Since the eye is more sensitive to brightness than color, having a colorspace that separates out brightness from color is more efficient than RGB.

If you were to make the most efficient 16bit direct color mode, what colorspace (YUV, LAB, etc) would you choose and what would the bit allocation be for each channel (e.g. Y8U4V4)?

Thanks!

Personally i would go for LAB value's because LED's get binned using a LAB specific coordinate system.(LCD screens, and such)

Greetings,

Thomas

An elegant fit would be using a 3 plane screen of bytes for this.
1. Plane U
2. Plane Y
3. Plane V

If Plane 1 and 3 are lowres a plane 2 Hires then it would match this format.

For a desktop RGB would work, for some applications however i would rather use a different angle.

Like i said for a game HSV would suit good.(especially 2d work)

For video Thomas is right about it being native to the files

@Gunnar could it be that simple and work?

It has to be looked at more as what are you gonna do then this is it One thousand.
For each application has it's own best fitting colour scheme and it can prolly be set in the file header or something like that.

1st word - 8bit Y, 8bit Cb
2nd word - 8bit Y, 8bit Cr

OR:

1st word - only Y for 2 pixels
2nd word - Cb and Cr same for 2 pixels.

second idea would be more efficient to operate with software i think.

it would be GREAT to have 3 8-bit planes, but luminance with full resolution and chrominance with halfxhalf resolution (or halfxfull meyby).

GREAT way to optimize memory bandwidth and amount of calculations.

even greater to play movies, YUV=>RGB and desubsampling done by hardware easy way.

useful not only for video i think.

If you still plan on using a VGA monitor, your stuck with RGB since that is the bus and protocol.(analog)

DVI works a little more, interesting. EXTERNAL LINK 
Anyhow honestly No matter what we'll do with the file of the image or a video file.
We have to send it to the display at some point.
If it is for a program for movie editing i can understand why people would prefer to stick with the file native format.

and to remain a little more on topic R5,G6,B5 since our eyes are more sensitive to green.

No offense intended to anyone guys, but the Amiga engineers must have worked thinking from how their output devices worked if you look at the standards.

Hence the need to be able to place the graphic palette use in the file header.
And i know your right Samuel about the need for alpha blend.

It's just a case of Mr. Z uses application Y using file type X.
It's not one to rule them all, we need to be able to use them all.
let the user decide what he needs to get the job done.
Let a designer or the graphic artist decide what would give the best effect for their work.

What we should be asking our self is the 16bit colour space enough for what i need it to do with it?
In somecases it will fit the bill fine, in others it won't.
Just give the User the tools to work.

(silly mode)
R8,A8 bit-plane 1
G8,A8 bit-plane 2
B8,A8 bit-plane 3

This could be handy if the editor needs to only check only colour(colour filter)
(/silly mode)

Greetings,
Thomas

Wojtek i'd advice looking at the thread name.
YUV is only a small portion of it.

YUV spread over two or more pixels is a good compromise for moving images but not really suitable for desktop/GUI usage in my opinion.  Look at the 'bleeding' on anything with a GUI in HAM mode as examples. 

The idea is with 16bits to get the best spread of human perceivable colours possible and that was the reason why LAB was defined so i'd look at that first.  Elsewhere there was talk of YCoCg also for textures but maybe it is relevant here as well?

Any advice on where to place the 0 degrees on Hue?

Again, RGB is off-topic.  It was disallowed in the beginning because it is inefficent.  No more RGB, thanks.  And it is to be 16bit too.

I think I like something like LAB. The Atari Jaguar had a CRY (cyan, red, lightness) which was nice.  They claim that the eye is least able to distinguish shades of green.  Although I think having a push-pull between red-green and blue-yellow (like LAB) is better than their choice of red-cyan and blue-yellow, because R-G and B-Y seem to be visual opposites. 

I think 8 on brightness would be best leaving 4-4 on the color, because the eye is most sensitive to brightness over color.  (So H8 would be out, because that is putting the most bits on color where the eye is weak, not brightness where it is strong.)  Of course, even YUV @ 655 would be much better than any RGB in 16bit.