Being in the digital domain, with the creative and distribution options it brings, is a good thing. However, it’s still a fact — even more so with the advent of HD — that compression is a necessary evil. With a wide choice of compression schemes available, and even more on the horizon, we could be looking at a future that is not so much Variable Bit-Rate as Very Baskin-Robbins.
It remains impossible to record full bandwidth digital signals on something you can carry around; you have to reduce the number of bits you need to record. Compression is the enabling technology that allows field shooting of digital video, which in turn has made ENG possible and allows us to use camcorders for dramas and documentaries, even in HD.
The basic premise of compressing any digital file is to remove redundant information — something you will not notice if it is not there anymore, or can easily reconstruct later — to save bandwidth. The same concept applies when you Zip a Word document, applies to motion pictures. However, in the case of video it means analyzing what is in the picture and calculating (some would say guessing) what the eye would not notice.
The easiest compression trick is to reduce color resolution, because the eye and brain resolve fine detail in luminance and are less sensitive to color. From the beginning we got used to 4:2:2 sampling, i.e. taking the color from every other pixel. Some compression schemes are even more brutal, using 4:1:1 or 4:2:0 sampling.
Then the compression engine has to look at content of the picture and describe it not pixel-by-pixel, but using a series of mathematical representations. Until, now compression schemes favored a process called discrete cosine transforms or DCT, essentially breaking down the picture into blocks. MPEG-2 and DVC compression schemes both use DCT compression, which can result in the familiar on-screen blocking errors in over-compressed MPEG-2 broadcasting.
DVC codecs have higher bit-rates, applying compression to each and every frame, while MPEG-2 employs compression over multiple frames, using a group of pictures (GOP) with an Intra-coded I-frame which describes the whole image, while the rest of the group just describes the changes from the I-frame. The more pictures there are in the group, the better the compression efficiency achieved.
Long-GOP compression is fine for transmission, but not so fine if you want to edit it. Most non-linear editing systems have solved the technical issues of reconstituting all the pictures to allow you to cut wherever you need — so-called ‘native editing’. At the same time, most quality issues are not immediately apparent, especially if you stick to cuts and simple transitions, for example in news and sports. The simple, every-day fade-to-black is still a challenge for long-GOP codecs however, and keying and complex transitions only serve to highlight the fragile nature of the codec.
Newer formats such as AVCHD, the long-GOP Sony/Panasonic consumer HD camcorder recording format, are about twice as efficient as MPEG-2 long-GOP. Unlike HDV, which is tape-based, AVCHD is disk-based and therefore more naturally non-linear-friendly. However, there is a drawback, as the computer horsepower required to unravel these formats increases dramatically.
(click thumbnail) |
| Grass Valley Edius | |
So, while ‘native editing’ is good, it has very specific uses and in some instances is still not yet viable, even with the fastest computers. This is an area in which the EDIUS™ family from Grass Valley™ excels, as it is built on a multi-format codec engine that gives the best of both worlds. It offers native editing where the application is suitable, and conversion to a more agile codec — in this instance the Canopus-engineered HQ codec - for more demanding effects and compositions or simply to provide the fluid editing experience we come to expect from non-linear editing, as in the case with AVCHD.
With realtime alpha-channel support built in, EDIUS’ HQ codec also caters for graphics-intensive work where the burden of uncompressed HD will slow most editing systems to a crawl.
We are also beginning to see a new generation of compression systems, tailored to the specific needs of acquisition. JPEG2000 has been chosen by Grass Valley as one of the formats for its Infinity™ digital media camcorder. JPEG2000 does not use DCT, but wavelet compression, a different mathematical process that allows it to work on the whole picture in a single pass.
This eliminates blocking artifacts, and the underlying mathematics mean that, under extreme pressure, picture degradation appears only as a slight softening of the image, not the blocking and freezes that we see with DCT schemes. It is an intra-frame compression scheme, so there are no GOP problems.
(click thumbnail) |
| Grass Valley REV PRO | |
Because it uses 4:2:2, ten-bit colour sampling, it captures much more picture detail than current camcorder codecs which typically only offer 4:2:0 at eight bits. This extra resolution really matters in post-production processes such as colour matching and keying because there is so much more colour information to work with.
Panasonic have also introduced a new HD codec - AVC Intra — based on the MPEG-4 AVC/H.264 compression scheme and claiming to be up to three times more efficient than MPEG-2. As its name implies, it uses intra-frame compression, and offers two modes: Full-resolution HD (100Mb/s, 1920 x 1080 / 4:2:2 @ 10-bit) and ENG/Economy HD Mode (50Mb/s, 1440 x 1080 / 4:2:0 @ 10-bit). In the same way that JPEG2000 is stored to a REV PRO disk or CompactFlash® on the Infinity platform, AVC-Intra also uses non-tape-based storage, solid-state P2.
While JPEG2000 and AVC-Intra formats both offer improved picture quality and more efficient compression, like AVCHD, both have an overhead in terms of the CPU power needed to decode them, which puts the onus on the NLE to deliver both native editing and an agile codec for more intensive editing work. At NAB 2007, Grass Valley will be previewing EDIUS version 4.5, which has the ability to edit all these formats.
The need for compression shows no signs of diminishing so it will remain necessary for the foreseeable future. The key for technological developments in the area is to make it less of an evil.