Learn More Contact Now
| | Leave a Comment(1) | TrackBacks (0)
Avid's DNxHD codec and Apple's ProRes are both riffs on the Cineform theme of high-quality digital intermediate codecs for editing.  Both offer more efficient compression than some competitors (especially at larger frame sizes), plus 4:2:2 chroma subsampling (which provides greater color fidelity than DV's 4:1:1 subsampling or MPEG's 4:2:0).  But the advantage of one of the other cool features -- support for 10-bit color depth -- has gone largely misunderstood or overlooked.  [Ed.: Some readers have asked me to emphasize that BOTH DNxHD and ProRes offer 10-bit color depth.  Apologies for any confusion.]

For starters, it's worth pointing out that each of the "extras" that ProRes confers may at first seem irrelevant: after all, broadcast NTSC still (for another month or so) uses YCbCr, and broadcast ATSC and DVB still use a variant of MPEG-2, with all the associated limitations.  Don't be fooled, though. 

Your post process almost inevitably involves changing the source image in some way or another, either through color correction, transitions, or any number of other processes -- and when you have all of the "extra" information in the ProRes picture, you're able to create an edited master that still has more information than you'd need for a "perfect" quality broadcast.  Similarly, you'd never edit in MPEG-2 directly (I hope) -- so using the higher-quality intermediate codec gives your compressor more "wiggle room" as the compressor tries to paint the highest-quality picture for the MPEG-2 transcoding step.

But enough of that ... more on bit depth specifically after the jump.
Background
Remember that your video is made up of a whole lot of points of color -- we call them pixels -- and that, in almost every major video application, we describe those pixels in a way that's technically about the same as "RGB."  That is, each pixel contains a value corresponding to the amount of Red, the amount of Green, and the amount of Blue that mix together to make the color that it's after.

8-bit codecs describe each of these channels (red, green, and blue) using 8 bits of information, or about 255 different possible values.  Incidentally, this is the same as "24-bit" color  ("True" color) when we're talking about computer displays or photography -- and it's a good balance between picture quality and the amount of data needed to describe the video. 

But 10-bit codecs spend an extra 2 bits of data per channel.  If you're working uncompressed, then yes, that does correspond to roughly a 20% increase in filesize.  But in exchange, you get 4 times as many levels of color for each color channel -- and this is a huge benefit as you're editing and mastering your video, even if you'll ultimately be using an 8-bit codec like MPEG-2 for output.

Actually, you may find one of my old articles on audio sampling useful here.  Think of the waves in the diagrams as measuring color rather than audio, and the discussion about bit depth is basically the same math-wise.


Practical
  • Shadows & Highlights -- If you've ever tried to check a range of video for broadcast-safety, you may have some idea about just how narrow that "95%+" range of luminance really is.  Put it this way: the human eye is better at telling the difference between slightly different shades when it's looking near the extremes of a luminance range.  This may or may not have something to do with the fact that areas in shadow or in highlight tend to encompass large areas of fairly similar colors.

    Ironically, this was less of a problem before digital video: in the same way as reds would "bleed" on analog sets, the border between similar color ranges would be a little "fuzzy."  But if today you're having problems with "contouring" in near-white areas of your video, the extra bit depth of ProRes (coming from a 10-bit source) may well be the cure to your woes. 

    Similarly, if you have over- or under-exposed video, those extra bits will give you substantially more freedom to correct blown-out whites or deep shadows.  If you were to try to rebalance video from 8-bit source, you'd exaggerate the problem of the few shades of color available at the extremes.  Say that my video has overexposed whites that I want to fix -- they're not blown out, but I want the colors between "Almost-White" and "Full White" to instead cover the range between "Kind-of-White" and "Full White."

    If I'm starting from an 8-bit source, then I'm trying to stretch a handful of shades of color to cover a larger range of shades.  But if I start from a 10-bit source, then I have plenty of shades of color in that "near-white" range to stretch over the number of 8-bit shades in a broader range of color.  And if I'm being totally confusing ... just roll your eyes at me and take my word for it?

  • High-Saturation Video -- Same principle as shadows and highlights.  Think of it this way: the red channel runs from 0% (black) to 100% (full red).  By the time I've told the video that I want this specific part of my video to be "red" in the first place, I've already had to use a lot of my range in the red channel.  So if I'm trying to distinguish between different shades of pure red, I run into the same issues as if I were trying to distinguish between whites.

    Thus, if you run into a lot of contouring in saturated portions of your video -- or even if you're having trouble trying to manipulate the exposure of those portions -- a 10-bit intermediate codec could well help you out.

  • Any Color Correction -- Mathematically speaking, the vast majority of color correction intrinsically involves some form of "stretching" certain ranges of color to fit different ranges.  If the intermediate video you're working on only has 8 bits of color data per channel, and your output codec uses 8 bits per channel as well, then you're going to have to "round off" some portions of the color range when you "stretch" them through a color correction.  Long story short, you lose color depth and quality.

    Remember, though, that 10-bit codecs have four times as many levels of color information as 8-bit ones.  So if you're using a 10-bit codec, then you could theoretically choose to drop the entire range from 0% luminance to 75% luminance, and the 75%-100% luma range that's then "stretched" to fit the full luma range would still use all of the levels available in your 8-bit output codec: ugly, but no depth loss.



If you have the choice, then codecs with higher bit depths are almost always a great choice for intermediate editing.  As more and more viewers switch to digital TV sets with higher resolutions, the contouring and flat colors of poorly-managed color spaces constantly become more obvious, even to the untrained eye.  And in today's world of cheap storage and fantastic compression ratios, the quality you gain is almost always worth the marginally-higher amount of disk space that higher bit depths require.



Comments?  Questions?  Clarifications?  Corrections?  I welcome your emails

Looking to build a top-notch digital video workflow on a budget?  GeniusDV's consulting services can help you maintain the highest quality every step of the way, from production all the way to DVD distribution.  It's more painless (and cheaper) than it sounds.  One of our Geniuses will personally contact you for a free, no-obligation initial consultation if you just leave your name and number.

comments  

horatio said:

hey, thanks for the article,10bit codecs and 8bit codecs sound like they increase quality but I never heard about them until today when I couldnt play a mkv file due to it being 10bit and had to download an older vlc version or are they not a good thing? Like will a 720p video ie. animation look better if its 720p and 10bit?

track.backs

Listed below are 0 links to blogs that reference this entry: ProRes, and 10-bit vs. 8-bit video -- what's the difference?.

TrackBack URL for this entry: http://www.geniusdv.com/weblog/mt-tb.cgi/1117

Receive FREE Tutorials by email:

  • Enrollment Cost: $95.00
  • 84 Lectures
  • Includes Practice Media
  • Interactive Quizzes
  • Official Certificate of Completion
  • One Time Fee - 30 Day Money Back Guarantee
  • Promo code GDVAVID 10% off!
  • Enrollment Cost: $39.00
  • 60 Lectures
  • Includes Practice Media
  • Interactive Quizzes
  • Official Certificate of Completion
  • One Time Fee - 30 Day Money Back Guarantee
  • Promo code GDVFCP 10% off!

about.this

Using Avid FX with Avid Titles was the previous entry in this blog.

Using the Final Cut Pro Track tool effectively is the next entry in this blog.

Find recent content on the main index or look in the archives to find all content.