Jump to content

Talk:ScRGB

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

color space diagram doesn't look right at all !!

[edit]

How can scrgb color space exceed 1931 color space ?

It doesn't make sense —Preceding unsigned comment added by 60.234.210.167 (talk) 00:29, 12 February 2009 (UTC)[reply]

Other color spaces like ProPhoto RGB color space also exceed the range of human vision and include imaginary colors. scRGB is probably the worst offender in this regard: about 80% of the color space consists of imaginary colors! That's the cost of maintaining backward compatibility with sRGB. OrenT (talk) 12:53, 31 May 2009 (UTC)[reply]

what would be the colorspace gamut if the RGB channels were encoded in floating point? 2A01:119F:21D:7900:E9D3:281:4EFC:8EA (talk) 05:37, 31 July 2019 (UTC)[reply]

The scrgb color space gamut is larger than the 1931 color space, it can represent colors that not only lie outside human perception, but lie outside reality as well. We call them imaginary colors. It is not a very efficient representation for the same reason, but it covers everything and a bit more...Carewolf (talk) 17:27, 31 July 2019 (UTC)[reply]

What are the sRGB issues?

[edit]

This entry mentions that ScRGB addresses "issues" of sRGB, but failed to menion what those issues are. Could this be added to this entry? Thank you! --24.249.108.133 17:55, 22 August 2007 (UTC)[reply]

  • The sources given are mostly blogs, which are not accepted as reliable per WP:SPS. It would be better to include more sources from reliable sources, such as magazines, newspapers, or published books. I am removing the "notability" tg for now, and will add a more specific notice. Sorted as part of the Notability wikiproject. --B. Wolterding 14:58, 7 October 2007 (UTC)[reply]

not sure where exactly this goes,but I am moving this notice to discussions:

"(this does not match table in the reference, somebody please check!)." —Preceding unsigned comment added by 193.222.161.5 (talk) 08:38, 27 October 2009 (UTC)[reply]

Accurate Math for non-linear

[edit]

I attempted to decode the Microsoft document linked in the article into a conversion from the RGB primary levels into the code values. However for the non-linear version, they had a table of example results and it does not seem to match what the math functions describe in any way. The text I inserted is my interpretation of the mathematical description. It would probably be a good idea to figure out what, if either, is correct.Spitzak (talk) 18:28, 27 October 2009 (UTC)[reply]

Arrgh, never mind, the table is correct. I had the sign of the 0.055 reversed for the negative case in my test code. Will try to fix the document. —Preceding unsigned comment added by Spitzak (talkcontribs) 18:44, 27 October 2009 (UTC)[reply]

Error on the graph or on the text?

[edit]

Is the phrase "the full range is -0.5 through just less than +7.5" correct? The graph shows at most +1.5, but as I'm afraid to introduce an error on an article I don't understand well and just passed by, I feel better asking here for someone to change it. Joaopaulo1511 (talk) 07:43, 24 February 2010 (UTC)[reply]

The poorly-labelled graph is presumably the classic constant-z slice through the CIE 1931 color space. So the axes are x and y. The article is talking about the red, green, and blue primaries. So you're talking about different quantities. --Imroy (talk) 08:09, 24 February 2010 (UTC)[reply]

The way I read it is that somebody meant to write 1.5 instead of 7.5. I'm not sure how that graph is scaled, but if it's correct (and linear) then 1.5 would be about right (the CIE shape appears within the correct bounds). The intro mentions coordinates in regards to x and y values on the CIE graph, in order to demonstrate that the location of the primary points are way outside what we can see. The red, green, and blue primaries are described by these coordinates. Lime in the Coconut 21:16, 2 March 2010 (UTC)[reply]

I am unsure if the graph is correct, but the "primaries" are in fact the corners of the small triangle. The larger triangle is the area that can be reached by using negative values for 1 or 2 of the primaries, while the small triangle is the area where all three are positive values. Larger positive values do not change the shape of the triangle as this is the gamut only for low levels of light when you are not clamping, any positive range limits results in a hexagon or eventually a triangle that is inverted from this one that eventually shrinks to a point at the brightest white that can be done.

I would expect the straight lines of the larger triangle to be about -.5 as far away from the parallel edge of the small triangle as the opposite point is. The diagram seems close but not exact. However this may be some non-linear XZ space (though that should mean the triangles have curved sides) or I just don't have the math right.Spitzak (talk) 23:05, 2 March 2010 (UTC)[reply]

I made a new graph after extensive and somewhat painful work trying to figure out how to plot the corners of the cube. xyY space is more complex than you might think, since the coordinates are X/(X+Y+Z) and Y/(X+Y+Z) and are meaningless if X+Y+Z is less or equal to zero. Slices other than X+Y+Z==constant produced very skewed triangles, so I finally chose this one because it is the largest value where sRGB is also a triangle.Spitzak (talk) 05:21, 10 May 2010 (UTC)[reply]

Acronym

[edit]

What does scRGB stand for? 75.118.51.238 (talk) 03:38, 19 January 2013 (UTC)[reply]

Less color resolution than 8 bit sRGB near 0.0?

[edit]

The article makes the claim that "The 16 bit scRGB(16) encoding is the linear RGB channels converted by 8192 x + 4096. Compared to 8-bit sRGB this ranges from about 1/2 the color resolution near 0.0 to more than 10 times the color resolution near 1.0." which seems to imply that a conversion from 8-bit sRGB to 16-bit scRGB would be lossy. This makes no sense as only 12 bits are needed to represent sRGB in a linear 0.0-1.0 range without loss of information and the increased range of 16-bit scRGB only takes away 3 bits of resolution. This would suggest that scRGB would have more resolution at any place in the curve compared to sRGB. To verify this I wrote a program[1] that losslessly converts the entire sRGB range into scRGB and back. Am I simply misunderstanding what that quote is communicating?

[1] http://ideone.com/FhLkBS — Preceding unsigned comment added by Det (talkcontribs) 20:52, 30 December 2014 (UTC)[reply]

It looks like the math got inverted. 8192/12.92/255 = 2.486 or more than twice the resolution. I suspect a math error inverted this and got 1/2. Fixing.Spitzak (talk) 00:27, 31 December 2014 (UTC)[reply]