The MPEG data (bit stream) can actually have the same number of pictures per second as the original material. The MPEG decoder will play these bit streams back at that frame rate. If the system in which the decoder is embedded must output to a different standard than the one in which the data it is playing was generated, it must perform frame rate conversions on that data. More on that later.
When film is transferred to videotape (via a process called telecine), the different frame rates of the source and destination must be reconciled. NTSC video runs at 30 fps, for example; thus, the 24 frames in a second of film must be adjusted to fit in 30 corresponding frames in video. This is accomplished by something called 3/2 pulldown, which duplicates fields in strategic ways to add the missing 6 frames each second.
Film destined for PAL videotape, in theory, needs to have the same sort of correction (24 to 25 fps) applied. In practice, it is normally ignored and the film is transferred frame for frame, resulting in a videotape show that runs faster than the film show. It stays in sync, but the running time on PAL tape would be 4% (1/25) shorter and the audio 4% higher in pitch. Sometimes the film is just shot at 25 fps so that the resulting transfer will be precise.
LetUs assume that we have videotape at either 25 or 30 fps that we wish to digitize and encode into MPEG bit streams. As explained above, we have three choices for the frame rate in MPEG: 24, 25, or 30 per second. If the videotape was created at its native frame rate (and not transferred from film), the answer is easyQjust use that frame rate in the MPEG data: 25 for PAL and 30 for NTSC.
The process is more complicated when we have videotape that contains film source material, especially in NTSC. Here we have videotape to which extra information has been added (6 frames per second) from the original in order to satisfy the gods of videotape. We could simply digitize the frames of video as they are and create a 30 fps MPEG bit stream. This would propagate the 3/2 pulldown frames into the MPEG domain. It works fine and the results can be satisfactory. But this depends heavily on the quality of the source and the editing history of the show.
Since with a lossy compression scheme like MPEG (as opposed to a lossless scheme), we are always fighting the quality battle at CD bit rates, there is another way to handle this case. During encoding, it is possible to remove the fields inserted during telecine and create the resulting MPEG bit stream at 24 fps, thus matching the original source material. This must be done by the encoding system and involves analyzing the digitized video and identifying the duplicate fields. It is called an inverse 3/2 pulldown operation.
There is another reason for, and benefit from, doing the inverse 3/2 pulldown in MPEG. By using this method, we create a 24 fps bit stream, which requires less data to represent a given running time than a 30 fps version. There are fewer frames and, therefore, fewer bits. This means that the encoder can dedicate more bits (25%) to each frame of MPEG data, yielding a lower compression factor and better image quality. In short, the pictures look better.
Now that we have MPEG data being decoded for playback at 24 fps, we might ask how that will ever work on a CD-I player that produces a standard NTSC or PAL/SECAM video output signal. The answer is that the player must guarantee that an MPEG bit stream of any frame rate will be displayed as well as possible on any TV system. The following chart summarizes the possibilities.
MPEG PAL TV NTSC TV
(25 fps) (30 fps)
-------------------------------------------
24 fps ??? pulldown 3/2 pulldown
25 fps ok ??? pulldown
30 fps drop 5 fps ok
As can be seen from this chart, in the case of 24 fps data playing on an NTSC TV, the player must do the same 3/2 pulldown originally performed during telecine. A similar pulldown operation must be done on other combinations where the data has fewer frames per second than are required by the display system. In one case (30 fps data on a PAL TV), the player must eliminate data; we might call this a RpushupS operation.
If we can assume that the player hardware does as good a job of these pulldowns and pushups as can be achieved in the post-production world, then there is no net loss in taking advantage of the better quality afforded by the lower frame rates.
But there is one further complication to this process. Let me first explain it and then discuss why it is a problem. It has to do with something called the "cadence" of a 3/2 pulldown.
During a telecine transfer of film to tape, the 3/2 pulldown is applied to the source material and a regular pattern of duplicated fields is established in the generated video. This regular pattern is called the cadence. If we want to return this piece of video to its original film mode, our encoder must detect the cadence and invert it, or remove the duplicate fields. This is fairly straightforward if the cadence is very regular.
But let's assume that after our film has been transferred to videotape, we want to edit it. This is very common, because editing is generally cheaper on tape than film. So now we assemble a final show on videotape composed of billions and billions of individual shots from different parts of our videotape. It is very easy to destroy the regular cadence in our finished show, since we are not guaranteed that all of our cuts land on parts of the 3/2 pattern that mate exactly with each other. It is possible to edit tape in such a way as to preserve the regular cadence of 3/2 pulldown, but it is much more time-consuming and expensive and, therefore, not always done.
If we now take this videotape and give it to our favorite MPEG encoder and ask the encoder to perform inverse 3/2 pulldown in order to generate a 24 fps bit stream, there may be problems. It is certainly more difficult to detect the cadence when it is irregular, and if this process is not done properly, picture quality can suffer.
One can imagine some really horrible scenarios--MPEG encoding nightmares. Consider a show in which there is incredibly rapid cutting, with source material mixed from film and all sorts of videotape formats--BetaSP, Hi8, 3/4 inch, and hand-held VHS camcorder. Music videos are famous for this. It's a lot harder to make good MPEG from something like that than it is from a high quality feature film mastered on D1.
What does this all mean to someone producing material destined for MPEG? There are no magic answers, really, just lots of tradeoffs to analyze and balance. Quality, budget, expediency, and other factors will all determine the choices to be made.
The purpose of this note is to give you some understanding of the issues so you can be equipped for the analysis.
Copyright ) 1993 Philips Interactive Media of America.
All rights reserved.
P RΓVε 1ì NP LÇÇÇ|yÇé~{w|{}ààåçåäüüêê}uw{âàÇxtsstz{x{~~~é~}ÇââàåâéâäüÇ~éÇ|~{|üüâé{srâ}wuåéäàåêêçâäèë~rpq{üumkq{Ç|vv{