Chroma Subsampling & Bit-Depth EXPLAINED (8-Bit vs 10-Bit, 4:2:0 vs 4:2:2 vs 4:4:4)

Join us this week as we tackle the subject of chroma-subsampling and video bit-depth! We'll cover the difference between 4:2:0, 4:2:2, and 4:4:4 video, as well as 8-Bit and 10-Bit color depth. This topic can get a little techy and has some math involved, so we kept this lesson surface-level. After all, we are image creators... not computer scientists!
1:02 ► Intro to Chroma Subsampling
3:11 ► Breaking Down the Ratio
4:35 ► 4:2:0, 4:2:2, and 4:4:4
7:48 ► Bit-Depth: 8-bit, 10-bit, and Greater
9:22 ► Internal vs. External Recording
Stay up to date!
Instagram ► / lensprotogo
Facebook ► / lensprotogo
Twitter ► / lensprotogo
LensProToGo ► www.lensprotogo.com

[@chazfaz3595]

Some quick insights on how bit depth and color subsampling might affect post production... 1) If shooting in front of a green screen (to be able to remove the background), it's best use a camera that can save a file in at least 4:2:2. That last number is about chroma. Chroma Key software plugins use this chroma color information to separate the green background from the subject. Chroma Key software ignores the luminosity aspect. 4:2:2 provides the keying software a lot more subtle levels of color to work with than 4:2:0 does. That means guicker, easier, and cleaner automatic results (less laborious manual tweaking). 2) Same with color grading and other post-processing filters. Which provides more color manipulation options, and again means guicker, easier, and better results from 10-bit 4:2:2 than from 8-bit 4:2:0. 3) Also if shooting subjects with very subtle color gradations (where the color/luminosity changes very little over a big area), 10-bit will provide many more color options for your camera/computer to apply to the blend, and it should appear accurately. But if you are using 8-bit, there may not be enough subtle colors in the file to choose from to represent the blend adequately, and banding will occur. 4) Digital camera's are not really optimum for film noir B&W because of notoriously grainy results, but higher-end ones can do fine. 10-bit can provide better subtle results for portraying "someone hiding in the shadows" much better than 8-bit can. Then with a file that captures it all, care can be taken by manipulating these results to try to keep that visual effect in the shadows when saving compressed for the viewing medium. But if you filmed the shot in 8-bit you might not have captured the shadow detail in the first place.

[@jaymills1720]

What about 10-bit 4:2:0

[@chazfaz3595]

@@jaymills1720 I've not done any experiments. But I can almost guarantee that the extra bit depth of 10-bit color won't prevent chroma "destruction" and averaging done by 4:2:0 compression. I say this in context of using the chroma information for color keying with green screen. Of course10-bit color will allow the simultaneous use of 1 billion different colors on screen! But then sub-sampling still happens to every 8-pixel grouping in this scheme. With 4:2:0 sub-sampling, and wherever the 8-pixel groupings overlap the edge where subject meets green screen, and all 8 pixels share just 2 colors, how good is the key going to be?

[@nhuthan3802]

ur describing skills along w the visual representations r beautiful, exactly what a slow learner like me needs! thank uuuu

[@thesisunatorpodcast6850]

Thanks man, i was almost able to understand this on the first shot. I'm gonna come back with my buddies later.

[@8k60]

Thank you this is exactly what I needed

[@pinekill]

love how you explain things !

[@rdm5546]

Excellent description!

[@jorgengerstel]

What happened to the color red in this video?

[@user-tz1xu7sq9x]

i want to ask a question, does hdmi 2.0 truely support 4k 10bit 4:4:4? i noticed that all the monitor with hdmi 2.0 doesn't support this format. and only the displayport support this format.

[@AgentSmith911]

Will AV1 allow for video exports of rec2100 14 bit hdr 444 chroma subsampling videos? Or what about higher colors like 16 bit, 32 bit or 64 bit?

[@arduinoguru7233]

7:45 it not necessarily be bog file size since, with compress algorithm/codecs like H264 it take advantage of extra info to compress it in the next frame specially with 10 bit, I read that 10 bit video are smaller than 8bit video some cases 20% smaller .

[@retrowrath9374]

10-bit/12bit-YCb 4 : 22, REC2020, 4K and above for HDR content, it's mainly because of bandwidth being transferred access the HDM or DP cable, that's really all people need to know

[@JacobGagoOfficial]

thank you

[@vishnukv2755]

Create a video about video codec it's very helpful

[@ecromancer]

For bit depth, it is 2^x where x is the bit depth. 2^8 is 256 shades of red, blue, and green each. So the number of colors is (2^8)^3 = 2^24 colors for 8 bit color depth and 2^30 colors for 10 bit color depth. This is how they got 16.7 million colors for 8 bit and 1 billion for 10 bit.

[@GeorgeGraves]

IMHO - this is a very poor way to present this. Yes, it's mathematically correct, but you really should talk more about how the human eye looks at color, and how luminance is more important to detail then color, and compression sacrifices color information first. Talking about what 4:1:1 does to colors, ,and go from there. Even going back to video tape and film as a reference would give much more of a "real world" understanding to the subject. Source: I'm a retired Avid editor that started back in the u-matic days all the way to the intro to HD.

[@GBftw]

I thought it was good, I knew bit depth before and roughly what chroma sub sampling was, but now I could easily explain it back. More knowledge is better, but also takes more time and that is a problem for KZbin.

[@lnz3704]

Damn i cant agree more. Just about to write that before i see your comment. I believe on top of this theory, most audience want to see the real example, real difference in term of images. Side by side. Altho not sure if youtube support it

[@mikol.douglas]

I’m twice disappointed. First to know that his presentation and explanation is insufficient, then to see that I can only learn about BMWs on your channel.

[@GeorgeGraves]

@@mikol.douglas Whatever broski.

[@babiychuk]

Thanks for the video! What about color space - eg Rec 702 or Rec 2020. How does it connect to bit depth and chroma sampling?

[@mikafoxx2717]

Basically, that color space mostly just tells the device what the intent of each of the primary colours should be. 255 Green for sRGB would be a lot less monochromatic pure green, than say P3 or rec2020. Some software will clip off the unreproducible colors to what the panel is capable of. Or if it doesn't, it'll look desaturated as hell or more saturated than intended. Use sRGB for sRGB intent, unless you know better. You don't want to reserve most of your color bit depth for colours the monitor can't reproduce, unless you got 10-12 bits to make up for it, like HDR 10+ would.

[@jadefinchscene5644]

ok, since you are doing this on YT, it may be helpful to include information about YT compression and what happens to your video when you upload it. just might be a helpful reference point that many viewers may be able to grasp. good video though, not knocking it in any way. just keep in mind if you ever do an updated version. =)

[@jeffhalebopp]

If not pushing a color grade, does 10bit 4:2:0 mostly prevent color banding such as in the sky? Or do you need 4:2:2? Does 10bit 4:2:0 prevent banding more the 8bit 4:2:0? I am still not totally clear on this. Thanks for your help.

[@retrowrath9374]

It's down to bandwidth and HDR, basically use 10bit/4:22 for 4K and above for HDR content like games and movies, most don't need to know the technical side and differences.

[@KevinMullarkey]

I can confirm that with my camera (Lumix S5) and TV (Panasonic OLED 10 bit 4k HDR) shooting in 4K50p HLG 10 bit 4:2:0 BT2100 results in zero 'sky banding' as opposed to the same but with 8bit where banding is noticeable but not as bad with HD rec709 like you see on TV broadcasts.

[@aevoss]

Fantastic job! What’s a photo site? Are you trying to say the photo site is a pixel? I think your trying to say each box of the 8 boxes represents a pixel. That wasn’t clearly explained. I’ve never heard of a photo site. Thank you

[@mergeform]

1080i 4:2:2 has the same clarity as black and white 480i. Reminds me of vinyl vs early CD's.

[@ArcanePath360]

Do you have a Cadillac called Christine?

[@IB1GMANI]

Should I set my a90k to 10bit and what about the color temperature thanks for the help and good video

[@user-yo2nh8so5o]

I also, am Italian and talk with my hands. I see you.

[@mschocker77]

Still confused. I understand what it means but still don’t understand HOW it works. Examples would be SUPER SUPER helpful.

[@shutterangle596]

very hard to understand ... need to be more specific