Sony's IMX Format
This article was written by Alistair Jackson of EditHouse, and was first published in Digital Media World magazine in October 2002.
With so many video recording formats around, it is initially surprising to hear that Sony have added a new one, however, they have their reasons. IMX is squarely aimed at the gap between Sony's news gathering SX format and their stalwart Digital Betacam range. Priced accordingly between these two formats, and clearly intended to combat Panasonic's successful DVCPRO 50 systems, it has already found considerable success in Europe.
As a standard, it has been gaining industry acceptance, with both Avid NewsCutter Effects and Matrox DigiSuite and DigiServer adding support for IMX to their systems. An Ethernet board for IMX VTRs is in the final stage of development and will allow the use of standard high speed networks to connect VTRs and workstations.
With similar specifications to DVCPRO 50, IMX has the further advantages that its data format is fully MPEG-2 compliant. This supports Sony's use of MPEG-2 in their SX range, and provides a basis for future developments in their server and interface technology.
While MPEG-2 has been with us since the early 90s, new ways of using it are still being developed. That's because the various committees of the Motion Picture Experts Group (MPEG) make a point of only defining the specifications for the relevant MPEG data stream, whether it be MPEG-1, MPEG-2 or MPEG-4. They leave the whole issue of encoder and decoder implementation to the individual manufacturers. If it puts out a stream of bits that fit within the MPEG-2 specification, then it's an MPEG-2 compliant encoder, if it can decode any MPEG-2 bit stream then it's a compliant decoder.
MPEG-2 is generally seen as a distribution format, and as such it has been very successful as the video compression standard for DVD and for Digital Video Broadcasting (DVB). However, it has not been taken up so readily for content acquisition and post-production work. To understand why this is, it is necessary to understand how an MPEG-2 data stream is put together.
By analysing a series of concurrent pictures, MPEG-2 is able to choose the best encoding method for each frame. This series of individual pictures is known simply as a Group of Pictures or a GOP. The first picture in a GOP is always a full image compressed in a JPEG like fashion. Because it contains all the information required to make the full picture it is called an Intra Frame. This I-Frame is used to establish the base image. After that, the encoder has the choice of sending another I-Frame, or a P or B-Frame. A P-Frame is a Predicted frame which is based on the preceding I-Frame, and only includes the information that has changed from one picture to the next. A B-Frame is a Bi-directional predicted frame that looks for a changing path between both the picture that precedes it and the one that follows it.
Because of this there is a common belief that you can't edit MPEG-2 streams. While this is not precisely true, MPEG does add considerable complexity to the process. A device that can cut between any two MPEG-2 compliant video streams, requires huge amounts of processing power and memory to simultaneously hold both incoming GOPs and to then re-encode them as a new GOP, with appropriately place I, P and B frames. A further disadvantage is that for repeated edits, changing a frame backwards and forwards between different frame types will degrade the image.
However, Sony has cleverly taken advantage of the fact that while an MPEG stream can be made up of a series of I, P and B frames, it doesn't have to be. The standard simply says that a GOP must start with an I-Frame, which can then be followed by P or B Frames. The Betacam SX format creates MPEG-2 GOPs of only two frames - one I and one B. The higher quality IMX format has only one picture to a GOP - a single I-Frame.
By only using I-Frames, IMX does not have an issue with edits. In fact, we are back in the same ballpark as DV based formats. However, in this case we have an MPEG-2 compliant stream. The idea is that you can load this tape footage onto a disk, and you end up with an MPEG-2 file. It is not as small as an MPEG-2 file that takes advantage of P and B Frames, but it is compliant with the standard.
IMX is seen by Sony as a key element for is MXF (Material eXchange file Format) vision for converging broadcast quality video into an IT Infrastructure. A crucial part of this concept is the eVTR board, which allows IMX machines to interface to an Ethernet network. This allows for VTR control and for transfer of Audio and Video over a LAN, WAN, or even the Internet. The board buffers several frames from the tape, and if necessary pauses the tape until the buffer requires refilling.
Obviously this could be a very long process if you're feeding material via the Internet, but if you are using a GigaBit LAN, then you should be able to bounce multiple MPEG-2 streams around your facility real time. As IMX tape decks are backwards compatible, you are able to use it to put material from any Betacam tape on a network. For instance, put your SP tape in the IMX VTR, it converts it from analogue to its own flavor of MPEG-2, then the whole lot is fed onto your network.
While the impressive compression ratios available with MPEG-4 are currently the talk of the town, MPEG-2 will remain the key compression tool for broadcast television for many years to come. It rules supreme in the area of video distribution and, thanks to formats like Betacam SX and IMX, it is making further inroads into acquisition and post-production applications.
(Addendum: Since this article was written, SMPTE have ratified the IMX format with the nomenclature D-10, based on SMPTE standard 365M)