Official 4:4:4 / Chroma Subsampling Thread - Page 2

Originally Posted by thepoohcontinuum

I agree. A vast majority of people who use their TVs as a full time computer monitor choose LCDs over plasmas (less power consumption, less image retention issues, 1:1 pixel mapping, size constraints, etc).

Still, I'll go ahead and create a table for plasmas. I won't spend much time actively researching plasmas, but I'll update it with any info that I randomly come across.

Originally Posted by pete4

This chroma issue was part of "input lag" thread started by gamers. At some point they decided, they didn't care about chroma and requested separate thread. My point is, the full RGB is important only if one uses PC for writing documents, spreadsheets and stuff with small fonts. When playing video on PC (HTPC), you can't be affected by this since video is already recorded at most at 422, probably even lower. The thread was started here, therefore should stay here and for example I don't read other forums, maybe except 3D and for viewing 3D, chroma is never an issue since they cut resolution at least by half anyway. BTW, this forum is among the most active forums here and has bigger exposure.

Originally Posted by 51dueller

Well I finally got the macro function working correctly on my camera.

Insignia NS-32E570A11 with DVI to HDMI


Insignia NS-32E570A11 with HDMI and audio


BenQ for comparison purposes

Originally Posted by thepoohcontinuum

When you say "420>444", do you mean those renderers upsample 420 footage to 444?

If so, I'm curious to know if there's any noticeable difference in quality. Given that 420 has some of the raw chroma information missing, going to 444 means you have to use an interpolation algorithm to fill in the gaps.

Originally Posted by Jan Willem

When the RGB 4:4:4 full range output of the renderer is left unprocessed by the display device (no digital processing, only digital-to-analog conversion for the panel), any form of chroma up-sampling in the renderer should give a better picture over nearest neighbor (especially if full picture resizing or aspect ratio correction is used by the renderer).

Originally Posted by Qaq

Correct.

http://screenshotcomparison.com/comparison/60342
Here some HQ vs Nearest Neighbor (ATI native) chroma subsampling algorithms. Do you see a difference? I don't with my "chipo LCD", but people with good plasmas do. Personally I'm not so familiar with these things, but here a response from someone who is:

Originally Posted by thepoohcontinuum

B4) Why do TV / Movie studios do chroma subsampling?

For the most part, it’s because of modern limitations in data storage capacity and transmission bandwidth.

For example, all blurays are subsampled down to 4:2:0 at the mastering studio (this is according to official bluray spec). As a result, most bluray titles have an average video bitrate of 25 Mbps. Assuming a 2 hour movie, you need a storage medium that can hold ~22.5 GBytes; which a bluray disc can do no problem. Now lets do 4:4:4 (no subsampling) on that same 2 hour movie. Going from 4:2:0 to 4:4:4 quadruples the amount of information required. So that 25 Mbps becomes 100 Mbps, and 22.5 GBytes becomes 90 GBytes; not even a dual-layer bluray disc can hold that much data.

If a bluray disc can’t hold it, don’t expect satellite/cable/fiber to transmit this kind of data either.

Note: the numbers used in this example are under review; will update when straightened out

Originally Posted by thepoohcontinuum

When HDMI audio extensions are enabled, *something* causes 4:4:4 to fail. I have no idea if this is a TV issue, or a video card issue, or a HDMI issue, or an EDID specification issue, or a combination.

Originally Posted by Lessen

Those numbers are rather wrong
Note: for all of this I assume 8 bit component depth.

4:4:4 RGB or YCbCr has 3 full-resolution components, so 8*3 = 24 bits per pixel total. So one frame is 1920*1080*3*8 bits = 49.8 Mbit. At 24 fps this results in a datarate of 49.8*24 = 1.2 Gbps.

Chroma subsampling only affects the two chroma planes, not luma. So luma is full resolution (1920*1080) while for 4:2:0 each of the chroma planes is half resolution in each dimension (1920/2 * 1080/2). So one frame of 4:2:0 is (1920*1080 + 1920/2 * 1080/2 + 1920/2 * 1080/2) * 8 bits = 24.9 Mbit, half of 4:4:4. At 24 fps this results in a datarate of 597 Mbps.

But comparing uncompressed bandwidth numbers is only meaningful for HDMI, not Blu-ray or HDTV transmission as those are always compressed. And over HDMI, only 4:2:2 and 4:4:4 are supported, not 4:2:0.

It's a little hard to compare how different subsamplings affect compression, but to start with chroma has much less information than luma no matter the subsampling and thus compresses extremely well, to the point that the great majority of the bandwidth goes to luma data. Anyway, with modern codecs (H.264), 4:4:4 does not take significantly more bandwidth than 4:2:0, and in some cases (upsampling a 4:2:0 to encode as 4:4:4) might actually take less bandwidth due to the larger prediction modes available to 4:4:4 and better interpolation filter (H.264 only.)

The main reason movies are all 4:2:0 is simply because that's the way things were done long before MPEG (it's the original lossy compression, by a factor of two.) The main reason I can think of for modern TVs to subsample is to reduce the required processing power for all the processing done. Or maybe for a PenTile matrix, but no TVs have them as far as I'm aware.



RGB support (4:4:4) is required for all HDMI devices. This is the native format of the framebuffer in computers, making YCbCr support completely optional. There is absolutely no reason for the video card to convert to YCbCr, let alone subsequently subsample to 4:2:2 before sending the data over HDMI. The only place the conversion and subsampling would logically take place is in the TV.

I'll have to try my TV later, I'm honestly quite surprised that getting 4:4:4 from TVs is hard.

Originally Posted by Lessen

It's a little hard to compare how different subsamplings affect compression, but to start with chroma has much less information than luma no matter the subsampling and thus compresses extremely well, to the point that the great majority of the bandwidth goes to luma data.
....
The main reason movies are all 4:2:0 is simply because that's the way things were done long before MPEG (it's the original lossy compression, by a factor of two.) I'll have to try my TV later, I'm honestly quite surprised that getting 4:4:4 from TVs is hard.

Originally Posted by thepoohcontinuum

Thanks for that very informative post (and you too pete4). Question though... According to the post here, most bluray movies have a video bitrate of around 25 Mbps, how does that number relate to your theoretical calculations? Going from 4:4:4 1.2 Gbps to 4:2:0 25 Mbps is one heck of a compression ratio.

Is it because there's two-factor compression going on? One compression during chroma subsampling (at the recording source), and one during the encoding process?



I'm only aware of one TV that has a RGBW subpixel layout, and that is the Westinghouse LD-3280 (pic). Is this a true PenTile panel? Who knows. Someone on AVForums mentioned it was a CHIMEI LCS-MVA along with the spec sheet (link).



Yup, the search for a 4:4:4 capable TV is a hard and arduous journey As pete4 mentioned, you would think a TV would have enough "intelligence" to know its receiving a 4:4:4 signal and bypass all the TV/movie processing junk. But that's unfortunately not the case.

And please do check your TV for 4:4:4, the more test results we have the better.

Originally Posted by thepoohcontinuum

There's probably some post-processing effect that's hard-coded and can't be disabled. I know with some 2010 LG's, they had an option called Edge Enhancer, which is basically a Sharpen function. Even though this was disabled in the menu, the TV was *still* sharpening things. This especially screwed up text since text pixels were being oversharpened and created a halo-effect around the letters.

Originally Posted by Qaq

I can confirm this with my LG LCD. In PC mode (RGB) Desktop looks to sharp, photos look sharper and video affected a little bit too. Thats why I think its better for these TVs to use YCbCr transport in PVP mode over videocard's overlay (like Arcsoft TMT does). HDCP-encripted video streams seem not touched by video driver (directly at least).

Originally Posted by thepoohcontinuum

Thanks for that very informative post (and you too pete4). Question though... According to the post here, most bluray movies have a video bitrate of around 25 Mbps, how does that number relate to your theoretical calculations? Going from 4:4:4 1.2 Gbps to 4:2:0 25 Mbps is one heck of a compression ratio.

Is it because there's two-factor compression going on? One compression during chroma subsampling (at the recording source), and one during the encoding process?