well, if I win the lottery.

 

Some ideas, in order from most sensible, to most desired:

 

Canon 20D with Canon EF 17-40mm f/4.0 L USM*

 

* This lens seems a fairly good comprimise between quality and price. The are some trade-offs, compared to the following lens, max. aperture, for one.

 

Canon 20D with Canon EF 16-35mm f/2.8 L USM

 

Canon 1D MKII with Canon EF 24-70mm f/2.8 L USM

 

Canon 1Ds MKII with Canon EF 24-70mm f/2.8 L USM, and?

 

This site has extensive canon lens summaries:

 

http://www.the-digital-picture.com/Default.aspx

 

Also, google "fred miranda <lens name here>", for lot's of feedback.

Marcos,

 

Nother breakthrough: I've been using sRGB as my working space. And when converting the 16bit linear files, first assigned the Minolta posi linear profile, then convert to sRGB.

 

The results were often cyan cast, and the cast would persist with autolevels, unless I tweaked the red, green and blue.

 

Just now, was working on a particularly difficult to balance image. Our granson had a bright red hooded top on, in bright sunlight. Autocolor was producing a murky cyan result. And his red top appeared to be "bleading" on some edges.

 

On a hunch, I shut it down, set my working space to Adobe RGB, and took it from the beginning. And of course, the step after posi linear assignment was to convert to Adobe RGB.

 

Now, autocolor worked flawlessly, it felt like I was looking through a window into the scene. Major improvement. My son's telling me now, I should be calibrating my monitor, to really be seeing the true results. Oh well, I'll keep on plugging.

 

Of course, one of your first bit's of advice was regarding color space. Takes a while for things to sink in with me :0

Regarding step 4 of your orig. workflow, and your mention of ProPhoto, I had a look at this:

 

http://www.luminous-landscape.com/tutorials/prophoto-rgb.shtml

 

Oh well, about 4 hours to take-off, I better take a break.

Hi Mendel,

 

Just wanted to let you know that I'll be going out of town for a week, on vacation, and should be back around July 21st.

 

Very briefly, though: I've been experimenting and tweaking with a Photoshop action I found on the Internet, and have obtained *great* results with it (scanning as 16bit linear and applying the corresponding profile while opening in Photoshop.)

 

I'll post more detailed instructions on how to use the action (as well as further comments regarding digital cameras, etc) when I'm back. In the meantime, I thought you might want to start playing with the action, so I'm attaching it to this post.

 

As you'll see it's all pretty self-explanatory and automated; I only included a couple of stops, one for adjusting the cropping (assuming an image of 1350dpi,) and another one near the end to adjust curves for gamma correction.

 

Hope you find it useful and look forward to your comments.

 

A final note: I would NOT reccomend using this action on files scanned at a resolution higher than 1350dpi, as the action contains numerous steps that require tons of RAM.

Hi Marcos, gave it a try. Not bad results, though I think I'm satisfied with the levels/curves/saturation workflow we've been discussing. For me the main hurdle has been dialing in the exposure in msu. I've found with that optimized, the rest is always doable in photoshop.

 

Well I got a 20d with a 24-70 f2.8 canon lens while on vacation. It really put things in perspective. When we got home I put the color negatives in high gear, outputting 16 bit linear, then processing them in ps. So I think I'm done. The scanning times for color negs are a killer! I hate the thought of shooting any more color negatives. Also, when you see the crisp detail from an 8 megapixel slr... Funny, they're not really that far apart, but the grain of films scans really is distracting.

 

I'm getting back into the scanning of slides now, about 2000 worth, taken over the last 30 years or so. I think I will concentrate on getting the best 16 bit linear output I can achieve, with accompanying photoshopped finish gamma images for viewing, and follow up with cleaning later, when time permits.

 

One thing I've found lately, I can create an icc profile similar the the Minolta supplied posi linear profile, but with Adobe rgb (or whatever) as a starting point, but setting the gamma to 1.0. This can in turn be used th bring the 16 bit linear file's gamma up to normal, but the color balance is different (a little warmer) than the posilinear route.

 

Oh well, keep on plugging.

Did you try with vuescan under the "Filter" tab the "restore color" and "restore fading" ?

it works quite well with FUJI Reala and FUJI SUPERIA.

 

Brgds -Gilles

I've experimented with the restore color functions a bit, though not with color neg's up to now. I'll give it a try. Though as I say, I think my days with ANY kind of film are numbered, due to my new toy (20d), and I'm pretty happy with my Reala results now.

 

Also, I'm kind of browned-off by Vuescan: it's frustratingly *off* white balance, and poor cleaning. Plus it's toasted my work so many times, so for now I'm trying to work within Photoshop.

Hi Mendel, sorry for taking this long to reply. I got back from Puerto Vallarta a few days ago, but between the accumulated work load and an ear infection I caught for swimming in the sea, I haven't had much time for Photoshoping.

 

I completely agree that scanning times are a killer, which is the reason I try to avoid any tweaking with the MSU and, instead, do all the work in Photoshop. So far, scanning as 16bit linear (without touching the exposure settings in the Minolta software) I'm able to get acceptable times: about 2 and a half minutes per frame at 1350dpi, with ICE on.

 

I used to worry a lot about getting the best possible exposure in the Minolta software, but after many tests it looks to me that this is not very important, as the final scan contains all the data and then some --which gives Photoshop plenty to work with (this when scanning as positive, of course.)

 

Also, Photoshop's "Neutralize" function seems to be better than the MSU at compensating for the negative's orange mask. Moreover, considering that any tweaking with the MSU can double or triple scanning times, this approach seems to be the most practical.

 

I've finished polishing the action I mentioned in my last posting, and I'm quite satisfied with its results --I'm attaching it at the end of this posting. I added a small routine towards the end, using a Selective Color adjustment layer to counteract the remaining cyan cast in the warmer colors, as well as neutralizing the blues (particularly useful for sky scenes,) which tended to be shifted towards yellow/cyan.

 

Also, after some experimenting, I think levels work best for the final gamma correction, giving a more "contrasty" look to the image than curves; with this in mind, I also added a final levels correction with a stop dialog (pre-set at 0.75) for the user to select the desired amount of gamma correction according to each individual image.

 

So, now that I'm finally getting acceptable results, I've begun scanning my entire negative archive at 1350 dpi, 16bit linear ("keeper" frames are scanned at full resolution, of course.) I then use the action in Photoshop --in batch mode-- for a quick post-processing.

 

I'm saving the processed images as JPGs (compression level 10 in Photoshop), while keeping the original 16linear ones as "negatives".

 

On the Digital Camera topic, congratulation on your recent purchase of the Canon 20D. If money was not a concern, that's the camera I'd go for. However, since money *is* a concern, I'll have to settle for the 350D, which I'm sure you know it's very similar to the 20D.

 

There's a lot to be said in favor of digital cameras. For one thing, they have finally (in my opinion) come of age and are able to compete with film, especially the ones in the 8 megapixel league and above. And when one considers the lack of grain, and the much faster workflow of working with RAW files versus scanning/processing negatives, I think digitals win the day.

 

I guess we won't be discussing film scanning much longer. I've really enjoyed this thread, and thanks to all your input I've been able to greatly improved my workflow and reduced scanning times. You and I seem to share a meticulous attention to detail which is not very common, as well as the patience to try time and again the different techniques we have discussed -- all in the quest for the "perfect scan" :o) Because of all this, I must say: it's been a real pleasure.

 

Hope you have fun with your new "toy", and if I come across any new method of getting the most out of the Scan Elite I'll be sure to let you know.

 

Best,

 

Marcos / www.aukereman.com