negative transfer: another try

[christoph]

justin was kind enough to send me a roll of Vision2 500T from his tests and bravely i attempted to capture it with my setup that's still under construction..

here's what i got (click for 1K res):



please dont complain about the dust, it's a "quick and dirty" transfer in the truest sense... and i'm not terribly happy with the colors. but since i'm not sure if i have a lot of time the next days so though i'll post it anyway.

also note that it's possible to use a different scaling algorithm to preserve more detail, but this will also enhance grain:


the main prob with neg transfer is not so much the orange maskin, but rather the low contrast that has to be stretched (ironically exactly the opposite of the difficulties in reversal transfers)
to illustrate, here's what the original linear capture looks like (with histgramms):



and for those interested in playing around with the cc themself, here's a lossless 16bit png file

++ christoph ++

[ccortez]

Looks very, very good Christoph.

The original is better than anything I ever started with doing neg xfer. I really think you're getting there. I was never sure if part of the reason the original is so low con was partially b/c of the optical correction i was using to negate the mask... and i never got past having to open the dv camera way up to pick up much at all... anyway, these look far less soft and washed out than almost all of what I've been able to accomplish. I'm sure you've seen much worse hist's than that, i certainly have.

It still feels a little "thin" but I think you can probably do wonders with it in CC. I can't wait to see what people come up with playing with the full res version.

BTW, I don't mind the extra grain in #2 and to me the sharpness/detail is worth it.

[paulcotto]

Looks very promising. There are several great programs out that are designed for color correcting. I have been playing with a demo of Apple's shake>

http://www.apple.com/shake/

I am impressed with the results so far, has anyone else tried it?

Regards,
Paul Cotto

[VideoFred]

the main prob with neg transfer is not so much the orange maskin, but rather the low contrast that has to be stretched (ironically exactly the opposite of the difficulties in reversal transfers)
to illustrate, here's what the original linear capture looks like (with histgramms):

Then this is good news for me: dynamic range from the digital cam is less important with negative.
The stretching is a software matter.

But I bet the backlight diffusion is the most important factor, to avoid the typical negative artifacts.

Nice work, beautiful first result. 8)
Filmlook is awesome, this proves it again.

PS: You sure have captured it very dark!

Fred.

[christoph]

http://www.apple.com/shake/
I am impressed with the results so far, has anyone else tried it?

ps: i should note that while shake is awesome, it's not intuitive to use and you have to know quite a lot about the basics of digital image maniulation to get the most out of it. most people will be better off with after effects and the like.
++ c

[christoph]

Then this is good news for me: dynamic range from the digital cam is less important with negative.
The stretching is a software matter.

But I bet the backlight diffusion is the most important factor, to avoid the typical negative artifacts.

i wish it were that simple..
the most important thing is having a camera which can resolve *very* subtle color changes... for best results it should have true 12bit (my camera records 10bit (on the paper) and even with averaging 3 frames you still can see the chip noise if you look carefully)...
stretching might be easy with software, but if there isnt enough different colors to stretch to begin with you'll get heavy banding and/ or chip noise.

PS: You sure have captured it very dark!.

uh, i should have mentioned that it's a linar 16bit file directly from the RAW conversion.. you'll have to apply a hefty gamma change to get to something which looks normal.

[Scotness]

Wow that took a lot of work - using photoshop I had to go through levels then color balance then levels again - then a bit of saturation - then dropped the brightness a bit - then more colour balance (a bit of red for skin tone)



Scot

[VideoFred]

the most important thing is having a camera which can resolve *very* subtle color changes... for best results it should have true 12bit (my

I see..
This makes me think what these Rank machines are using for the sensor.
Something custom made, or available in the industry?
These are line scan modules, right?
Then how do they know if a frame changes?

uh, i should have mentioned that it's a linar 16bit file directly from the RAW conversion.. you'll have to apply a hefty gamma change to get to something which looks normal.

This is far beyound (beyond?) my knowledge...
But I believe you and I learn from you

Fred.

[Blin]

i should note that while shake is awesome, it's not intuitive to use and you have to know quite a lot about the basics of digital image maniulation to get the most out of it. most people will be better off with after effects and the like.
++ c

You can get a free demo here:

http://www.apple.com/shake/trial/

[paulcotto]

Not to mention that it has been reduced in price to $499.99. I get a discount trough my company and the price is $399.00 I don’t think you can get afteraffects at that price can you?

Regards,
Paul Cotto

i should note that while shake is awesome, it's not intuitive to use and you have to know quite a lot about the basics of digital image maniulation to get the most out of it. most people will be better off with after effects and the like.
++ c

You can get a free demo here:

http://www.apple.com/shake/trial/

[Andreas Wideroe]

I see..
This makes me think what these Rank machines are using for the sensor.
Something custom made, or available in the industry?
These are line scan modules, right?
Then how do they know if a frame changes?

See my post on page 2 of this thread:
viewtopic.php?t=13052

Gives you an idea of the difference between professional telecine machines and projector-like scanners with cameras connected.

The difference in dynamic range is huge.

Off to Amsterdam...

Cheers,
Andreas

[VideoFred]

Interesting!

Andreas wrote:

What would the typical dynamic range for these cameras be in a 1:something ratio?

I have a Hitachi HV-D30P camera for my Sniper (same used in the Flashscan8 machines) and I'm not happy with the dynamic range.

Both the FDL-60 and FDL-90 have a 1:7500 dynamic range. The FDL-60s a 1024x1 pixel CCD and the FDL-90s an improved CCD chip 1024x1 pixels, one for each RGB colours.

I think the Spirit has a 1:15000 dynamic range.

Can I assume 1 is the darkest pixel and 15000 is the brightest possible pixel, without blowing out?

The problem is, often they do not mention dynamic range in the specs.
And if they do, it is in db. I know 60db is not bad.

Can someone explain more?

Fred.