With virtually all the data in the blue channel, typically the
noisiest, it's no wonder the noise filtering algorithms are bent out
of shape. ;-) There's virtually no G channel at all (should be half
the data) and only 5% of the normal complement of R channel.

I downloaded four of the PEFs and the catalog and played with them on
two different directions:

- to create a false color rendering, I started by manipulating the
Tone Curve on the channels to spread out the range of values in the R
and G channels, then did a bunch of tweaking in the HSV panels. The
Hue and Saturation panels in particular proved very very sensitive
within certain ranges. I reached what I thought was an interesting if
odd false color look and applied it to a vcopy of the other three
files. It worked pretty well, but needs more development work.

- to create a monochrome rendering, I started with the camera
calibration and changed the basic camera mix feeding into the other
tools. Then some small modifications in the Basic and HSV panels
neutralized tones.

With either of these renderings, I'd output the results as is to TIFF
files and do any further editing in the TIFF RGB space, rather than
with the tools operating on the raw data—the settings are too fragile
to give much adjustment range at this point. If I'd thought of it last
night, I'd have opened the Adobe Profile Editor and created a custom
camera calibration to do what I did in the Camera Calibration panel
... It's more precise and has more range than the panel's tools.

It's fun to work with pathologically messed up raw data ... ;-)

G

On Mon, Aug 20, 2012 at 9:51 AM, Bruce Walker <[email protected]> wrote:
> On Sun, Aug 19, 2012 at 9:38 PM, Larry Colen <[email protected]> wrote:
>>
>> On Aug 19, 2012, at 6:07 PM, Bruce Walker wrote:
>>
>>> Hey, a Blue Man Group cover band! :-)
>>>
>>> That looks like a challenge. I'm game. How about posting a .dng on DropBox?
>>
>> I made a small lightroom catalog with the raw files (.pef) and posted it in 
>> a hidden directory on red4est.com
>>
>> If you want to give it a try, ping me, and I'll email you the url.
>>
>> The .pefs are still uploading, and will take quite a while, I expect.
>
> Larry, I've had a chance to play with one PEF (8698.PEF) and here's
> what I found.
>
> The extreme contrast between levels in the RGB channels appears to
> freak out Lr. A few functions don't work like they should at all (eg
> using the eye-dropper to detect points in the tone curve: total fail).
> It also freaks out the Noise Reduction algorithms so they botch the
> image by posterizing it.
>
> But I got a decent, albeit grainy, b&w rendition. Try this:
>
> Start with all controls at their Lr defaults.
> Your detail-over-smoothing problem is solved by reducing the "Detail:
> Noise Reduction: Color" to zero (it defaults to 25). Do that first.
> Lens Corrections: Enable Profile Corrections: on
> HSL/Color/B&W: Black&White Mix:
>   - 1st set all sliders to -100
>   - set Blue to +88
>   - set Purple to about -63 (adjust to balance shadows and noise)
> Tone Curve: Lights: -40, Darks: +40
> Basic: Shadows: +69
>
> The image should look pretty good now. No noise reduction at all, so
> it's quite grainy.
>
>
> If I was going to do a really proper job, I'd set all the Lr controls
> to their defaults, set Color NR to zero, then send the the image to
> Photoshop to edit. There I'd extract the blue channel and chuck the
> others (copy-all in the B channel in Channels, then paste to a new
> layer). After that play with curves and apply some NR with a plugin.
>
> Or else, after extracting the Blue channel in Ps, just save the image
> back to Lr and use the  Luminance NR and tone curves there. Set the
> Detail slider up to 80 or so when raising the NR.
>
> Cheers!
> --
> -bmw
>
> --
> PDML Pentax-Discuss Mail List
> [email protected]
> http://pdml.net/mailman/listinfo/pdml_pdml.net
> to UNSUBSCRIBE from the PDML, please visit the link directly above and follow 
> the directions.



-- 
Godfrey
  godfreydigiorgi.posterous.com

-- 
PDML Pentax-Discuss Mail List
[email protected]
http://pdml.net/mailman/listinfo/pdml_pdml.net
to UNSUBSCRIBE from the PDML, please visit the link directly above and follow 
the directions.

Reply via email to