Tuesday, July 22, 2014

Chroma Subsampling

Last week I started off with the 4K image of my backside...

I am going to go a bit further into the rabbit hole...

One of my biggest objections to the whole 4K thing comes from this topic of chroma sub sampling...

The whole idea of it is that we as humans stink at seeing colors, but can determine brightness quite well.  That is to say in more technical speak, we can determine luminance information much better than we can determine chrominance information.  Because of this fact, we have figured out that if we create an image with a certain resolution for the luminance side of the world, we can decrease resolution in the chrominance side of the world, and still have a relatively good experience.

I know several of you right now are saying that color is just as importance as brightness, but I would tell you that for the most part you may be inaccurate.  There are several standards out there right now that use chroma sub sampling and have taken off like wildfire.  One standard that has used chroma subsampling is DVD Video, but that's small potatoes to the largest... which is Blu Ray.

That's right the Blu Ray video spec calls for chroma subsampling.
You have not watched a movie on DVD or Blu Ray that is full bandwidth ever.
So the amazing visuals you thought you were seeing are NOT all that great.  They could be made better.

Blu Ray (and DVD incidentally) uses what is called 4:2:0 chroma sub sampling...
What is this 4:2:0?

The first number represents the horizontal Sampling reference
The second number represents the number of chrominance samples in the first row of J pixels
The Third number  is the number of chrominance samples in the second row.

For your convenience, A nifty Chart:
4:1:1     4:2:0     4:2:2     4:4:4     4:4:0
1234 J = 41234 J = 41234 J = 41234 J = 41234 J = 4
(Cr, Cb)1 a = 112 a = 212 a = 21234 a = 41234 a = 4
1 b = 1 b = 012 b = 21234 b = 4 b = 0
¼ horizontal resolution,
full vertical resolution
½ horizontal resolution,
½ vertical resolution
½ horizontal resolution,
full vertical resolution
full horizontal resolution,
full vertical resolution
full horizontal resolution,
½ vertical resolution

So as you noticed, the full resolution is noticed as 4:4:4...  You will also notice that our amazingly awesome 1080p video from blu rays, is actually 1/2 resolution vertical, and 1/2 resolution horizontal.
Via the Magic of Math, you get 1/4 the resolution as the 4:4:4 chroma subsampling.

You are literally using a full resolution black and white image in combination with a 1/4 resolution color differential image to get the resultant image...

So you have never seen a full resolution image from a blu ray (or a DVD player for that matter).

Now how does this relate to may backside, and 4k?  well hopefully you still don't want to see a 4K resolution image of my backside, but as it pertains to 4K, it comes down to bandwidth...

The Math of figuring out bandwidth required for signal rates isn't quite as easy as it might seem as there are plenty of factors out there that would seem like nonsense, but when you get down to the nitty gritty, it makes perfect sense.   If you would like to see the math, let me know and I will do it out for you.

The end results are:

4:4:4 1080p 60 Frames per Second data rate, with 8 bit color depth, and our bandwidth is 4.46 Gbit/s
4k Video (which is essentially 4x 1080p) at all the same stipulated factors requires bandwidth of

17.84 Gbit/s!!!!!!

Our current HDMI 1.4 spec limits us to a max data rate of 10.2 Gbit/s...

Hopefully most if not all of you could agree that 17.84 is greater than 10.2...

So how do them TV folks tell us that 4k is actually happening...  They do it by chroma subsampling...  If they only include 1/4 of the color information, you can get a passable 4K signal...

The result is that the 4K video at 60 Frames per Second with 8 bit color depth and 4:2:0 chroma subsampling weighs in at about 9 Gbit/s...  Less than the 10.2 limit.

So again, it all comes back down to this...
4K in full resolution simply does not exist for the masses yet...

Let me know what you think about the proliferation of less than full resolution color sampling...

Oh, yeah, and apparently our eyes are terrible at determining color resolution.

original                             Luminance                      Chrominance

1 comment:

  1. I would like to see the math. Please show all your work.