CIEDE2000 and CIELUV versus CIELAB
CIEDE2000 and CIELUV versus CIELAB
- Subject: CIEDE2000 and CIELUV versus CIELAB
- From: Steve Kale <email@hidden>
- Date: Mon, 26 Nov 2007 22:03:15 +0000
- Thread-topic: CIEDE2000 and CIELUV versus CIELAB
Thanks to all that replied.
As noted earlier I had simply coded, for example ATAN2(b1,a1) instead of
ATAN2(a1,b1). I misunderstood Bruce's annotation of the formulae.
I did also mean 1994 rather than 1974. Apologies.
I need to reread the paper to try to understand the discontinuity issue but
it likely goes clean over my recollection of university mathematics. My
understanding is that CIEDE2000 has replaced the 1994 calculation as the
recommended methodology for calculating colour differences for most
applications (albeit with the caveat noted in the paper).
I'm using it for the computation of colour differences between target Rec
709 or SMPTE-C primary/secondaries and measured performance of my home
cinema plasma display. (I suspect from reading the below that the
discontinuities observed are unlikely to be an issue in my situation but
perhaps that's wishful thinking....)
To that end, I have one further question:
I understand that the CIELUV space is still preferred by many working with
video. In fact I've heard one commentator say that CIELAB is preferred for
reflective media applications but CIELUV is preferred by those working with
emissive applications. I find such comments puzzling as I had always
thought that CIELAB had proven itself over time to be the more "robust" (by
which I assume people mean "more perceptually uniform") colour space and
that computation of colour differences using LAB-based dE equations (1994 or
2000) represented best possible estimates of perceived differences in
colour. Else one would have thought that LAB would have died by now.
(Furthermore, I don't understand why one perception of colour would depend
on emmission/reflection.)
So is there any basis for using LUV or is this just a case of old habits die
hard?
Regards and thanks again
Steve
> From: <email@hidden>
> Reply-To: <email@hidden>
> Date: Mon, 26 Nov 2007 12:05:07 -0800 (PST)
> To: <email@hidden>
> Subject: Colorsync-users Digest, Vol 4, Issue 425
>
> From: Graeme Gill <email@hidden>
> Reply-To: <email@hidden>
> Date: Mon, 26 Nov 2007 11:21:40 +1100
> To: <email@hidden>
> Subject: Re: dE(2000)
>
> Steve Kale wrote:
>> Thanks. I actually found this before I got this digest. Turns out I had
>> misunderstood Bruce's annotation for the inverse tangents (whatever they
>> are) and had the coordinates around the wrong way.
>
> Basic Trigonometry, something you would have come across early
> in high school maths - <http://en.wikipedia.org/wiki/Tangent>.
> atan is often uses in converting polar co-ordinates to rectangular
> co-ordinates.
>
>> The spreadsheet you
>> referenced below helped me find the error. I have not been through their
>> paper in detail but noticed the mention of "minor discontinuities". What's
>> the sense out there as to 2000 vs 1974 dE? My assumption is that while
>> still not perfect, 2000 is still an improvement on 1974.
>
> DE2000 is regarded as being superior, but there are mathematical
> issues that can affect some uses. The computation of a mean
> hue angle is a discontinuous operation, as two colors that have
> opposite hues, have a mean that could be one of two directions,
> also opposite. This is nicely illustrated in Figure 1 & 2 of the paper.
>
> Two colors of almost the same hue that lie either side
> of zero angle, also illustrate a problem with hue difference
> as well. The sign of the hue difference changes depending
> on the way the modulo difference is handled, and for DE2000,
> the sign of the hue angle difference affects the result.
>
> The sort of things where these discontinuities pose a problem,
> are where it's uses for numerical optimization, or the
> first derivative is needed for some calculation.
>
> Graeme Gill.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden