[Tig] Is there anybody out there

[Deanan]

oops. forgot to set it as plain text:
 
bobineinc at yahoo.com wrote:
> That "test" is an eye opener sort of speak!
>   
I'm out of town next week but I'll try to do it for you the week after.
(if I forget, please remind me :)

>
> Unless we are astronomers in need to see  big objects
> colliding  million miles away in HDR  to try to
> predict a catastrofic gamma burst, we could be
> throwing away colors! Correcting a picture that is
> very flat and very muddy require very large and less
> controllable moves that could show large step-overs in
> certain colors. It could be a disadvantage to say the
> least because if we're not super careful, we could
> exceed the dynamic range of some encoding by making
> corrections that are worse than we would make with
> less range.
>   
I think we're mis understanding each other. In particular there are 
different
uses for HDR in capture vs correction. I agree that correcting in HDR
directly is not a great idea. However, capturing in HDR is another story.
In the same way that a negative has a higher dynamic range than a film 
print,
rec709, etc., a digital camera with wider latitude can have the same 
advantages.
Having a larger latitude give you the option to transfer the wider latitude
to the output medium more optimally according to your preference.

For example, when we capture raw, we're capturing the full range of the
sensor in linear to light. By doing so, we can then adjust the exposure
in post to a certain extent rather than baking it in on set to fit into
rec709. Because we're not compressing the highlights from 14bits
down to 10bits, an overexposure can more easily be mapped down.
On the other hand, you also have the full range to work with so if
you chose to pull down the highlight data so you can see the detail
in the highlights, the data is still there. In most cases though, a more
contrasty mapping is used rather than a wide range, low contrast mapping.

After the mapping stage, you then take it into CC as you normally would
in print density or rec709. Some of the colorists we've talked to actually
prefer to get the data in it's unmapped state because it gives them
more options CC rather than some arbitrary HDR->visual space mapping.

> Grain is what film is made of and noise is missing
> data.  Film is sharper than digital capture. Have you
> shot a resolution chart with both film and digital
> capture? Well film has the advantage because each
> grain can record whether there is a "line" or
> "no-line"!
>   
Although, what happens when there is no line on film (ie. a smooth flat 
surface)?
There is detail (grain) where there is an absence of detail so in this
sense grain is similar to noise because it's missing data. I'm not saying
this is a bad thing (I quite like grain) but it's one of the grey areas
between noise and grain. Grain definitely adds the advantage
of increasing the perceived sharpness because it's everywhere.


> This proves that film is sharper and binary. 
The only real way is to do a side by side test and judge for yourself.
You're welcome to come shoot anytime with our camera if it interests you.
In my opinion, each has it's own advantages and disadvantages.

> the shadows. Every time people try to use smaller
> pixels for cameras, they get more noise,.
>   
Definitely true. You definitely have to balance the use of larger
pixels with resolution. Going for lots resolution and sacrificing
dynamic range in my opinion not worth it. Likewise, to dig more
into the shadows in film without blocking up, you shoot a faster
film (or rating) which in turn increases the grain size.

> As for skine-tone film is more forgiving because it
> takes more grains than pixels to make a shade. 
>   
At the same time if you shoot with a grainier stock, the grain
hides the wrinkles and replaces the detail with grain structure
so you still perceive some sharpness but not the detail
you really want to hide (wrinkles). Some B/W stocks do this
particularly well.


> Isn't it that a digital camera captures colors in
> forms of gray-steps and encodes those levels in forms
> of color at the raw conversion? Does that mean more
> colors clipping in a smaller working space and more
> bits needed in wider working spaces? 
>   
I'm not sure if I understand your question completely.
Clipping is function of many sensor and camera designs. The actual
design of our sensor on a hardware level is that the sensor sites
have an inherent shoulder of two to three stops above white.
So instead of going linear all the way up to white and then
clipping everything off, the sensor responds to more light
at the very top in a more natural way.


>
> I suspect that Biedny, Monroy, and Moody from ILM at
> the time also knew a thing or two about HDR. After
> all, they were the power behind that program called
> "Photoshop" which had basicly HDR capabable.
>   
I've used photoshop from version 1.0 which I think was
about the same time (or a bit later) as I was using Radiance.
For many years, Photoshop was strictly 8bits per channel although
they were no doubt aware of higher range stuff. Don't forget
to also include Thomas and John Knoll as the powers behind
photoshop :)


Cheers,

Deanan