On 11/01/2012 10:11 AM, Ulrich Pegelow wrote:

Hi!

> Am 01.11.2012 09:47, schrieb Alexander Wagner:
>>
>> Is there currently a way to see/work the individual channels
>> in dt?
>>
>> Though it is a bit "abstract" to work only in L, a or b
>> displayed in grayscale this sometimes helps, and especially
>> sharpening should work on the L-channel (mostly at least),
>> so "just display L" could probably already solve the OPs
>> issue.
> That's not possible currently. What you see is in any case
> a conversion of Lab to RGB.

Sure.

> One could think about some
> fake blend modes to map L, a and b into an L'a'b' where a'
> and b' are set to zero and L' represents a mapped version
> of the original L, a and b.

I have to get used to your blend mode stuff. I don't think
in this way (yet).

What I had in mind was a display similar to decompose in
gimp: you have 3 channels by default L, a, and b (or as you
put it: their representations in RGB on the monitor) are
visible. This gives me the full image. Now I could imagine
to switch off either of them, e.g. a and b leaving me only
with L displayed in grayscale. Like the three decomposed
layers in gimp that do not make up a full colour image (at
least not at the moment).

> However, this is likely not to happen short term. We also
> need to pay attention not to invent too many blend modes,
> else people lose the overview :)

Well, I think one problem is indeed that dt works with the
whole stack all the time. For me at least this is always a
bit difficult. If I adujst X it operates on the output of a,
b, c but not d cause d works after X in the chain. This is a
bit more complex then working on layer X and then combine it
with a and b and stack c above it. Sometimes I play with
settings but I have no idea in the first hand what will
happen. It's a more try&error approach than an analytical
one in the sense that I learn: first do this then do that by
experience not by "this works because". Being form the
theoretical sciences this is "different".

Also I have many complex plugins that do really great things
but I can not come down to the simple basics. Like the
mentioned Gimp plugin "Lab sharpen". It works well but I
did not understand why till I got some instructions on how
to do it by hand what gave me a better idea of the
parameters.

>> When I use Lab sharpening in<this other great software>,
>> I usually decompose the image use unsharp mask on L only
>> and, if there is some noise, blur a and/or b slightly
>> using a gausian e.g. Though there's a plugin for Lab
>> sharpen, I admit that I usually do it manually, since I
>> understood how it works. Simply cause I can see the
>> effect easier. I think this is probably what the OP
>> intends. AFAIK I can not this kind of workflow replicate
>> that in dt at the moment, as sharpen works in Lab but on
>> all channels simultaneously, right?
>
> We already do sharpening only in L and leave a and b
> unchanged.

Ah, so this is hardcoded in the plugin.

> If you want to additionally apply some denoising on a and
> b only, take module lowpass with blend mode "color".

Ah. Sounds like a good idea. However, this is a gain a bit
different to my way of thinking I admit. As soon as I apply
the low pass filter it is applied somewhere in the chain (I
believe you that the chain is sensible) but it can be there
only once. Right?

-- 

Kind regards,                /                 War is Peace.
                             |            Freedom is Slavery.
Alexander Wagner            |         Ignorance is Strength.
                             |
                             | Theory     : G. Orwell, "1984"
                            /  In practice:   USA, since 2001

------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_sfd2d_oct
_______________________________________________
Darktable-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/darktable-users

Reply via email to