[argyllcms] Re: Creating camera profiles with Argyll

  • From: "Pascal de Bruijn" <pmjdebruijn@xxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Tue, 11 Nov 2008 17:36:24 +0100

On Tue, Nov 11, 2008 at 4:00 PM, Graeme Gill <graeme@xxxxxxxxxxxxx> wrote:
> Pascal de Bruijn wrote:
>>
>> Though I'm curious about three things:
>> 1. Ultra quality seems extremely slow... Much Much Much slower than
>> High quality... I've
>> had high quality compute in a minute or three... While I've had Ultra
>> quality running for
>> tens of minutes (after which I aborted). Is this normal?
>
> Yes. "Ultra" can take an extreme amount of time (hours)
> to create the B2A table.
> Don't use it unless you have a specific reason.

Right. Got it.

>> The point why I was considering using ultra is to get 4096 entry
>> resolution instead of the 2048
>> which high quality produces. The general reasoning would be my camera
>> produces 12bit data,
>> which "means" 4096 entries.
>
> Right, but you need to take into account that the 1D lookup
> curves have 16 bit entry values, so the number of entries
> has no specific relationship to the bit depth you are
> working with.

Ok, so 2048 entries should be fine then...

>> 2. The generated profile has a A2B0 table and a B2A0 table, but do I
>> really need both? I think
>> Nikon generates a high res A2B0 table, and a low res B2A0 table. Can I
>> do something similar
>> with Argyll?
>
> You can choose a lower resolution or no B2A table at all
> using the -b flag (ie. -bn). The B2A table is there so that round
> tripping can be done on the device colorspace.

Ah, great! In my case it's probably best to not generate the B2A table at all
then. Since it would be silly to convert an image into the camera color space.

>> 3. I also noticed Argyll identifies itself with 'argl', however all
>> other profiles I've
>> inspected so far use caps to identify themselves. Just of out curiosity,
>> why
>> the lower caps?
>
> No particular reason, that's just the tag I registered I think.
> Many other tags are lower case.

Thanks for the information,
Pascal

Other related posts: