Video Compression (SD-DVD Edition!)

All of the Digital Video you watch in your Home Theater is "compressed".  ALL of it!:  Whether it comes off an optical disc, or read from a media file you play, or as a program you stream from an Internet service, or even as a program you watch from a local, on-air, Digital TV station.  Compression reduces both the amount of space necessary to store the video, and -- often more important -- the data rate needed to transmit that video or to read it off of storage.

But too MUCH compression will damage picture quality in ways difficult to ignore!  In this post I'll cover the basics of video compression, along with a case study of video compression as applied to SD-DVD movie discs -- where the sins of poorly implemented compression are rife!  I'll end with some recommendations on what to look for when picking up an SD-DVD title to maximize your odds of getting great picture quality.

Video compression techniques are the subject of deep study in both Information Theory and Digital Signal Processing.  In the few decades since the introduction of Digital Video into consumer electronics, great strides have been made in the amount of compression that can be produced while maintaining the same, or even better, picture quality.  Fortunately, these developments have been matched with the introduction of ever more powerful signal processors, which can do this work fast enough to keep up with movie and TV frame rates -- but at costs still appropriate for Home Theater gear.

Another area of study is in Perceptual Science -- or more bluntly, "How much can we get away with?"  What sorts of damage to picture quality are viewers LESS likely to notice?

This is particularly true for Internet streaming services.  These services operate on the principle that, "The Show Must Go On!"  That is, if there's a (hopefully temporary) slowdown in your Internet service, or if there's congestion on the servers feeding the program to you (and, simultaneously, to LOTS of other people!), the streaming company will attempt to keep the show going by automatically switching to a lower quality version which will still work at a lower data rate.  This can be accomplished both by sending a lower video Resolution, and by cranking up the video compression.

 And extra compression is the thing they try FIRST, because viewers are almost certain to notice a sudden drop in Resolution.  Indeed, these services want to operate at a high level of compression even under ideal transmission conditions, since that gives them more margin of safety before they might have to react if, for example, your Internet service slows down.

But high compression means SOMETHING has to give, and that's were the Perceptual Science kicks in.  For example, it is pretty common for Internet streaming services to severely limit the data rate for dark scenes.  This not only reduces your ability to see near-Black details, but also means dark scenes are more likely to exhibit Compression Artifacts, such as Macroblocking (about which, more below).

However, casual viewers pay less attention to dark scenes, and, perhaps even more important, many viewers don't even have their TVs set up properly to DISPLAY dark scenes all that well.  So any cruft produced by such over compression down there may not even be seeable!  As I said above:  "What can we get away with?"

Playback from optical discs (or media files) is different in that the data transmission rate for the video content is known -- and not subject to external events like congestion from other viewers also trying to access that content.  So the folks authoring such stuff know, in advance, just how much compression they'll need.

However they can STILL run into problems.  For example, the media might be limited in its storage capacity.  If what you've got is too big, you can either trim some of it off, or crank up the compression to MAKE it all fit.

And the authoring tools used to do the compression also require sensible choices to be made.  For example, how much can we limit the data rate for scene backgrounds (which may have little detail or fewer changes from frame to frame)?  And what of action scenes which are so complex and so rapidly changing that they don't compress well?  Is simplifying them by, e.g., blurring the action, better than allowing some Compression Artifacts to creep in?

The huge success of the original SD-DVD movie discs was a testing ground for all such questions, because SD-DVD is nothing if not resource challenged!  It is limited both in the storage capacity on the disc and in the data rate available when reading the disc.

SD-DVD discs come in different configurations.  The playing surface may have 1 or 2 "layers" of content on it.  Having read one layer of content off that playing surface, the player can go back and read that surface AGAIN to get the second layer.

And the discs could also be single-sided or double-sided.  A single-sided disc has playable content on just one side.  The other side usually has a marketing label printed on it.  A double-sided disc has playable content on BOTH sides.  Flip the disc over to play the other side.

The most popular combo was the dual-layer, single-sided disc.  So let's take a look at the numbers for that one.

The capacity of a dual-layer, single-sided SD-DVD disc is 8.5 GB (GigaBytes -- Billions of Bytes of data, where each Byte is made up of 8 Bits).  SD-DVD discs are "variable bit-rate" media, meaning that there's a range of data rates that might be found on the discs.  But the allowed range of data rates has to be nailed down -- built into the standard -- so that the optical disc drives, and their associated data electronics can be designed.

For SD-DVD, the nominal data rate is 4.7Mb/s (MegaBits per Second -- Millions of Bits per Second).  That's only just a little higher than the LOWEST data rate recommended for the standard -- which is 4.0Mb/s.  But it is less than HALF the HIGHEST data rate allowed for the standard -- at 10.08Mb/s!

The nominal capacity of a dual-layer, single-sided SD-DVD disc, then, is right around 4 hours of content.  The math works like this:

4.7Mb/s x 60 Seconds x 240 minutes / 8 Bits per Byte approximately equals 8.5GB

The difference being "overhead data", which is used to manage the format on disc.

But if you want to use minimal compression, and record your movie onto the disc at the HIGHEST allowed data rate, the capacity drops by more than HALF -- to under 2 hours of content!

The math is comparable for the other SD-DVD disc types.  A single-layer disc has nominal capacity for about 2 hours of content, but only 1 hour at the highest data rate.  A double-sided, single-layer disc has the same capacity as a single-sided, dual-layer disc with the added annoyance that you have to flip the disc over to get to the other side.   (Due to various manufacturing limitations, dual-layer, double-sided SD-DVD discs were pretty much a non-starter in the marketplace.)

Now keep in mind that EVERYTHING on the disc consumes some of that capacity.  In particular, if you buy a disc which has two ways of playing the movie -- for example both "Widescreen" and "Full Screen" -- that means the movie is on the disc TWICE.

Every audio track also takes space, including Commentary tracks.  As does every Subtitles stream.

And then there's Extras content -- all that stuff Studios love to put on discs to tempt you into thinking you are getting a more valuable disc.  Interviews, documentaries, behind the scenes footage, and "making of" pieces ALL take up space on the disc!

And it's not just a matter of space on the disc.  It's also a matter of how that data rate gets divided up!  So let's take a look at what you can accomplish with that MAXIMUM data rate of 10.08Mb/s.

First of all it's not all "content".  Some of it is control overhead.  Allowing for that reduces the Maximum, content data rate to 9.8Mb/s.

Then there's audio and subtitles.  Even though you are only listening to one audio track and possibly viewing one subtitles track (out of the available choices), the data for all the choices gets read in parallel with the video.

Assuming 3 Dolby Digital audio streams (alternate languages) at 448Kb/s each, and 4 "sub picture" streams (subtitles) at 10Kb/s each, the MAXIMUM video bit rate is now down to 8.4 Mb/s.

TECHNICAL NOTE: The format supports up to 8 languages (or alternate multi-channel audio encoding formats), subtitles for up to 32 languages, alternate video streams (camera angles) for up to 9 streams, menus, and of course the ubiquitous "Extras" all of which also take up space on disc depending upon how the content producer decides he wants to divy up the available capacity. But the more of that stuff you put in, the less space you have for high quality recording of your primary video track.

So let's assume you want to do the highest quality possible video on a single-sided, dual-layer, SD-DVD, with 3 audio streams and 4 subtitle streams. That means you will want to compress your video only so much as to get it to fit into that max video bit rate of 8.4 Mb/s.  Some of the discussion that follows will make more sense if you've already read my prior post on Digital Video.

Forget about the studio masters of the digitized film for a moment and concentrate on a 480p frame. Twenty-four times a second (i.e., the frame rate for film based content) you will need to record luminance and color data for 720 x 480 pixels. If you took a computer approach to this and did that in RGB at only 8 bits per color, that's 720 x 480 x 3 colors x 8 bits per color x 24 = 199Mb/s.

To get that to fit into your available 8.4Mb/s you will need a net compression factor of 23.7! And that's the MINIMUM compression you'll need. Any additional compression damages the image further.

TECHNICAL NOTE: You may be wondering how the 24 fps film rate becomes the 30 fps video rate for a traditional TV. Well that's done by the player as part of the DVD standard (whether or not the player also includes de-interlacing) in order to keep from having to store even MORE frames/sec on the DVD for its most common use as a storage mechanism for films. This is achieved by duplicating certain, interlaced, half-Frames (called "Fields") in a regular cadence. Doing this is easy. What's hard is doing DE-interlacing properly in the presence of both film and video frame rate stuff coming off the DVD and thus maybe -- or maybe not -- in the presence of these periodically recurring, replicated "fields".  This is made worse since films are sometimes edited with digital editing equipment which ends up spitting out video frame rate results!

That humongous, 23.7 compression factor is achieved BOTH by cutting corners on the amount of data recorded in each and every Frame AND by additional compression which takes advantage of the similarities in successive Frames.

The most important corner which is cut as regards each Frame is that the color data is not really stored at the same pixel resolution as the luminance data, in accordance with choices which were made long ago for a standard, broadcast, color TV signal.  This is the distinction between YCbCr 4:4:4 and YCbCr 4:2:0 I detailed in my prior post on Digital Video.  Think of this as a form of preliminary video compression which takes place on each Frame before the REAL compression algorithm takes over.

That additional compression is possible because real-world video is "coherent".  That is, a given Frame of video looks very much like the Frames on either side of it.  If you could figure out a way to record just the DIFFERENCES between Frames, it would take far less data than fully recording each and every Frame.

And THESE are the compression algorithms I alluded to at the top of this post.  The one used for SD-DVD rejoices in the name "MPEG-2".  It comes from the "Moving Picture Experts Group", established by ISO (the International Organization for Standardization) and IEC (the International Electrotechnical Commission) in 1988.

TECHNICAL NOTE:  If you spend any time trying to follow these standards bodies, you will likely find yourself frequently asking, "Can I Buy a Vowel?"  These days the MPEG-2 standard is also known as H.222 / H.262, and is controlled by ITU (the International Telecommunications Union of the United Nations).  The MPEG-2 standard itself was published in 1995.  Newer, more sophisticated, compression standards are used for Blu-ray discs, and even newer for UHD (4K) Blu-ray discs.

So when you allow for the chintzy color signal, the additional compression you need to glom out of your MPEG2 encoder is a more modest level, say about 15x.

Now compression doesn't come for free. It is "lossy". The more compression you have to use the worse off the resulting image will be. And indeed at any given LEVEL of compression there are CHOICES the authoring tools get to make according to where in the image to focus the efforts of compression. Badly made choices result in visible artifacts. (Over time the Studios got better and better at this.)

MPEG-2 works by dividing each Frame into rectangular chunks called Macroblocks, and then tracking how each Macroblock moves from that Frame to the next Frame.  The more Macroblocks you use, the more of this tracking you have to record, so the level of compression is smaller.  But if you use too FEW Macroblocks (to get a higher level of compression) they become too large, move too much from Frame to Frame, and/or contain too much internal change INSIDE the block to do a good job of representing the next, successive frame.

This results in artifacts -- visible defects in the picture quality.  This particular type of artifact is called, reasonably enough, "Macroblocking".  It looks like the image Frames -- or sizable portions of the image -- have broken up into moving rectangles which appear and disappear, apparently at random.  It is pretty much impossible to ignore.  If you see Macroblocking then what you are seeing is excess video compression.

There are other types of artifacts that result if the MPEG-2 compression algorithm gets cranked up too high.  Another common one, called "Mosquito Noise" looks like little dots (the mosquitoes) hovering around sharp edges -- such as when text appears in the image.

But this is doable!  That is, if the film is not too tough to compress, MPEG-2 can handle that remaining compression.

However, that means you are going to have only 2 hour capacity on the disc! Well by the time you add in MORE than 3 audio tracks, fancy animated menus, and all the desired "extras", the net result is that you have to crank up the MPEG2 compression HIGHER to make it all fit -- which damages the video. And that's where the studios have a choice. Will people buy discs based more on a high quality transfer, or because there are DVD-ROM games added onto the disc?

And this is where YOU come in, as a now-savvy buyer!

First of all, try to find out what media type was used for a given SD-DVD.  Sometimes you can find this in fine print on the packaging.  In other cases you may need to do some sleuthing on-line at web sites which catalog this info.  In particular, is the disc dual-layer or single-layer?

Next check what's been crammed onto that disc.  Does it include more than one version of the movie?  What about Extras content?  Has that been placed on a separate disc?

Ideally what you want to find is that a dual-layer disc has been used to store only one copy of the movie, and that ALL the Extras content (other than, perhaps, a Commentary audio track), has been placed on a separate disc.

Also consider the LENGTH of the movie, given those capacity limits I mentioned above.  If the movie is over 2 hours long -- as, for example, with some Extended Edition releases -- it really should be divided across more than one disc.

For a while, the Columbia TriStar division of Sony was marketing what they called "Superbit" SD-DVDs as if they were something totally new.  They weren't.  They were still just SD-DVDs.  But they were SD-DVDs authored with extreme care for maximizing the video bit rate.

So for example, any movie Extras were always on a separate disc.  And if the movie was long enough it was divided across two discs.  In addition, the Superbit discs used very simple menus and strictly limited the number of audio tracks and subtitle streams.

TECHNICAL NOTE:  There are OTHER ways to screw up the transfer of a movie than just over-compressing it.  For example you could get sloppy and end up crushing near-Black details, or getting the color balances wrong.  Extra care was taken in these other aspects as well to maximize the picture quality of these Superbit SD-DVDs.  In the end, it is the combination of these factors which made those discs look so good.

Also take particular care when looking at discs that include both the Widescreen and Full Screen versions of the movie on one disc.  There are two common ways to do this.  First you could use a double-sided disc and put each copy in the single layer found on each side.  Flip the disc to choose which you play.  Or you could put the two movies on a single-sided dual layer disc by putting one version in each layer.  Done that way, a menu selection on the disc can be used to decide which one will play.  But in EITHER case, you've now glommed each version of the movie into a SINGLE layer -- half the capacity of a dual-layer disc.  And unless this happens to be an unusually short movie, this means both versions will likely be over-compressed.

As I said in my earlier post on SD Video Aspect Ratio, the amazing thing about Standard Definition Video is just how impressive it can look when it is done right!  BUT, there is no margin for error.  Cut corners at any point in the video processing chain from camera to screen and you WILL get picture defects that WILL be seen by viewers.  Over-compression is just one such mistake.  However, by following the recommendations just given, you can maximize your odds of picture bliss!  Keep in mind that Studios which have taken particular care with compression, have quite likely ALSO taken care with other aspects of authoring the transfer to disc.

But Studios want to make money, and they know people get attracted to discs that have LOTS and LOTS of stuff crammed onto them.  Indeed, they hope to get your money MULTIPLE times for the same movie title!  So the first release of a movie might be a quick and dirty (literally) transfer, which is then followed up by a higher quality release for those who know the difference, which is THEN followed up by a  "collector's edition" which includes that high quality transfer PLUS Extras content on a separate disc.

Since it remains the case that many great movies have yet to be re-released for Blu-ray or UHD -- and may NEVER be! -- it pays to be a savvy buyer.  Know what version of SD-DVDs are available for a given movie, and try to limit yourself to the ones that are authored well.