By Hiro Yamada, President and CEO of Canopus Corporation

Regardless of the compression format chosen for nonlinear video editing, the desired end product is usually a high quality master tape that can be duplicated for distribution or aired for broadcast with minimal loss in quality. The better the source material and the cleaner the editing chain, the better the end result will be. This has driven the highest end systems to handle video in a completely uncompressed format. Due to the lower cost of the equipment, most videographers use MotionJPEG- or DV-based systems for video editing. New MPEG-2-based editing systems are just coming on the market, so which format is really best for non-linear video editing?

For the last 7 years, Motion JPEG has been the compression format of choice for capturing and editing analog video. The lower data-rates, good image quality and lower cost editing equipment resulted in M-JPEG’s wide acceptance. However, the lack of good data-rate regulation limited M-JPEG to the PC environment, where an individual video frame could be any size, and prevented it from becoming a viable encoding standard for tape-based camcorders.

In 1996, Sony and Panasonic introduced the first DV camcorders and changed the world of the video professional and prosumer forever. For the first time, digital quality video recording was available at the then "affordable" price of $2500 to $4500 (now as low as $799 with Digital8). The DV format data size was smaller than M-JPEG and it captured video directly to tape with all analog to digital conversion performed by high quality circuitry in the camera. In addition, the DV camcorders introduced a completely digital interface called IEEE-1394, also referred to as iLINK or Firewire, to transfer the pure digital video and audio data directly to the PC in perfect lip-sync. Further improvements in camera optics and CCDs have produced a quantum leap in quality from standard analog S-Video/Hi8 sources.

The net result is that pure digital acquisition, editing and distribution is now possible at a very affordable price. The DV standard has eliminated the analog-to-digital and digital-to-analog conversions previously required with M-JPEG and lowered the data rate of the video stream down to 3.7MB/second versus the nearly 7MB/second for similar quality with M-JPEG. Because every analog-to-digital and digital-to-analog conversion reduces quality, the DV format retains the highest possible quality by keeping video in the digital domain throughout the acquisition process. The lower data rate means you do not have to spend as much money to configure a DV editing system because you can use lower cost components such as IDE RAID drives versus SCSI. Going forward, as video professionals add DV cameras to their analog cameras, it makes more sense to edit both the analog and DV content together in the DV format. This keeps the digital content pure and reduces the size of the compressed analog video in your system.

Recently, with the increasing acceptance of the DVD movie format and Direct Digital TV to the home, the MPEG-2 compression format has been considered as a possible video editing format. This seems logical because it offers superior quality compared to current VHS videotapes, broadcast TV and even Laserdisks. A significant obstacle to MPEG-2 is that, unlike DV where the data rate and image quality are consistent across all implementations, MPEG-2 image quality can vary significantly by the encoding system used and the processing power put behind it. The DVD movie you watch is the result of an intensive, multi-pass MPEG-2 encoding process to maximize quality at a specific data rate. Do not expect to see this level of quality on a low-cost, single-pass MPEG-2 editing solution.

Another obstacle to MPEG-2 editing involves the application of IBP frames in the video stream and the large amount of data lost in order to achieve the low data rate. An I-frame is a full video frame containing the same information that a M-JPEG frame would contain. In fact it is nearly identical in format and algorithm to the M-JPEG frame. P- and B-frames contain mostly frame differences and need to refer to the nearest I-frames to be reconstructed. P-frames reference the nearest previous I-frame while the B-frame refers to both the nearest previous and future I-frames to be reconstructed. Using all three frame types, MPEG-2 compression can be very high while maintaining exceptional image quality. Until recently, MPEG-2 editing systems edited in I-frame only, in effect, making them almost identical to editing in M-JPEG with the same higher data rates and the above mentioned analog-to-digital conversion issues inherent in the M-JPEG standard.

Newly introduced MPEG-2 editing systems use Main Profile at Main Level (MP@ML, IBP) or Simple Profile at Main Level (SP@ML, IP). The highest quality at the lowest bandwidth is MP@ML. The next step down is SP@ML. SP@ML, using I- and P-frames only, requires higher bandwidth and adds unnecessary loss. Since the P-frame is not a full image, the MPEG-2 decoder and editing software must reconstruct a full image from the last I-frame and the changed portion represented in the P-frame. A typical B- or P-frame contains only about 5% of the data found in an I-frame. The reconstructed frame seldom reproduces the same quality as a full I-frame. It is like looking at a photographic print. The photo image itself is very high quality, but making a copy, or editing the photograph should not be done from the picture itself. You must go back to the negative to achieve the highest quality. In the case of MPEG-2 IBP editing, the negative does not exist, it was never captured that way.

When all is said and done, the basic principles of video acquisition and editing have not changed. To achieve the best output quality, it is important to minimize the number of times you convert your source material to another format in your editing process. Each unnecessary digital-to-analog, analog-to-digital step, or transcoding between compression formats, decreases quality. In addition, the compression scheme must keep enough data available during the editing process to maintain image integrity.

The best option for the highest quality output is to maintain data in the original acquisition format if the source is digital, keep it in digital format throughout the editing process and keep it in the same digital format throughout the process.

With DV cameras becoming more and more affordable and the DV format offering low data rates while maintaining high image quality, the editing format becomes clear. Edit in DV. And since MPEG-2 is currently the best format for video archiving and distribution, the best solution is to edit in native DV format, and save the final master in MPEG-2.

Hiro Yamada is President and CEO of Canopus Corporation.

Based in San Jose, California, Canopus designs and manufactures award-winning graphics and video editing products for the PC.