[Openicc] New Topic: data to share

[Chris Murphy]

At this bitdepth there is enough precision that TRC encoding isn't particularly relevant. If anything we are better off with sRGB primaries because UI controls for hue editing will be more well behaved with sane primaries like sRGB than crazy ones like ProPhoto RGB, but with 32bpc float we're not limited to encoding colors inside the boundary formed by the primaries. And with L*a*b* encoding, which differs from the actual color space CIE 1976 (L*a*b*), likewise I don't see why changing to this encoding has any benefit at all. There are real colors which cannot be encoded, and there are encodings for which there are not real colors. It's kindof a goofy space, useful sorta, as a PCS, but not for image editing and has no benefits for storage.

Last I find ECI-RGBv2 pointless. There's no benefit to L* encoding. The ECI recommendation sways the user from encoding 8bpc as L* anyway, and 16bpc or higher precision encoding increasingly obviates the relevancy of TRC selection.


Chris

On May 17, 2011, at 6:53 AM, edmund ronald wrote:

> I agree. 24 or 32 bit LAB with headroom is the way to go for archival purposes. 
> 
> 
> 
> On Tue, May 17, 2011 at 2:27 PM, Graeme Gill <graeme at argyllcms.com> wrote:
> Scott Geffert wrote:
> When we assume sRGB in any part of the imaging workflow we are essentially clipping
> many colors that most modern printers (and even displays) can reproduce. In fact, Adobe
> RGB clips about 10% of these very printable colors which to me is unacceptable. When
> you are discussing "internal" spaces or untagged spaces for "no color management" I
> would argue that sRGB is not the way to go and many feel that ProPhotoRGB is too large
> but eciRGBv2 is ideal. It is L* based, it has a 5000k white point and the gamut is
> 
> There is no compunction to use RGB spaces for wide gamut. Use L*a*b* or something
> that is inherently suitable for wide gamut (even scRGB might be better than any of
> the RGB spaces you mention).
> 
> It's not a good idea to use a wide gamut space as a default space for "mystery meat" RGB
> either, if it doesn't match typical expectations, which are often set by displays.
> Looking at it statistically, your extreme measured colors are at the tail of the
> probability curve, and therefore could be safely ignored for most everyday practical
> purposes -i.e. if you are one of the few people who care about accurately reproducing
> such extreme colors is important to you, don't use an unlabelled colorspace to store the
> color values !
> 
> Graeme Gill.
> 
> _______________________________________________
> openicc mailing list
> openicc at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/openicc
> 
> _______________________________________________
> openicc mailing list
> openicc at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/openicc