Talk:Luma (video)

NTSC origin

I don't think this word was created in the NTSC (~1953) era. For example, some of the members of the original NTSC committee published a book (under Hazeltine laboratories) explaining how NTSC works. They use the term luminance, and make note of the incorrect order of operations.

also see [1] Glennchan 06:06, 16 November 2006 (UTC)

601 vs 709 numbers

The coefficients listed even before Glennchan's big edit are from the 601 gamma-compressed luma versus the 709 linear colorimetric luminance. I think the latter numbers are probably not what we want here, and are unnecessarily confused with the former. Is this just a misinterpretation of what the 709 spec is saying, or what? Dicklyon 23:48, 19 November 2006 (UTC)

Rec. 709 also uses gamma-compressed components in calculating luma. The transfer functions on Rec. 709 are different; in practice Rec. 709 and Rec. 601 material is interchanged without translating/converting the transfer functions. I'm not sure what you're saying about the latter numbers confused with the former. They are clearly two different sets of numbers????Glennchan 01:17, 20 November 2006 (UTC)

I'm saying that the weights shown associated with 709 are the weights for linear R,G,B. Can you point me at the 709 documents that suggest using such numbers with gamma-compressed R, G, and B? Dicklyon 01:28, 20 November 2006 (UTC)

The Rec. 709 documents would have it, although you gotta pay for em. What you're looking for is implicit in the information presented at [2]. Rec 601 <--> 709 is interchangeable if you use a matrix to account for the different luma co-efficients. Consumer devices also omit this matrix. This wouldn't be possible unless Rec. 709 used gamma corrected components. Do you believe me now?

Perhaps you are confused since those weightings are correct for linear light RGB components. Rec 709 luma simply uses those same numbers, but with the *wrong* order of operations. This can cause very incomplete luminance/color separation for fully saturated colors (like an error of 60%). Poynton shows the calculations at [3].Glennchan 08:09, 20 November 2006 (UTC)

Forgive me if I remain somewhat incredulous. You may be right, but I'd like to see it from someplace other than an old Poynton rant. Dicklyon 15:55, 20 November 2006 (UTC)

[4] I have no idea if that document is supposed to be on the internet, but it is. If you look carefully, it details the Rec. 709 transfer function as well as Y'CbCr encoding.Glennchan 04:21, 22 November 2006 (UTC)

Fascinating. So maybe we should list all the coefficients from Table 6-9. What is 709-5? A section? or a variant of 709? Is this really the specification used in HDTV? Go ahead and fix it as you see fit. Dicklyon 05:49, 22 November 2006 (UTC)

As each standard is updated, they change the number they stick at the end. Rec. 709 is a set of recommendations about HD formats AFAIK. Most HD acquisition formats and broadcast formats do follow it. I think the transitional 1035i formats were created before Rec. 709, and they do not follow Rec. 709.Glennchan 05:06, 24 November 2006 (UTC)

RGB values of example colors with (supposedly) same luma

It seems to me as if the tables of RGB values do not show colors with the same luma, but (roughly) the same relative luminance. E.g. looking at the CCIR 601 table for light colors:

R'	G'	B'	Luma	Luminance
255	147	255	0.7514	0.5877
255	170	170	0.7663	0.5863
211	211	0	0.7331	0.5841
192	192	255	0.7811	0.5886
200	200	200	0.7843	0.5860
122	244	0	0.7047	0.5918
0	235	235	0.6460	0.5857
0	250	125	0.6314	0.5857
0	255	0	0.5870	0.5870

The other tables show similar discrepancies. --Fhoech (talk) 12:09, 3 April 2017 (UTC)

There's two approaches to fixing the discrepancy, either change the RGB color values in the tables to actually make them the same luma so that the existing description becomes correct, or change the description to correctly describe the existing tables. I opted for the latter approach as it's less work. Fhoech (talk) 10:28, 4 April 2017 (UTC)

From what formulas these values have been calculated? Now in tables are NTSC (1953), but nowhere in the text. PawełS (talk) 19:00, 29 November 2017 (UTC)

As per section 2 of the article, for CCIR 601 (uses NTSC 1953 coefficients), Luma Y′ = 0.299 R′ + 0.587 G′ + 0.114 B′ and conversely rel. luminance Y = 0.299 R + 0.587 G + 0.114 B, where R, G and B (and R', G' and B') have been converted to the range 0..1, and R, G and B also had gamma expansion applied (i.e. taken to the power of 2.2) Fhoech (talk) 23:33, 30 November 2017 (UTC)

For 0,250,120

Luma Y′ = 0.299 R′ + 0.587 G′ + 0.114 B′

0 + 0.587*250^2.2 + 0.114*120^2.2

noramlization 0.587*0.980392^2.2 + 0.114*0.490196^2.2 =

0.587*0.9574 + 0,114*0,2084 =

0.562 + 0.0238 = 0.5858 . Correct? PawełS (talk) 11:39, 2 December 2017 (UTC)

0,250,120 are gamma-compressed components, so to calculate luma Y', gamma expansion must not be performed. Gamma expansion is only required to calculate rel.luminance Y from gamma-compressed components. Fhoech (talk) 13:44, 31 December 2017 (UTC)

@Fhoech: In the article there is ,,where the prime symbol ′ denotes gamma compression". I want to replace this with ,,where the prime symbol ′ denotes gamma compressed values different from the sRGB nonlinear values which use a somewhat different gamma compression formula, and from the linear RGB components." Based on Grayscale#Luma_coding_in_video_systems.

I wanted to show different colours, (with RGB values I can choose in a graphic program, gamma compressed, right?), with the same brightness so I think I should move BT 709 tables from here to the Grayscale article with recalculated values according to the formula of gamma expansion. Is it a good idea? Now I don't know whether there is a point to keep tables showing NTSC colours on sRGB monitors. PawełS (talk) 22:52, 9 January 2018 (UTC)

That seems reasonable to me. Not sure what to do about the tables. The appearance of the R'G'B' values still depends on the characteristics of the actual display device (and possibly color management in the browser, which may interpret the values as sRGB regardless). Fhoech (talk) 21:17, 14 February 2018 (UTC)

Can we put somewhere in the article the NTSC coefficients, and state that they are the NTSC values? Or maybe they could go on the NTSC page itself. Gah4 (talk) 00:07, 5 April 2018 (UTC)

short

I always thought luma was short for luminance, and chroma short for chrominance. In a long discussion, it gets tiring to see the long words so many times. Also, I always thought that gamma corrected signals were to match the characteristics of CRT display devices. Gah4 (talk) 10:15, 21 May 2019 (UTC)