Re: [Lcms-user] conversion sRGB to L*a*b* fatal error

[<ma...@li...>]

Hi,

>good news for you.
>conversion with icctrans works quite well - in both directions.
>  now i know how to deal with it :-) values are the same as in
>  photoshop - except for rounding differences.
>
>i must admit - hmmh - i made a beginners mistake...
>when choosing color values with photoshop 'color picker' i was in
>  L*a*b* color space - and so was very often out of RGB gamut...
>  got a hot flush when i realised this.
> sorry Mart=ED for this false alarm, my only excuse is that i _am_ a
>  newbie to color models and color handling.

Don't apologize :-) You have done a great job checking the engine=20
and then reporting what seemed a bug to the list. This is the way we=20
uncovered bugs, so this is really a contribution, and it is welcome.

>in that situation i threw my eyes upon the L*a*b* color model which
>  hopefully gives me what i want - at least it has one value to adjust
>  'lightness'.

Yep. Lab is very adequate for doing image manipulation. In first place,
is near perceptually uniform. This is good for most operations. Then, it
has Luma separated from chroma, again a nice feature.=20

Here are some "tricks" on Lab:

 - Addition works far better than in RGB. Adding two Lab values does =
give the resulting Lab color.
 - To modify "brigtness" just add o substract a constant to L
 - To modify "contrast" apply a exp() on L.=20
 - You can also use polar form of Lab LCh. There are a couple of =
functions to convert Lab <-> LCh
 - You can "desaturate" or "oversaturate" color by =
incrementing/decrementing C in LCh
 - Saturation of color is given by s =3D C/L
 - Hue angle is given by h

Best Regards,
Mart=ED Maria
The little cms project
http://www.littlecms.com
ma...@li...



  ----- Original Message -----=20
  From: Andreas Sch=F6mann=20
  To: lcm...@li...=20
  Sent: Saturday, March 16, 2002 9:34 AM
  Subject: Re: [Lcms-user] conversion sRGB to L*a*b* fatal error


  hi Mart=ED,

  good news for you.
  conversion with icctrans works quite well - in both directions.
    now i know how to deal with it :-) values are the same as in
    photoshop - except for rounding differences.

  i must admit - hmmh - i made a beginners mistake...
  when choosing color values with photoshop 'color picker' i was in
    L*a*b* color space - and so was very often out of RGB gamut...
    got a hot flush when i realised this.

  sorry Mart=ED for this false alarm, my only excuse is that i _am_ a
    newbie to color models and color handling.

  maybe you like to know how i came to lcms...
  i intend to make a tool to adjust brightness of each individual pixel
    of a picture - for now JPEGs.
  to do so i modified the IJG source code - needed three days of
    hard work...
  when i came to the point where i had access to the sample data i
    wondered how to modify the RGB values in order to modify
    brightness - but without changing the 'color' or at least my
    perception of a color becoming 'lighter'.
  whatever this _really_ means - need to find out more about it -
    i figured this must be difficult modifying three independent color
    values at once.
  in that situation i threw my eyes upon the L*a*b* color model which
    hopefully gives me what i want - at least it has one value to adjust
    'lightness'.
  but how to convert RGB to L*a*b* from within C code? now i know
    the answer and am very happy with it.

  thank you Mart=ED and all others involved in lcms project!

  andreas

    ----- Original Message -----=20
    From: Mart=ED Maria=20
    To: Andreas Sch=F6mann ; lcm...@li...=20
    Sent: Saturday, March 16, 2002 3:14 PM
    Subject: Re: [Lcms-user] conversion sRGB to L*a*b* fatal error


    Hi,

    > i am trying to validate the 'icctrans'-values for sRGB =
IEC61966-2.1 to L*a*b* conversion.=20

    Humm... I have checked this utility extensively, and seemed to work. =

    Are you sure Photoshop is using sRGB as workspace (Edit-Color =
settings-Working spaces RGB)?

    I'm also attaching latest source, with some small additions. But =
anyway the old one did work properly to me...=20

    This would give sRGB to Lab

    icctrans -v

    And this Lab to sRGB

    icctrans -i*Lab

    If this don't work to you, could please send any values that makes =
it fail? Thanks.

    Regards,
    Mart=ED Maria
    The little cms project
    http://www.littlecms.com
    ma...@li...






      ----- Original Message -----=20
      From: Andreas Sch=F6mann=20
      To: lcm...@li...=20
      Sent: Friday, March 15, 2002 5:10 PM
      Subject: [Lcms-user] conversion sRGB to L*a*b* fatal error


      hello marti,

      i am trying to validate the 'icctrans'-values for sRGB =
IEC61966-2.1 to L*a*b* conversion.
      i want to validate against adobe photoshop - i simply use the =
'color picker' to get values.
      i am unlucky :(
      neither the built in profile 'cmsCreateLabProfile(NULL)' nor =
'lcmstiff8.icm' nor 'tifflab8spac.icm' give me the photoshop values.

      i built lcms 1.08 with cygwin on win2000, build looked ok.

      what do you think?
      maybe i missed something?
      i'm not sure whether i use the right profiles.

      regards,
      andreas