WWDC2004 Session 730

Transcript

Kind: captions
Language: en
I hope you've been having a good day
thanks for coming to our session and
it's the evolution of standards I'm very
pleased to welcome you to our panel
discussion today and that panel
discussion is going to be led by Dave
singer who is one of our standards
experts at Apple so Dave thank you very
much so yes welcome to the evolution of
standards this is a moderated panel
session we're going to go through a few
slides at the beginning and give you
time for questions at the end we're
going to give you a rough idea of what's
going on what's happened recently what
may be coming up out of the standards
organization and I'd like to introduce
with you the distinguished panelists I
have here so on the left here a leading
luminary in the video coding business a
published author and a respected figure
worldwide in video coding Barry Haskel
of Apple Computer gives me great
pleasure and a small squeeze sort of
excitement to introduce the chair of the
International MPEG audio subgroup a sage
in the audio coding business again a
leading figure in the international
audio business skyla Quackenbush I'm
bringing out the back I'll talk a little
bit about about system stuff we're going
to introduce the session you asked
questions we have a lot of fun so I'm
going to go through the standards
organizations at least some of them that
you may have heard of and who they are
roughly what they do give you a couple
of things about some things that I think
are exciting about what's happened in
systems related standards recently a
couple of slides on licensing but I'm
going to turn it over to Skyler who's
going to talk to you about audio and
then we'll do a video update at the end
we will be taking lots of questions at
the end and we'll take straight for
clarification questions during sessions
so if you have a clarification question
don't hesitate to interrupt but if it's
a general question hold at the end we're
going to leave plenty of time for it
unlike what you might expect we're going
to be frank honest even forthright about
questions like when will Apple introduce
a particular technology or what legal
advice do you have for this and so on
and in order to save time however I'm
going to give you that question now the
answer is we don't know you might want
to save those questions for some other
time so the standards organizations this
is the most boring part
you can sleep for a minute okay ISO an
ITU two very big organizations ISO has
two important working groups for the
quick time related people working group
one under SC 29 the joint photographic
experts group the people who bought your
classic JPEG JPEG 2000 and so on
they're doing lots of exciting work I'm
not going to talk about it today working
group 11 the motion pictures expert
group where we all go to they're the
people who brought you mpeg-1 mpeg-2
mpeg-4 and then pick with lots of other
numbers behind it too and then the ITU
has the video coding expert group which
you've heard about which is where the
h.264 work started for example on lots
of other h-series codecs certainly
listen F&E I know that's been talked
about a lot at this session there's two
really big organizations there they're
both partnership projects they're
partnership projects between standards
organizations believe it or not mostly
national standards organizations 3gpp
sets the standards for roughly the
people who are in the GSM family of
cellular standards the ones that started
in Europe but are now basically
worldwide 3gpp - follows the other track
the CDMA cdma2000 related set of
families you'll find them both have open
web sites in which you can find a lot of
information - more interesting ones
Society of motion picture and television
engineers they do a lot of the standards
work for production and broadcasts and
so on that's simply and you'll find lots
of information on their website and
finally those of you are interested in
streaming the standards for
internet-based streaming or the protocol
dispersed specifications and so on they
come from the internet Engineering Task
Force IETF now there are also out in
that field industry consortium there
aren't really standards organized as
such these people do collaboration
agreements and so on implementation
agreements where they all agree to
implement the same thing they do a lot
of the testing sometimes some of the
more business related discussions
happening consortium like this so
internet streaming media Association is
ma they do they do interoperability
agreements for the internet and for
ip-based networks Open Mobile Alliance
they do that for cellular and for
telephony related high-level functions
TVB they do it for digital video
broadcasting if their title implies
mostly based in Europe DVD forum there
people who talk about shiny discs and we
had a lot about that yesterday in the
h.264 session
finally the MPEG industry forum put this
one laughs but it's actually a good
place to go to this is one where they
promote and explain MPEG technologies
they have a lot of information on their
website they tend to keep a lot of news
there may be a good place for you to
start and you'll find lots of links to
other useful places from the MPEG if' so
with that I'm going to just give you a
little quick update of two things that I
think are really exciting in systems
unfortunately I know systems is not that
exciting so I'm going to talk about file
format I'm going to talk about file
format simply because it's my baby
ok so you're just going to have to humor
me for one slide what's been going on
with the ISO file format the ISO file
format is we told you last year and the
year before was adopted into MPEG as the
basis for the mp4 file format then it
became the motion JPEG file format then
the 3gpp file format under 3 GPP - file
format and the SD video card
associations file format so you can see
this powerful much going going places
and really going that with the AVC the
h.264 up the committee realized that
because they commit the the codec itself
offered had such a variety of features
and functions we could do with better
file format support so there's lots of
fun stuff in there you can group samples
together so you can do layering and
subsequent selection of an AVC session
in an ISO based file you can get much
more dependency information because the
dependency structure of AVC is so much
more complicated we give you clues in
the file format or at least we enable
you to put clues in the file format so
you can disentangle the dependency
information there's an amendment in
under way which includes those
extensions and actually a lot of other
stuff too a lot of you have been saying
guys the ISO file format approach to
metadata is very weak and we knew that
so we put in a really good foundation
for metadata support and explicit
support for MPEG 7 but it's also open to
storing other kinds of metadata so if
you're a metadata oriented person this
amendment may be for you and then
there's support there for content
protection so you can do protected
rights protected DRM content in the file
format in a standard way so we can
finally we got a standard container
format for exchanging rights protected
content which is a I think a
breakthrough in the industry an
interesting development going on in 3gpp
release six is almost done
a terrific amount of stuff in there for
those of you who follow the cellular
world this is probably the reason I
think we'll will become the one that
becomes prevalent in the in the 3G world
very rapidly not only does it include
the h.264 codec which will obviously the
video feature but in the system's area
they finally put content protection in
there so we can talk about rights
protected content in the cellular well
there's buffer management and adaptive
streaming across the network so we can
expect better quality streaming better
adaptation between the handset and the
server back at the of the operators
place and indeed there's RTP
retransmission support so if a packet
does get lost despite all the lower
layers attempts to get it through to you
there's retransmission support at that
level there's also quality of experience
monitoring which is going to be
interesting it's going to enable the
operator to work out hey that sucks for
you and not charge you so much before
you actually had to call them and
complain that it's such of course
they're hoping it won't suck at all and
then there's multicast and broadcast
support so there's lots of opportunities
in business now to send out real-time
live content in a multicast broadcast
fashion which puts less load on the
network and obviously should cost less
to the end-user so with that I'm just
going to give you a really quick update
on licensing so we normally do licensing
licensing is perfectly ordinary there
are companies whose job it is to do
basic research and development and then
license it and we all appreciate that
because we couldn't get that number of
experts together in the same place and
get the same result right so licensing
is a normal part of a high-tech business
it became difficult over the last few
years we know but we think it's getting
easier again almost any successful
market needs an interoperable standard
you've got to have multiple players in
the market who can interoperate
providing the different products that
fill out all the niches both in terms of
function and in terms of placement that
are needed to make a market mature for
multimedia standards licensing models
have included encoder/decoder
manufacturing and bundling fees that's
perfectly normal you haven't seen them
in the past because they typically been
embedded in the chips that were put into
the products so that chip cost a little
bit more because it paid a license
they've also been content replication
fees that's perfectly normal again so
when you buy a DVD there's a there's
amount of money going back to the rights
owner here who who developed the
standards for how you make that DVD
that's how they make their money
we typically cover a good chunk of what
we can for you so you can use the
technology without having to worry too
much so the question is of course okay
if Apple's covering a good chunk of it
do I need a license
well you might one of my strongest
suggestions to you and I mean this
seriously is read the License Agreement
we give you we actually spend a lot of
time writing that License Agreement and
it really breaks our hearts to think
people are just going to go agree
without having read it yeah and
particularly it breaks our hearts when
they go agree without having read it and
then they have to call us to say what do
they say so if you're in doubt about
your use read the License Agreement
it might explain it and if you are still
in doubt you can do two things you can
work out well if I'm doing a product of
my own an apples license doesn't apply
maybe I need to go to the licensing
authorities or look at the standards
itself most standards when they
published have an annex of people who
claim IPR and that standard you could go
and read that and contact them there are
two big organizations in the business
who cover a lot of the multi media
licenses you might need MPEG LA they did
then take two famously and take four
video I'm systems license and they're
handling some of the ABC licensing and
Paykel a comm for your licensing they do
a lot of audio licensing and they also
have some of the ABC licensing and via
licensing comm they have FAQs on their
websites they have press releases they
have synopses of the license agreements
and then if you're feeling really really
keen you can actually read the legalese
of the license agreements themselves so
with that I'd like to hand it over to
Skyler who's going to give you an
exciting audio update thanks Dave okay
so I want to begin by reviewing the MPEG
audio standards talk about where MPEG
has been and where MPEG is going I'd
like to begin the timeline with the
compact disc 1982 which I think kicked
off the whole digital audio revolution
more than ten years after that MPEG
standardised MPEG 1 layer 3 or mp3 so
that is looking already like a pretty
old guy but still holding an important
position in the marketplace
several years after that and take two
issued advanced audio coding and this
also exists in the mpeg-4 standard this
is arguably the state of the art in
perceptual audio coding several years
after that 2003 last year MPEG issued
the high efficiency AAC codec this
builds on AAC we'll talk about how that
works but in fact it uses a different
dimension exploits a different dimension
to get more compression in 2005 next
year I anticipate MPEG will issue a
standard on lossless audio coding so
what we see is I'm giving you a timeline
view here but what we see is that we've
got compression and then we kind of take
a step backwards and say oh let's do
something else
let's do compression but I want all the
bits back and then finally we're about
to begin a new work item in spatial
audio coding we're back in the
compression business but again we're not
just doing a perceptual audio codec
we're exploiting a different aspect and
that actually soundstage compression to
give you total bitrate compression so
here's the timeline of what MPEG has
been doing I'm going to stand over here
and so let me give you another view of
this and look at performance so the top
line is the compact disk and if you look
at the right hand side I want to look at
its per channel per second which kind of
normalizes my whole rate because we'll
see that we're going to go from two
channel to multi channel in the course
of looking at the various specifications
and I've reversed the timeline and we're
looking at on the right hand column bits
per channel per second so the ALS the
audio lossless coding it's at 325
kilobits per second that's a lot but
that's about two to one compression and
you get all the
Spang so it's bid exact I think that
number is it's 48 kilohertz sampling
rate it's not a perceptual coder
it's just compress it mathematically and
give me the bits back which is a
departure for MPEG now the next line is
mp3 it's the oldest of the MPEG coders
it's a perceptual audio coder and it
gives you good compression not the best
and typically people I mean I think it
should be run at like 80 kilobits per
second per channel so 160 kilobits for
stereo a lot of people run at 128 a lot
of people run at 64 but at 80 it really
sounds good now I said the next after
that was AAC and in effect really sounds
good at 64 kilobits per second per
channel and then affect MPEG benchmarked
it as an EB you broadcast quality codec
at that rate so as I mentioned before
mp3 and AAC are perceptual audio coders
you're getting compression and it
primarily exploits aspects of human
hearing to shape the coding noise so
that you don't hear it now the next
standard is high efficiency AAC and
again another slide will show you how it
works
but here we're not just doing perceptual
coding in fact it builds on top of AAC
and it uses AAC but then as a feature of
coding the bandwidth so called bandwidth
extension it gives you compression by
coding the bandwidth and a clever way so
there's shape noise then with coding and
in the bottom line s C stands for
spatial audio coding a whole other
dimension in compression by exploiting
soundstage compression in a perceptual
way so it is a perceptual coder but it's
not shaped no ways it's how we perceive
a soundstage and can we do clever things
to code that in a very compact way and
the answer is yes and so we see that to
recap empty
380 kilobits AAC 64 kilobits HDAC AAC a
huge leap forward 24 kilobits per second
and finally spatial audio coding 21 so
we are really marching forward but we've
had to use different tricks to get there
so let's talk about what are those
tricks how does it work
okay now first I get ahead of myself
so we those are the MPEG standards where
are they in the marketplace are they
being used so let's talk about that
first
mp3 it's everywhere I mean it's old it
has taken ten plus years to be really
successful in the marketplace so
successful it's never going to go away
so MPEG is usually proud of that piece
of technology AAC which I say is
definitely one better the state of the
art it's being used by Apple iTunes and
the iPod cool great realnetworks wraps
Rhapsody uses AAC at a very high bitrate
they want to really bring you great
digital audio so they run at 192
kilobits per second for stereo and in
fact you will see AAC and a number of
tree G cell phones HD AAC
it's the stuff that makes XM satellite
radio go satellite radio is what I call
expensive bits and you need compression
when bits are expensive and so they can
only dedicate like 48 kilobits per
second to a broadcast and you need some
heavy compression to deliver
entertainment quality audio now AAC I
told you real at 192 kilobits per second
this is excellent quality it sustains
the your inspection and preview in the
best listening room hea AC we're talking
huge compression to get over that
channel listen to it in your car
enjoy it but it's at what I call
entertainment quality so we talked about
compression but it's in the think of it
in the context of the application so
it's
all reference listening room kinds of
qualities ALS again this is in
work-in-progress and MPEG I think it's
going to have a lot of application in
professional music archiving when
professional goes v channel which is
happening now that's a lot of bits and
even if you can give me a factor of two
that disc farm got a lot smaller so
they're going to go with that and we
will see that one of the advantages of
standards is longevity you know I got
the data base it cost me a fortune
can I please read it 10 20 30 years from
now so standards are good in that
capacity
finally spatial audio coding it's a new
work item it's going to begin at the
next MPEG meeting in July I'm very
excited about it and what I think is so
great is that I told you about HEA AC 48
kilobits per second to my car
what if you want to do that and say now
it's 5.1 channel well I think the
spatial audio coding could make that
happen so I think it's a very powerful
technology than a local and open a lot
of markets but we're just beginning
right now I'm going to talk about these
technologies and very quickly give you
the 50,000 foot view of what makes them
work this is mp3 a lot of boxes just
look at the blue boxes what you see is
the 32 band filter Bank that's common
four layers one two and three layer
three follows it by at 18 point mdc-t
the result is high-resolution filter
Bank and this is the first component to
get compression it's what I call
statistical compression you need to
Decorah late the signal and just in a
mathematical sense you get compression
the second component is the perceptual
model which is up on the left this goes
one better and says okay I see what the
mathematical model says and
statistically speaking I can
rid of some debts but now the perceptual
model says well what can you really hear
and so based on that model it says okay
it's important to code this you don't
have to code that at all and so it is
really driving the allocation of bits in
the whole coder and then finally the
blue box noiseless coding is essentially
an entropy matching block and it's
needed to deal with some mathematical
aspects of compression and so and it's
in a very first order approximation the
filter Bank is like a factor of two
compression the perceptual model about
2.5 and the lossless coding another
factor to you get to about a factor of
ten in this classic perceptual coder
framework now AAC it's better what
really happened a lot more boxes I'm not
going to go into that but now you got a
1024 point filter Bank even higher
resolution it made a difference but you
get the same perceptual model the same
noiseless coding so in the first
approximation it's just a cleaner
looking layer 3 in which you don't have
the two filter banks but just one same
principles apply the classic perceptual
audio coder now what is this other stuff
I was talking about high efficiency AAC
well really it's AAC kohler with
something else and what you begin is
with an encoder that down samples of
signal and just codes the first half of
the bandwidth with AAC but it has some
side information the encoder looks of
the whole thing and it says hey can i
approximate that upper band by what I
know about the lower band and I send
some hints and that's the side
information the answer is yes and it
works incredibly well so in the decoder
you've got a classic AAC decoder you get
back the hatch mark here and then you
essentially map it up to the high band
fix it up equalized with the side
information and you get the wideband so
the point is you got as a foundation the
classic perceptual coder but the new
dimension is this bandwidth coding we
have to use a new idea to get more
compression this is one of the new ideas
now spatial audio coding is another case
of a new idea you have a five channel
signal you present it to the encoder and
the first thing it does is it mixes it
down to let's say two channels and
encodes it with a classic AAC coder
sounds like HEA AC so far but now
instead of talking about bandwidth we're
talking about soundstage what can I tell
you if I have two channels about how to
unwrap it
two font the drums are in the right
channel note they're really in the right
surround so I have some signed
information that says how can I give the
decoder hints to expand that so it's
soundstage coding based on a solid to
channel Foundation and in fact it works
incredibly well we have some early demos
I expect to begin this work in July and
the first demo indicate that I can do
5.1 I'd give a hundred and one hundred
and ten bits to AAC and maybe twenty to
thirty two kilobits to the side
information and you know in the first
audition you can't tell the difference
between the coded and the discrete five
channel based band so I think it's
really exciting technology and I think
it's going to have a huge impact in the
marketplace but that's in the future
finally lossless coding we do a few
things in MPEG I want to just bring you
through one this is a classic lossless
compression architecture there's nothing
new here it's got a time domain
predictor and you follow up by an
entropy coder why what's new well the
answer is nothing but it's MPEG so if
you look at the end big standards
there's compression but nothing lossless
and so as I said before you know I'm the
I'm the DVC I save all my programs I got
dis farms that stuff's worth some money
I want to make sure I can read that 20
years from now and standards provide a
kind of a comfort level where they can
say well the standard won't change
I've got multi-vendor support I'm going
with the standard as opposed to a
proprietary solution where that company
maybe they're going to go out of
business
so it doesn't have to be new and better
it can just be standardized and that's
the role that we want to fulfill with
this and I think it's a legitimate one
okay I'm going to give you one demo
it's a AC versus high-efficiency a AC
and in fact I'm going to do it at 32
kilobits per second I just want to show
you that that this is a new thing and I
think you'll see that relative to a AC
there's a remarkable difference
delivered by HEA a sync so we can go to
the demo machine okay we're not going to
go its back okay so 32 kilobits I'm
going to play AAC
[Music]
it's band-limited I can't do better than
that with this classic architecture so
let me just bounce between that and HEA
a stage
[Music]
okay with expansion and they could be
different
muffles low uninteresting you get more
staying bitrate
so I wish I could demo the spatial
coding but I don't have live channel
here but I just wanted to show you that
MPEG has got some new stuff going again
this is at a bid rate which is 32
kilobits per second for stereo that's I
would have said that was ridiculous
several years ago but it is a place in
the marketplace it sounds awfully good
and as I said XM radio runs out at 48
but in this room 32 says everybody says
I hear the difference so you kind of
kind of push the compression so you
realize what's going on but can we have
the slides back again but I think it the
answer is cool cool technology okay so I
have one more slide find out more two
URLs the one is the MPEG homepage the
second one is the audio homepage so I
certainly invite you to go to the audio
homepage and then there's some books on
MPEG these Jen tend to be all of MPEG
but you'll find a chapter on audio so
there's a couple for your reference and
that's all I have thanks caller and now
do something completely different a
video we're going to try and cover these
three questions what are they what can
they do for us and what are they going
to do for us in the future well
everybody loves pictures and the test of
the video progression is whether we can
see the moon in the upper right hand
corner and we can that's great but they
take an awful lot of bits if you don't
have any compression Home Video
telephone for example 6 megabits just
for home video telephone high-definition
TV gets you up to one and a half
gigabits an incredible number of bits if
you can't do compression for example
now here's what what compression can do
for us for the home video telephone we
can get compression ratios of 50 with
very very high quality or push it up to
as high as 300 if we're comfortable with
the usable
I won't go through all of these but down
at Entertainment Television a
high-definition television we can get
compression ratios in the range of 40 or
50 with very high quality or even with a
little more for quality that's usable
something like you're used to on your
cable television the current video
coding standards look like this and here
are the dates in which they were
introduced h.261 i guess was one of the
first ones to catch on in a big way
that's used even today in video
conferencing internet video conferencing
still uses this a little bit mpeg-1 came
on in 1992 that was aimed mostly at
compact discs kuroky machines use MPEG 1
almost exclusively MPEG 2 was jointly
done with the itu itu gave it this
number h2 6 to MPEG gave it this well
it's got a bigger number actually but
everybody calls it MPEG 2 and that's
what's used in today's digital
television HDTV DVD and personal video
recorders a whole bunch of things use
mpeg-2 this is probably the most
successful video standard ever h.263 is
an itu standard follows on h.261 it's a
better video conferencing standard
mpeg-4 part 2 as the video came on a few
years ago aimed at PC and Internet
applications it's more than the
compression standard it just has a
plethora of things that it's trying to
do with mpeg-4 and last but not least
just coming on a few months ago is h26
this again was jointly developed by ITU
and MPEG ITU gave it this number h.264
mpeg-4 AVC and that's a bit of a
confusion you'll find people bouncing
back and forth I tend to call it h.264
David tends to call it ABC and different
people just bounce back and forth so if
you can just plant in your brain AVC in
and h.264 are the same or the same
and it's evolving a little technology I
don't know how much technologies
introduced here actually how many were
at the h.264 session yesterday same time
same place a lot of people well and some
weren't
okay well I'll mention a little
technology of course we all like color
pixel a color pictures and we know we
send color pic they come out of the
camera as three components red green and
blue and those can be digitized of
course but if you just do that you end
up with a lot of bits we don't send the
red-green-blue we convert that with a
linear transform into luminance and to
chrominance signals it's a reversible
transform that but with this transform
that enables us to subsample the two
chrominance signals the chrominance is
usually less samples per picture than
the luminance a little more technology
jargon iframes P frames and B frames
what do these mean well what it means is
typically we don't send the pixels
themselves we divide the picture up into
small blocks I believe this is 8x8
blocks it's 8 pixels by 8 pixels
sometimes they're 4 by 4 and instead of
sending the pixels we transform those
into transform coefficients it's so
discrete cosine transform is what's used
by everybody today historically there
were other transforms but this is what
we use so we don't send the pixels we
send transform coefficients and iframes
means well we just send the transform
coefficients in one frame very similar
to the way JPEG does the coding does not
depend on any other frame now we also
use motion compensation if we want to
send this block in the current frame we
try to predict it from the previous
frame and since video often has motion
the previous frame the block is in a
different position so we use that
shifted block as a prediction of the
block we're trying to send in the
current frame and we send the difference
between
in the current frame and the previous
frame if this is an 8 by 8 block we get
64 differences we transform the
difference and send that and that's what
we call a P frame the motion compensated
predict and frame it depends on one
other frame we also use by directional
motion compensated prediction in this
case if we wanted to send this block we
calculate the same motion compensated
prediction in the previous frame but we
also sent calculate a motion compensated
prediction in the future frame we then
calculate the two differences from the
current frame to the previous frame
current frame with the future frame add
up the differences divided by two and
send the average and that's what's
referred to as a B frame by directional
motion compensated predicted frame this
coding depends on two other frames we
put it all together let's you get a
block diagram that looks like this I
won't go through all the components
here's a predictor predictor goes into a
subtractor you calculate the difference
transform the difference quantized code
it send it off to the transmitter
decoder does virtually the same thing it
calculates a prediction but the received
signal is a different signal so that has
to be added to the prediction to get the
decoded image so that's how all these
coders work now an important issue on
these standards is that they specify the
bitstream they tell you what the bits
are and they tell you how to decode
mostly they tell you how to decode those
bits they do not specify the end coder
they do not tell you how to produce the
bits that's a difference between the
audio in the video standards the
encoders in fact continue to improve for
years and years and years after the
standard is finalized
mpeg-2 encoders are still improving
every single year mpeg-4 video part 2 is
video there are all kinds of other parts
has a wide range of bit rates and a wide
range of options now here's the picture
we use to get to describe mpeg-4 mpeg-4
was aimed at coding
separate 3d objects separate audiologic
separate video objects and to allow
those objects to be coded and
transmitted separately and composed at
the receiver at the decoder to produce
the final picture the final picture
might be choosable at the transmitter
might be choosable at the at the
receiver I think it's fair to say this
vision is not implemented very much
anywhere the academics are having with
having a great time with it hasn't
really hit the marketplace there are
other features like text-to-speech
patient body animation wireframe models
sprite coatings of backgrounds just a
wide variety of features the newest
video coding standard when you heard
about yesterday one we're introducing at
Apple this week actually is the
h.264/avc standard much better
compression than mpeg-2 approximately
twice as much as mpeg-2 on the
compression side so for example for PC
or cd-rom applications which typically
mpeg-1 handle compression ratio of 55 50
well these compression ratios as you
know depend on the quality of the coding
so this is a range 55 you can double at
over a hundred for a normal TV
compression ratios of 60 figure double
lot you can get over a hundred double
the number of channels that you can send
on the digital transmission and for HD
again doubling that would would mean you
could get double the number of HD
pictures so that you can on today's show
cable now what's in this h.264
technology you heard much of this
yesterday and I won't go through these
in great detail there's a 4x4 transform
instead of an 8x8 transform the motion
compensation blocks instead of all being
the same size can have different sizes
if you're on the edge of an object you
want to have small motion compensated
blocks there's ad blocking filter
many of these compression algorithms if
you really squeeze them down you begin
to see the 8x8 blocks so there's a D
blocking filter that des compensates for
that what else there's an intra
prediction instead of coding the little
blocks separately they can use
predictive coding to encode the blocks
there's some error resilience techniques
if you have enough bits you can actually
send the pictures several times so if
one of them doesn't make it during the
transmission you can use one of the
others you don't have to send the whole
picture you can send pieces of the
picture some pieces are more important
than others multiple reference frames
this coder allows you to use in your
motion compensation a whole bunch of
previous frames not just one so you can
choose which of the reference frames you
want to use what oh what's coming up
well an MPEG in h.264/avc an amendment
just being worked on now is called
fidelity range extensions at the mouth
full we typically call it fixed
that's one syllable acronym perfect it
allows up to 12 bits per pixel prior to
this most of these standards use 8 this
12 may be extended actually to 16 it
allows higher color sub Sam a higher
color resolution typically the color
resolution has been subsampled by 2 in
both horizontal vert and vertical
directions the sub sampling can be what
the sub sampling can be completely
eliminated for professional applications
it has an optional 8 by 8 DCT on some
videos the 8 by DCT works a little
better than the 4 by 4 DCT so they threw
in an optional one and an alpha channel
is being considered currently another
one that's just getting underway
this is MPEG 21 now so you may ask how
to send big come up with these numbers 1
mpeg-2 mpeg-4 and pic 7 and pig 21
answer said we don't know but we have no
idea it's a marketing thing I think
scalable video coding scalable video
coding has a checkered history in this
business MPEG 2 had a scalable video
coding profile it was never really
implemented as far as we know it's a
great match for networking that has
multicast you know the multicast IETF
mechanisms for the Internet
unfortunately multicast is really isn't
everywhere that's the problem so if you
want to do scalable multicast broadcast
you may not be able to get everywhere so
it's not used very much but MPEG 21 is
going to give it another try and we'll
see what happens
h.265 that's more or less a placeholder
actually it'll be the follow-on to h.264
at this stage it's probably a research
effort and that's how most of these
video standards come about actually a
research effort starts and then
technology is either adopted or tossed
and then after quite a number of years
and in the case of MP of h.264 the
development of the standard took place
with with contributions from a lots and
lots of people the MPEG video standards
typically have more than a couple
hundred people contributing to these
standards so h.265 we have high hopes
for I'll give it six or seven years
maybe another issue on these standards
of course is the processing power that
we have to deal with a lot of the ideas
are not particularly new but with the
incredible processing power that our
chip friends have provided us we can now
implement these ideas so given a few
more years hopefully the curve will
continue to be will give us more
processing power in the future and we
can implement a lot of these ideas okay
that in a nutshell is
is so what video standards are all about