What is Chroma Subsampling | LSE - E05
Vložit
- čas přidán 12. 07. 2024
- Live Streaming Essentials - E05
In this episode, Leo answers the question of what is chroma subsampling. He explains everything you need to know about Chroma Subsampling, a compression technique from decades ago that is still used in modern videos.
What is 4:4:4, 4:2:2 and 4:2:0?
Learn how you can reduce the file sizes of your video and what it means for the image quality.
This video is part of his video series “Live Streaming Essentials” in which Leo talks about the fundamentals of live video topics to help you get started and grow beyond the beginner level.
Turn your Mac into a live video production studio with mimolive, download your free trial:
▸ mimolive.com/
0:00 Intro Chroma Subsampling
0:32 History Background
1:14 4:4:4
2:11 4:2:2
2:43 4:2:0
4:25 Recording 4:4:4: in mimoLive - Krátké a kreslené filmy
This viking just did the best video about chroma subsampling!
As it is said "If you can't explain it simply, you haven't understood it yourself". Fucking awesome explanation! 👍
this is the best video i have seen on the subject
This is the best explanation I have seen on the whole internet
Really well explained
Thanks
this video is old but still the best explanation video of Chroma Subsampling!!!!!!!!
Probably one of the best 5 minute explanations of anything I've seen. Great stuff!
Glad you enjoyed it!
Great explanation - you simply put what my professor couldn‘t explain to us in an hour, thanks very much!
Great presentation - so much better than many textbooks, and much easier to understand.
Excellent video! Best description of chroma subsampling I have seen so far. Excellent work!
This is a great lesson, thank you very much! 👍
Brilliant explanation. Simple, to the point, with useful visuals. Finally understood the concept!
OMG! Holy smoke! this is so far the simplest and easiest explanation for YCC colour format on the internet I have found so far! others are focusing too much on the channels like the Cb and Cr which I don't quite understand. but this video got me to understand all of the concepts with just one watch (and of course with some basic information from the "Computer science" channel) Thank you guys for making this video. I am not saying others are worse but it is just not a friend for people who lacks the basic information/knowledge in colour space and stuffs like that. Thank you again!
Thanks for the kind words.
EXCELLENT Tutorial. Thanks.
Simple and straight forward explanation !!!
Amazing video - possibly the best I've seen in this subject. Thanks for this!
Glad you enjoyed it!
Well done! Best explanation on this 4:4:4. I never understood it before, and other videos you can see they don't know what they talking about. This is well explained. I owe you a beer. Just fly me over to your town, and the beer is on me!
Thank You! Good job!
This was so informative! And I enjoyed the comedy.
thanks for the AMAZING explanation
thanks for the background information, feel a little bit smarter today.
Great to hear!
this is an excellent explanation!
Really helpful to understanding Chroma Subsampling with animation.
really helpful video, understood it right away, thanks
Very good explanation.
All other videos are confusing. This one is very clear
Thank you so much for this! I was losing my mind at work the other day, because my render would come out pixelated. I spent about 4 hours trying to figure out what was going on, trying out different software, computers and resseting settings only to find out it was becasue of the rare situation, where the pixels share the same color. I did have to use a different color, because I found out about this later on, but using 444 on the original color made it look correct!
Great to hear!
Superb explanation!
Thank you for explanation of exact meaning of numbers!)
Glad it was helpful!
verrry gud video, i will share thaad with my frind ranshid in india! thaank yuu so much !
Great Information - THANK YOU !
Our pleasure!
Thank you so much!
Great explanation
Thanks very much.
Thx mate!
best and easy explained thanks
I subscribed after 1 minute
Great explanation
me2!
Thanks it's so helpful
Holy hell that was helpful. Great video!
Glad to hear it!
fantastic video finally it makes sense also lol that this guy is giving a tutorial on greenscreen and keying
Leo I'm sure you hear it all the time - buy boy do you look young ! Keep rocking that young look. Like the beard. Oh, and good video. I keep seeing xxbit 2:2:2 and xxbit 4:4:4 and had no idea what they were talking about.
Thaanks!
great video, hope
your channel grows
I appreciate that!
I overclocked my 144hz monitor (BenQ XL2430T) to 210hz but I can only use it in 4:2:2, and that's why I'm here... Thanks alot for the explanation!
thank you
I loved this video. Thanks! Best explanation!
Question, which one should I use on my pc with 4k resolution, RGB FULL or 4:4:4? Edit: for HDR and 12bit color
Just want to mention having a 4k camera is going to give you a great chroma key if using 4:2:0, its about the 8.3 million pixels that are going to key your subject better than the color sample loss ! maybe in the big screen it would matter but doubt that they just have the money for that, thats all !
Wow!
Thats a great video
Thank you.
Well done, kid. Well done.
thanks
Bro -- Any detailed video on Video Codecs, Raw, Log & Uncompressed video please
thank ou
How is the shared chroma value decided? Is it an average approximation of all the chroma values to be shared, a more complex algorithm, or is it just picking one of the existing chroma values (and essentially saying, "okay, we're using this one") based on some arbitrarily defined pattern?
Other than that, good video. I now somewhat understand most of the wikipedia page on the topic.
Hey,
this depends on the implementation of the Chroma Subsampling algorithm, but for example with 4:2:0 it is *usually* done by averaging horizontally and then vertically to get one value!
But you can also write your own implementation and algorithm and use your "okay, we're using this one"-approach.
We are happy our video helped understand this complicated topic. Don't forget to check out our other LiveVideoEssentials
the best i love u
In short you never compromise with the brightness and keep reducing the color component making use of the limitations of the human eye's color perception abilities.
At the end of the day all these clever techniques exist because of only one reason which is limitations in computational power and network bandwidth.
So when we reach that stage where we have almost limitless resources and bandwidth at our disposal, will these techniques still exist ? or I am being too optimistic about future of quantum computing ?
I think narrative would have changed by then I guess and people would be talking about why compromise on quality when we have such ultra fast networks.
Brightness is the biggest factor for HDR but most people don't know that oled displays have different HDR nit values than any backlit display due to the perceived brightness from the much superior contrast ratio. I think an OLED has to put out 550 nits to do HDR properly while any backlit display needs to be around a thousand nits or more.
What would be the difference between RGB 444 6 bit and YCBCR 422 10 bit
nice
so, when is recommended to use 4:4:4?... i feel like it make no sense to use it.
PD... Thanks for making this video, Im really bad at this subject in the university :(
4:4:4 makes sense for high-professional work, such as professional archiving data, or science.
Silly question: Could 4:2:0 be referred to as 2:1:0? And 4:4:4 be 2:2:2 or 1:1:1? Thanks!
Dear Shawn, if you have a good reason to change the international convention of naming subsampling methods - you could do it. I won't recommend it.
How does DSC factor in?
Doesn't things change with HDR where your dealing with 1024-4096 shades of color and 1000-10,000 nits of peak brightness. As well as a color gamut like Bt.2020, so the lossy nature of chroma upsampling becomes even more apparent?
I would have to agree. Your display would need to be able to handle the full color gamut and IMO brightness is the biggest factor on how good HDR material looks. OLEDS need around 550 nits to properly show HDR well I need backlit display needs to be either a thousand nits or slightly over a thousand nits. This is due to the perceived brightness of oled's due to the much superior contrast ratio of any backlit TV no matter how many zones it has
I feel like RGB is more suited for computer applications and subsampled content is more for video and television media. Is this a good distinction or is this an oversimplification?
RGB was causing black crush only when there was a mismatch in how the Blu-Ray (or other media device) was outputting and how the TV was had it's Black level setting.
RGB comes in two variations, Limited (16-235) and Full (0-255). And whichever you set as an output, the TV must match it in it's Black level settings. Otherwise you'll get either washed out blacks (when output is RGB Limited but TVs black level is set to Full), or crushed blacks (when output is RGB Full but TVs black level is set to Limited).
As for how the content is passing through from mastering phase until you watch it, everything starts out in YCbCr (also known as YUV or YCC) and ends up in RGB to be displayed by the individual pixels of your TV (unless you have a CRT TV, in which case it remains YUV).
The bottom line is, everything digital is RGB, everything in the movie/TV/broadcast industry is YUV. Everything that's being displayed by a computer is generated in RGB, everything that is being sent to your TV (be it, TV, Blu-Ray, etc.) is YUV.
So at some point from the post-production/mastering until you watch it on the TV, you need to convert from YUV to RGB. The only question is, when do you do it. And the shortest answer is, as late as possible. Preferable at the very end, at the TV side.
Chroma subsampling was something added to decrease the bandwidth required to push picture information in any given media format. As 4K resolution goes, you can't send for example HDR 4K @60Hz as RGB, there's not enough bandwidth in the HDMI spec.
As it's been said, all movies and TV shows that comes from either a Blu-Ray player, or streaming services are "done" in YUV 4:2:0. So the ideal way is to give that exact signal to the TV, and then the TV will take that signal and start working some magic and transform it into RGB to send it to each individual pixel.
Now because of some weird and complicated reasons, HDMI standards limit the amount of available configurations of resolution, bit depth, chroma subsampling and color space. So, for example you can't send 4K @24Hz, 10bit, YUV420, Rec.2020 but you CAN send 4K @60Hz, 10bit, YUV420, Rec.2020 or 4K @24Hz, 12bit, YUV422, Rec.2020, both of which require more bandwidth. PLEASE NOTE THIS APPLIED TO HDMI 2.0B AND NOT 2.1
Because of this, with the recent launch of 4K content, there's been a lot of talk about this, because UHD Blu-Ray is served as 4K @24Hz, 8bit, YUV420, Rec.709 and HDR UHD Blu-Ray is 4K @24Hz, 10bit, YUV420, Rec.2020 the exact setup that can't be send via HDMI to a TV. So instead of sending the content untouched to the TV, you have to do some chroma upsampling on the media device and upsample from YUV420 to YUV422 or YUV44 or RGB (if the framerate allows it) then send it to the TV where another conversion will take place.
So, to answer your initial question, always use YCbCr when the content is Movie/TV/Broadcasts, always use RGB when the content is Games.
@@JoshFisher567 I learn from ntt-electronics com
" The difference in Image Quality between 4:2:2 and 4:2:0 Chroma Formats in Cascaded Codec Connections "
4:2:2 Keep Chroma still in range.
4:2:0 Shown more degradation after encode/decode operation.
@@JoshFisher567 it supports 4K bro i see on Wikipedia
At 4:45 you talk about the data savings
4:4:4 has 16 values
4:2:2 has 8 + 4 values so 12, so surely this is 75% the data (not 2/3rds)
4:2:0 has 8 + 2 values so 10, so surely this is 62.5% the data (not 50%) ?
Am I missing something?
The color data used in chroma subsampling is represented in YCbCr color space. The Y channel is luma, while Cb and Cr are the red and blue chroma color channels. With 4:4:4, each pixel has all three channels, so that's a total of 12 values for 4 pixels. With 4:2:2, only half of the pixels keep all 3 channels, so that's 3+3+1+1=8 values for 4 pixels (66% compression). With 4:2:0, only 1 in every 4 pixels retains all 3 channesl, so that's 3+1+1+1=6 values for 4 pixels (50% compression).
hello plz can you help me i'm using 2 monitors and the main one is using vga to hdmi cable and the 2nd is coneccted to the other port i've recognized that movies and games became so much dark i search this again and again until i found what is called the (RGB) in nvidia control panel i change it to YCbCr444 and colors are back to normal but when i install a new game
the first monitor in going to sleep id why plz can you help
Could you please enable automatic-captions?
There are 4 2 2 8 bits?
4:57 how are data savings numbers obtained? I thought 4:2:2 should be 50% and 4:2:0 should be 25%.
Each pixel has 2 components: luma (Y) and chroma.
Chroma is the combination of two sub-components: the blue-difference (Cb) and the red-difference (Cr).
Therefore, in total there are 3 components for each pixel: luma (Y), blue-difference (Cb) and red-difference (Cr).
In 4:4:4 we are dealing with a 4x2 pixel matrix, where each pixel is formed by the three components described above. In total there are 4x2x3 = 24 pieces of data.
In 4:2:2 there are 8 luma components in total and 2 Cb and 2 Cr components per line. This totals 8 + (2+2)*2 = 16 pieces of data. This is a 33.3% reduction when compared to 4:4:4.
In 4:2:0 there are 8 luma components in total and just 2 Cb and 2 Cr components sampled from the first line (and repeated, but not sampled, in the second line). This totals 8 + 2 + 2 = 12 pieces of data. This is a 50% reduction when compared to 4:4:4.
ClarkVideo Wow thank you! I understand now. Another question though, how common is YCbCr used compared to RGB? Where will I see YCbCr and RGB usage respectively?
Very informative video but I think it's a litte ironic that the green screen in this video looks kinda trashy.
But at least I learned something new, thanks for making this.
So ,After all, which one saves more Dandwidth?
4:4:4. 4:2:2 or 4:2:0?🤔
The pure amount of pixels / data follows this form: 4:4:4 > 4:2:0 > 4:2:0.
@@Boinxsoftware which have more the best quality when watching videos and movie
the sound of the brush at the beginning ist annoying af :D but good video anyway (Y)