Kodachrome Revisited

[pfigen]

Well, I've scanned hundreds of my own Kodachrome 25, 64 and 200 images on a Howtek drum scanner, which can actually "see" the color of the base of the film. My scanner is profiled with a Hutchcolor Velvia target and here's what I see:

When I scan Kodachrome, or any film, I always include all of the film including the rebate edges. When scanning Ektachromes, that rebate edge will come in somewhere around 6 Red, 5 Green and maybe 3 Blue. The scanner actually sees the color of what we perceive as infinite black but really isn't. When scanning Kodachrome of any variety, the RGB numbers for the film base come in more like 8 Red, 6 Green and 25 Blue. The scanner is seeing the bluish color of the film base and reporting back as such.

My solution, whether scientific or not, is to manually push the black point of blue channel down to where it would be when scanning E-6. Trident - the software that runs the Howtek - has a very easy way to click down those values independent of each other, and dropping that blue channel down in the shadow end also affects the entire image, but on a gradated basis. It's almost magical when you see the color come right to where it's supposed to be with each click.

As I said, I've scanned many hundreds of my own images and this has worked every time. As for people who complain about whether a particular method is scientifically accurate or not, it really doesn't matter. I have yet to see an image of mine or anybody else's that could not be improved upon either in the scanning software or in Photoshop later on. I'm going for the best possible image and can't wait the last sixteen rolls of Kodachrome I sent in after Christmas to arrive in the mail.

As for the drum scanner operators who claimed they couldn't scan Kodachrome, that does not surprise me. So many were just scanning bots who never really understood the why of what they were doing. They were the same people who thought drum scanners sucked for scanning color negs as well. I imagine most of those folks have been retrained for other jobs by now anyway.

[Schewe]

As for the drum scanner operators who claimed they couldn't scan Kodachrome, that does not surprise me. So many were just scanning bots who never really understood the why of what they were doing. They were the same people who thought drum scanners sucked for scanning color negs as well. I imagine most of those folks have been retrained for other jobs by now anyway.

Those people have been long gone for a decade...(and yes, I heard worse actually about scanning negs than Kodachrome).

[Ernst Dinkla]

Ok, Ernst, I guess you just want to continue to throw stuff out there, and ignore any questions.

Cliff,

I come back to this issue when I have more time but there is one thing that I object to: Your cyan dye plot isn't the 1:1 explanation for all that happens here. And it if was and solved all the issues in one clear way I wouldn't object at all. What seems to me a key is that Kodachrome unlike all the other slide films has been balanced for 3200K viewing and that it got other dyes in its chemistry for that reason and/or the different process used. Your cyan plot falls within that scheme. The next question is whether the dyes and the cyan especially show more metameric issues brought to 5000K (not-continuous) scanner light than for example Ektachrome brought to 3200K projection light. If so that would explain the acceptance of mixed slide projections in 3200K. Another example is the IR aspect of the cyan dye has been discussed for ICE but not in relation to sensor RGB filters and the sensor itself.

I think these threads should be seen as brainstorming, to get once and for all the factors on the table that really matter. There has been a lot of confusing information available, we have seen that so far. On one hand I think that it is because it is approached in an unscientific way (be careful with Hutcheson though) on the other I think there are that many variables that can lead to different results and by that to other conclusions, right or wrong.

To the others who want to have an instant solution: there are many out there, all with the same claim to be the sole and right solution. Pick one that suits you. This thread may not lead to a simple solution or to a solution at all.


met vriendelijke groeten, Ernst Dinkla

Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/

[crames]

I've read all the Kodachrome threads. Nothing was finalized or proven. Didn't learn any knew information that would tell me what everyone was trying to achieve scanning kodachrome. Out of frustration I had to ask myself just what the heck is Kodachrome suppose to look like since everyone here seems to know from examining kodachrome images, from where I haven't a clue.

Did you miss the first thread in the series? http://www.luminous-landscape.com/forum/index.php?topic=50262.msg415125#msg415125

Kodachrome was designed for projection viewing (at least the more recent versions), so I think the gold standard would be the way it looks when projected.

A problem with Kodachrome scans is that the color can be ruined by a blue cast and excessive contrast. The scans have to be adjusted somehow to make them look more like the projected slide. It seems that a lot of Kodachrome scans you see on the web haven't been adjusted very well, or at all, and therefore look crappy.

Some more recent versions of scanning software such as SilverFast and NikonScan have special Kodachrome settings that give very good results.

I can never understand how you can have a discussion on color without showing color. It's all talk, spectral curve graphs and mathematics with no connection to their results.

Along with graphs, which tell a lot if you know how to read them, there were in fact some example images posted. You know you have to be logged-in to LuLa to see them?

I suggested a very simple method using levels to improve Kodachrome scans. Why don't you try it on some of the crappy ones you find on the web?

[Tim Lookingbill]

Trident - the software that runs the Howtek - has a very easy way to click down those values independent of each other, and dropping that blue channel down in the shadow end also affects the entire image, but on a gradated basis. It's almost magical when you see the color come right to where it's supposed to be with each click.

Yeah, Peter, that looks like it does something similar to ACR's Split Tone slider. I guess from what I've gathered from these threads an ICC profile derived from an iT8 target can't do that.

Crames, I read that thread and saw the image samples and visited the links. All I got from it was either Kodachrome isn't suppose to look blue or it is but for projection viewing, or it depends on what version of Kodachrome you're scanning. It was just too confusing.

However from what I gathered from those image samples and comparing them to the first four beach scenes in that last Getdpi.com forum link I posted, Kodachrome sees light beige colors in caleche roads, sand and other warmish pastels as having a light ashen putty hue and shadows having a blue to blue green cast which isn't accurate, but quite pleasing from a color design sense.

But then I notice how all four of those beach shots changed hue in sky blue and the color of the sand. They're the same beach and same time of day. Don't know if that's user error or the inconsistent nature of Kodachrome.

[Mark D Segal]

Hi Mark,

It should be easy to get at least a rough idea about what is being done. Scanning an image with a gray scale twice, once with the Kodachrome media setting on and once with it off, and comparing the RGB values (or the density values) should reveal something about how the tone curves are being manipulated.

I happen to have a couple of rolls from the last batch of K64 on the way back from Dwayne's, including a few frames of a Color Checker, so will try looking into it a bit more when I can.

Cliff

Hi Cliff, yes I agree - there isn't all that much in the scan software which one couldn't reverse engineer with enough logic, trial and error. Your question about what the applications are doing when you select Kodachrome as the media had the word SPECIFIC in it, so that's where it got hard to answer because the developers won't get too specific. But just because it's their secret sauce doesn't meant others can't devise the necessary workarounds - such as shifting the blue curve and/or other such tweaks with several tools to similar effect. The one thing I have gleaned from my inquiries and experience on this matter is that each scanner model reacts differently to the same images/same media, so the recipe should be scanner model specific for best results. With the extra effort, decent scans can be made from Kodachrome systematically, but Jeff is correct that it does take more work than scanning E-6 type films.

[davidh202]

Truth is I haven't yet Jeff, and I never did say, I knew better !
Though intelligent and mechanically inclined, I am rather technophobic and I have repeatedly put it off for fear of failure. I was so eager to try again until I started reading all the tech stuff. I had done a lot of research a few years ago and just could not afford a top end film scanner so I purchased a Microtek I900 that I thought would be a compromise.It came bundeled with Silverfast but needed to upgrade to their pricier program with their Kodachrome "solution"nearly 8 months ago.I also
questioned  the merits of their scanner profiling "solution" on this forum a few months ago. After Mark Segal replied that  scanner profiling is basically for the birds I gave up on that, and have yet to decide what course of action I should take.I don't want to invest $400 -600 on software that really wont do the job as they say it will.
It just seems that no matter what topic comes up there are always differences of opinions and very different takes on what works and what doesn't.
That was all I was trying to express in my post. It seems that unless one can carry on a very technical conversation on this site we are intimidated and brushed aside as totally ignorant by you.After reading the threads now I am as confused as ever.
This site has had such good recommendations and I came to learn but.....

[Mark D Segal]

Truth is I haven't yet Jeff, and I never did say, I knew better !
Though intelligent and mechanically inclined, I am rather technophobic and I have repeatedly put it off for fear of failure. I was so eager to try again until I started reading all the tech stuff. I had done a lot of research a few years ago and just could not afford a top end film scanner so I purchased a Microtek I900 that I thought would be a compromise.It came bundeled with Silverfast but needed to upgrade to their pricier program with their Kodachrome "solution"nearly 8 months ago.I also
questioned  the merits of their scanner profiling "solution" on this forum a few months ago. After Mark Segal replied that  scanner profiling is basically for the birds I gave up on that, and have yet to decide what course of action I should take.I don't want to invest $400 -600 on software that really wont do the job as they say it will.
It just seems that no matter what topic comes up there are always differences of opinions and very different takes on what works and what doesn't.
That was all I was trying to express in my post. It seems that unless one can carry on a very technical conversation on this site we are intimidated and brushed aside as totally ignorant by you.After reading the treads now I am as confused as ever.
This site has had such good recommendations and I came to learn but.....

I think you have sadly misunderstood my position on scanner profiling. I've written two articles published on this website explaining the merits and limitations of scanner profiling and reported results in some depth. You do need to use an appropriate scanner profile for your scanner. The only question is whether you can get away with using a canned profile (for example SilverFast comes with a large number of them customized for the scanner MODEL), or whether it is better to make custom profiles for your specific scanner. That depends on the consistency/quality of the scanner manufacturing from one unit to the next. So to be clear: use an appropriate scanner profile and make it a custom profile if the canned profile isn't good enough! Oh - and I should add in case this is the source of the confusion: one cannot profile a scanner for negatives - only for positive.

[crames]

My solution, whether scientific or not, is to manually push the black point of blue channel down to where it would be when scanning E-6. Trident - the software that runs the Howtek - has a very easy way to click down those values independent of each other, and dropping that blue channel down in the shadow end also affects the entire image, but on a gradated basis. It's almost magical when you see the color come right to where it's supposed to be with each click.

Your approach makes sense. As you say, pushing the black level down in the blue channel will subtract blue on a graduated basis - proportionately more blue is subtracted in the shadows, as you go higher in the tone curve proportionately less is subtracted. This is a reasonable adjustment since the excess blue is highest in the shadows and drops to nothing gradually as you go up to the white point. Do you make any adjustment to the red channel?

The Trident software has a density curve control in the Tone & Cast window. I'm wondering if you could just steepen the blue density curve a little and lower the red density curve to get a neutral gray scale? It seems to me that working with densities in scanner software is a natural way of manipulating the dye densities in scans; too bad it's only available at the high end.

[davidh202]

My bad  Mark,I didn't splain myself correctly
My question in that thread ( http://www.luminous-landscape.com/forum/index.php?topic=47258.msg393601#msg393601)  was about using   scanner software to create custom paper profiles as Silverfast says you can, not a scanner profile,
And it was Andrew Rodney that had replied, not you  

[crames]

All I got from it was either Kodachrome isn't suppose to look blue or it is but for projection viewing, or it depends on what version of Kodachrome you're scanning. It was just too confusing.

No, it's not supposed to look blue, any version, when it's viewed the way it's supposed to be viewed.

However from what I gathered from those image samples and comparing them to the first four beach scenes in that last Getdpi.com forum link I posted, Kodachrome sees light beige colors in caleche roads, sand and other warmish pastels as having a light ashen putty hue and shadows having a blue to blue green cast which isn't accurate, but quite pleasing from a color design sense.

It looks to me that most of those images at the GetDPI link have the blue curse.

In Levels, try a Red mid-tone (gamma) slider setting of 1.2 and a Blue mid-tone setting of .9, especially on the beach, vegetables, air show, and dog. Adjust to taste.

Removing the blue cast can really make the color pop, don't you think?

Cliff

[pfigen]

"Your approach makes sense. As you say, pushing the black level down in the blue channel will subtract blue on a graduated basis - proportionately more blue is subtracted in the shadows, as you go higher in the tone curve proportionately less is subtracted. This is a reasonable adjustment since the excess blue is highest in the shadows and drops to nothing gradually as you go up to the white point. Do you make any adjustment to the red channel?"

That's exactly how it works. There is virtually no effect in the highlights, although you could easily put one in if you needed to. I usually don't need to touch the other channels, but it all depends on the particular piece of film. You do whatever it takes to make it right.

"The Trident software has a density curve control in the Tone & Cast window. I'm wondering if you could just steepen the blue density curve a little and lower the red density curve to get a neutral gray scale? It seems to me that working with densities in scanner software is a natural way of manipulating the dye densities in scans; too bad it's only available at the high end."

You could do that as well, but the curve controls are a bit weird - not at all like Photoshop's. There are three or four ways to accomplish the same result that are basically user choices in interface that all affect the same thing. Sometimes it's faster to use the approach that causes the least amount of trouble. I also think that when you have a scanner that can actually see the color in the base density, that changes the way you deal with the software.

[Tim Lookingbill]

Crames,

I thought those Kodachrome scans of the beach scene were quite pleasing and thought that was the color characteristic of Kodachrome everyone raves about. So that color palette I described isn't as intended? If so then that's disappointing to hear from a graphic design point of view. I mean if you ran those beach images through a color palette generator like at this site:

http://www.pics2colors.com/

I'ld guarantee it'ld make a color designer drool. Dan Bayer's Kodachrome Project practically emphases that hue of blue in most of the shots on his site especially images with reds like this:

http://www.flickr.com/photos/23585735@N06/3616350590/in/set-72157619583742644/ That really pops!

Wonder why those beach scenes aren't as blue as other Kodachrome shots? What's the variable that makes one bluer over another? Looks like it might have something to do with actual color temperature of the scene, but then I saw a beach scene at Photo.net Kodachrome thread that was all blue and quite dark.

[pfigen]

Here's what I consider a pretty good example of a Kodachrome 64 scan. This was shot in West Rawlins, Wyoming somewhere around 1990 or '91. It's an advertising sign for The Bucking Horse Lodge, which, according to online reviews, completely sucks.

[Ernst Dinkla]

When I scan Kodachrome, or any film, I always include all of the film including the rebate edges. When scanning Ektachromes, that rebate edge will come in somewhere around 6 Red, 5 Green and maybe 3 Blue. The scanner actually sees the color of what we perceive as infinite black but really isn't. When scanning Kodachrome of any variety, the RGB numbers for the film base come in more like 8 Red, 6 Green and 25 Blue. The scanner is seeing the bluish color of the film base and reporting back as such.

The lamps of the Howtek are halogen tungsten ones with an integrated mirror I gather. Any idea what their K value is and the spectral plot? I would expect a continuous spectral plot and 4000-5000K. Osram seems to have replacement lamps for Howteks, I could check that.



There is a page on Coloraid where extended color patches film targets are shown, a bit in the sense of Hutchcolor targets but unique test sets to improve profile creating software. I have not yet explored that but I see several scanners (a Howtek too) mentioned that were used with the targets. Wonder if that could shed a light on the lighting and RGB filter difference between scanners. More drivers are used too.

http://www.testdata.coloraid.de/


met vriendelijke groeten, Ernst Dinkla

New: Spectral plots of +230 inkjet papers:
http://www.pigment-print.com/spectralplots/spectrumviz_1.htm

[crames]

Here's what I consider a pretty good example of a Kodachrome 64 scan. This was shot in West Rawlins, Wyoming somewhere around 1990 or '91. It's an advertising sign for The Bucking Horse Lodge, which, according to online reviews, completely sucks.

That's a great looking scan!

[Mark D Segal]

pfigen: I agree with Cliff - it is a great-looking scan!

Ernst: I just went to the link you provided to Wolf Faust. Interesting stuff, but seems to be an abandoned exercise dating to 2006/07, and as he said "a work in progress". I don't know how important that is - sometimes dated material remains just as valid today as it ever was for shedding light on certain issues - I'd need to study it further to see what it informs us about the current thread topic. However, to use the materials he shows for doing one's own tests, (a) we need to obtain the slide set from him (still possible?) and (b) we need to have our own profile creation software. So those two conditions will determine who can experiment further with Kodachrome materials using this particular approach.

[Ernst Dinkla]

I didn't read it properly but I thought if one started first with copying the most saturated CMYRGB patches from the images files there it should give a rough idea on how selective each scanner lamp + filter combination is. There are driver variations too but one would assume that a true RAW scan should be the same. Wolf did take some slack on the results though, small deviations between sets and different scan operators.

Pity it wasn't done with Kodachrome targets.


met vriendelijke groeten, Ernst Dinkla

Dinkla Gallery Canvas Wrap Actions for Photoshop
http://www.pigment-print.com/dinklacanvaswraps/index.html

[crames]

I thought those Kodachrome scans of the beach scene were quite pleasing and thought that was the color characteristic of Kodachrome everyone raves about. So that color palette I described isn't as intended?

I don't know what the photographer intended other than what was posted, but I suspect that the actual slides on a light box or projected are warmer, more vivid and colorful.

Dan Bayer's Kodachrome Project practically emphases that hue of blue in most of the shots on his site especially images with reds like this:

http://www.flickr.com/photos/23585735@N06/3616350590/in/set-72157619583742644/ That really pops!

It pops, but if you get rid of some of the blue, it pops even more and has an increased illusion of depth. (Some of the blue is due to the sky light that lit the shot)

Kodachrome, as a general-purpose color film, is certainly designed to have, if not accurate color, then pleasing color. Some of the most important colors are skin tones. The skin in a lot of the shots at http://www.kodachromeproject.com are drab, bluish, not so pleasing - not what Kodachrome can really look like on a light box or projected. So I think the scans have not been prepared well to show what Kodachrome can really do. Speaking only of the color and not the artistic merit of the images, I would guess that a lot of people hoping to see "what's the big deal about Kodachrome," will be saying "so what?"

Wonder why those beach scenes aren't as blue as other Kodachrome shots? What's the variable that makes one bluer over another? Looks like it might have something to do with actual color temperature of the scene, but then I saw a beach scene at Photo.net Kodachrome thread that was all blue and quite dark.

The color temperature of the scene will have a large effect if it's not filtered to match photographic daylight (D54 or D55). As you underexpose Kodachrome, the colors are increasingly shifted to blue.

[pfigen]

Ernst - The Howtek does indeed use an Osram 8v20w tungsten halogen lamp. That lightsource is centered on a flexible fiber optics tube that is in turn focused through drum, film and into the lens, after which it passes through the scanner's aperture device, through a beam splitter and through three dichroic filter on its way to the respective side window 931B photomultiplier tubes. I don't know but I have to think that the characteristic of those dichroic filters have to have a part in the equation as well, as they define the actual color gamut of the scanner.

I believe that Heidelberg drums also used an inexpensive halogen as well, but the Optronics Colorgetters used a $600 high intensity lamp that was much bluer in comparison. I'm sure you have to take the entire system into account and not just the color temperature of the light.