DV Format &
Compression
The following material is provided
courtesy of R/com MediaSchool and author David
Barrett.
DV25
The most common form of digital video is referred
to as DV25. If you have a Sony, JVC, Panasonic,
Canon or other camera that has the DV or DVcam
logo on it, you’re working with DV25 (also
including MiniDV, DVCAM, and DVCPro).
There are two methods for recording a DV25 signal
– onto tape or onto a disc – either
a hard disc or in some new cameras, directly to
DVD. It’s a new and exciting prospect to
know that a video file can be created and saved
directly to a hard drive.
DV25 refers to a compression algorithm
(often called a CODEC). Various manufacturers
have made slight adjustments to the method of
recording and storing the DV25 data (hence the
variations in names as noted above), but overall,
the format is the same. By the way, if anyone
ever says to you, “I’m using uncompressed
DV,” you have permission to laugh. All DV
files are compressed. We’ll cover more on
compression in a bit. In fairness, if someone
does refer to “uncompressed DV,” they’re
probably trying to say that no additional compression
was added during the recording process.
DV25
Specifications
Compression Ratio: DV is compressed at a ratio
of 5:1 Unlike certain other formats, DV compression
is fixed (you can’t scale the rate of compression).
Data Rate: 25 Mbps Now, you know why we call it
DV25. The great thing about using DV25 on a computer
is that you can predict how much hard drive storage
you’ll need. Note the following (approximate)
examples:
1 Second = 3.5 MB
1 Minute = 215 MB
4 Minutes, 40 Seconds = 1 GB
1 Hour = 13.2 GB
DCT Compression:
DV25 uses a form of compression called Intraframe
Discrete Cosine Transform (DCT). It is fairly
similar to MPEG compression but is made up entirely
of I-frames. In addition, each frame stands along
– there is no reliance on other frames for
color or other data. As a result, DCT compression
is ideal for nonlinear editing.
Color
Sampling: Color Sampling could take up
an entire chapter in a book on video. We’ll
spare you the technical blah-blah and just let
you know that DV uses a color sample rate of 4:1:1
This number is a ratio of luminance sampling vs.
color sampling. The initial number – 4 –
refers to the baseline for sampling. In this case,
all digital formats sample luminance at 13.5Mhz.
If you see a format with color sampling rates
of 4:4:4, you’ll know that there is nothing
missing from the color compared to the luminance.
4:2:2 sampling is used by many “professional”
formats and in the analog world that would include
Betacam SP. 4:2:2 color sampling in the digital
video world typically is used for DV50 formats
(hence the 50 in the name). We’ll cover
more on this in a moment.
DV25
vs. DV50 and other video formats
I realize some of you may be starting to nod off,
but the functions and relationships of video formats
is critical information for a motion graphics
designer. DV25 will be thrown at you in an ongoing
manner, but if you don’t know its strengths
and weaknesses, especially when compared to other
formats, you could be complicating your project
results – so a little background can go
a long way.
Standard definition digital video
signals are typically 4:2:2 YUV component resolution.
This includes DigiBeta, D9, and DV50 formats.
The chrominance (color) components and Chroma
resolution are one-half of the luminance resolution.
This reduction in information allows the data
to be smaller –the average human eye is
far more aware of luminance than chrominance.
DV25 further reduces the chrominance components
and Chroma resolution to one-quarter the luminance
resolution. In most applications, the human eye
does not notice this reduction – and in
fact, most images contain relatively low saturation
colors, so DV25 is capable of producing remarkable
images – in some cases, better than professional
analog formats of old.
DV25 was more difficult to work
with when creating composited layers – as
you do in Photoshop and After Effects, due to
the lower spatial color resolution. The result
can be aliased images – they look like stair-steps
on diagonal lines. After Final Cut Pro, Adobe
Premiere, and After Effects and other applications
now have filters and other tools to help overcome
this problem – but you need to know the
structure of the data, or you won’t know
how to solve the problem. In this case, you can
achieve a quality matte edge by slightly blurring
the RGB channel that is related to the key channel
(such as green or blue). It’s also very
important that you apply the proper amount of
light to your subject – and if you’re
using a green or blue screen, make sure you create
enough separation and use a reverse key or backlight.
Blue or green “oversplash” on clothing
or hair can ruin all attempts at creating an appropriate
matte, especially on 4:1:1 signals.
If you have the opportunity to work
with DV50 footage, you will have more flexibility
than you will with DV25. You’ll also be
dealing with higher bandwidth video, which means
you’ll need a bigger pipe to transfer video
data and most probably, a third-party video card
to process and manage the data. There is a wide
range of third-party solutions for both Mac and
PC available. Whenever possible, use the component
connections (SDI) to connect DV50 edit or player
decks to your workstation.
Without doubt, DV50 images have
more data to work with – and as such, typically
retain more of it. In fact, most DV50 recorders
utilize two DV25 codecs to compress and decompress
DV50 material. The variations of formats within
the DV50 range are also extensive and we won’t
deal with that here, except to say that you can
work with either YUV or RGB native, lossless or
uncompressed video and 10, 12, or 16-bit processing
power. No matter the solution, it will offer higher
quality than DV25.
At the same time, DV25 offers a
better image than many professional analog formats
(some of them still in use today). The key is
to have a quality camera, a good shooter, strong
high-quality connections and a good DV codec in
place. There are many general distribution motion
pictures that include sequences, scenes or even
the entire film that was shot using DV25. Welcome
to the new world order.
Compression
This is also important when it comes to dealing
with video. We’ve already mentioned a few
codecs, but let’s briefly go over some compression
issues that may come into play as you work on
various DV projects.
There are many specific types of
compression, but two general processes. The first
type is typically called intraframe compression.
When manipulating video using intraframe compression,
each individual frame of video is treated separately.
As an example, Motion JPG (M-JPG) video uses intraframe
compression – typically in the range of
8:1 to 40:1. The 8:1 video will look pretty good.
The 40:1 video is really just to give you an idea
of the image when working with a nonlinear editing
system.
Interframe compression is unlike
intraframe compression in that it refers to changes
in the video data and only carries forward those
changes, allowing for highly increased compression.
Various forms of MPEG video use interframe compression.
It is possible to get signals with 200:1 compression
rates and a decent picture.
Codecs
Early codecs were used for very specific purposes.
There was a codec for use in offline nonlinear
editing, a different one for use in online nonlinear
editing – plus codecs for CD-ROM, distribution,
broadcast and so on. Of these, the M-JPG codec
may have been the most versatile as it could be
cranked up or reduced, depending on the application.
This was very convenient in early third-party
video cards used in both Macs and PCs and it was
how many early generation digital video productions
were created.
As I mentioned a few moments ago,
DV25 is not scalable (and DV100, DV50, DV25 are
not really scalable video as much as different
formats). Certain nonlinear applications such
as Final Cut Pro can render low-resolution proxy
files from source material – but this is
essentially a workaround (although a very acceptable
one).
YUV vs.
RGB
This is a critical part of preparing for professional
work using DV nonlinear editing solutions. When
using any number of image management applications
(Photoshop, After Effects, etc.), you have the
choice of selecting either YUV or RGB workspace.
Understanding the how and why is important, because
translating from one format to the other will
absolutely degrade the overall quality of the
original material.
YUV is the format video professionals
are used to working with. RGB is the format computer
designers are used to working with. If you’re
going to stay within the computer environment,
you can get away with RGB. If you’re going
to output to tape, you’ll need to use YUV.
Some very expensive systems for editing or compositing
video are RGB native, but convert the signal to
YUV on export – the cost of the conversion
gear keeps the reduction in quality to an acceptable
(read that as negligible) level.
Uncompressed
vs. Lossless vs. Compressed
Again, you could write a book about video formats
and the technical wizardry that goes on behind
the post-production scenes. But for our purposes,
let’s look at one more key issue: the number
of pixels being pushed around.
Uncompressed video (4:4:4) is a lot of data. A
native Mac or PC can’t even dream about
moving 2,000 by 2,000 pixels per frame around
on the screen. High-end nonlinear editors and
compositing workstations have support hardware
and software to move the data for the designer
or editor. Oh – and by way, uncompressed
doesn’t mean flawless. Every company that
offers “uncompressed” video management
is still using a technology of some sort to move
the video data.
Typically, in the personal computer
world, you’ll deal with some form of compressed
video. Most systems use M-JPG or DV compression,
which will range from about 3:1 all the way up
to 40:1. There are a few MPEG-2 nonlinear systems
that use a ratio of 2:1, but those typically will
not be in your line of work – although you
will output DV projects to MPEG-2 (for DVD work).
