Hard chroma subsampling, will it ever die? - AVS Forum
Forum Jump: 
 
Thread Tools
post #1 of 2 Old 01-13-2013, 07:36 AM - Thread Starter
Member
 
Eric D.V.H.'s Avatar
 
Join Date: May 2002
Location: Monterey, California, USA
Posts: 21
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Yes yes, shifting from RGB to YUV and back is still a good idea, and back in the '50s, simply deleting whole lines of pixels from the yellow/green and red/blue channels was (like interlaced scan) a pretty clever compromise.

But nowadays? We don't worry about the bandwidth impact of resolution going up to 1080p60 and beyond, even over the internet. Since the flexibility of destructive compression at variable bit rates allows us to more precisely choose what to sacrifice, HD is typically streamed well below the bitrate of SD DVDs without looking too shabby. Heck, if fractal compression ever catches on, the notion of frame resolution will mostly go out the window in favor of bandwidth consumption.

In spite of this, every consumer codec from JPEG to h.264 is chroma subsampled (and almost always at godawful 4:2:0,) producing increasingly huge blocky multicolored jaggies that absolutely can't be removed.

Imagine if, instead, after the RGB-YCbCr stage, codecs simply biased to compress each chroma channel at higher ratios (like HD greyscale.) Then, when there was something, even on part of the frame, that compressed really well but produces horrible chroma artifacts (like a slow pan over thin, sharp, colorful line art) it would often be in 4:4:4.

Even people that dive into the intricate minutia of codec architectures, like Jason Garrett-Glaser, seem to just take chroma subsampling for granted.

Has anybody in a standards body ever expressed interest in fixing this hoary old hack?


Eric,
Eric D.V.H. is offline  
Sponsored Links
Advertisement
 
post #2 of 2 Old 01-19-2013, 07:49 PM
AVS Special Member
 
Richard Paul's Avatar
 
Join Date: Sep 2004
Posts: 6,955
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 29
Due to the nature of digital video compression going from 4:2:0 to 4:4:4 wouldn't cause much of an increase in bit rate. The main reason that 4:2:0 is still used is because it reduces the complexity of the decoder and the amount of memory that the decoder needs. Since companies count pennies when it comes to designing their products this is the main reason that companies still advocate for the use of 4:2:0 chroma subsampling. In my opinion both chroma subsampling and interlacing are relics of the days of analog video but for economic reasons there is still a good amount of support for chroma subsampling.

I have a bit of hope that 4:4:4 consumer video might happen with HEVC. There is a chance that a 4:4:4 consumer profile could be released in the second version of HEVC but so far nothing has been announced. The good news is that HEVC does have the Main 10 profile which is a 10-bit consumer profile that would greatly improve video quality.

Also if you get rid of chroma subsampling it makes sense to use RGB video. Converting from RGB to 4:4:4 YCbCr and than back to RGB causes an unrecoverable loss of information. Charles Ponyton, who is the author of a book called "Digital Video and HD", wrote this online article on converting between RGB and YCbCr for studio video and here are some quotes from it:
Quote:
...
If each of the Y’, Cb, and Cb components has 8 bits of precision, then obviously the entire Y’CbCr cube has the same number of codewords as 8-bit R’G’B’. However, it is immediately obvious from the appearance of the transformed R’G’B’ unit cube in Figure 3 above that only a small fraction of the total volume of the Y’CbCr coordinate space is occupied by colors! The number of colors accommodated is computed as the determinant of the transform matrix.
...
If R’G’B’ is transcoded to Y’CbCr, then transcoded back to R’G’B’, the resulting R’G’B’ cannot have any more than 2.75 million colors!
...
Stay in R’G’B’ if your production situation permits. The first conversion to Y’CbCr will cause an unrecoverable loss of 75% of the available R’G’B’ codewords, and the first subsampling to 4:2:2 will cause an unrecoverable loss of half the color detail.
...
Richard Paul is offline  
Reply HDTV Software Media Discussion

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off