by Jim Mendrala
January 29, 1997
Television transmits a tremendous number of programs on film. More than 75% of prime time is originated on film. Most TV commercials are on film. TV with its higher frame rate of 30 frames per second (fps) requires that the 24 frame film image be converted to 30 fps with a 3:2 pull-down. DTV however takes the 3:2 pull-down out of the incoming video, converting it back to 24 fps when it enters the MPEG compliant encoder. Since 30 fps gives you better temporal resolution with live TV, we don't seem to realize that with film there isn't anything to be gained by converting 24 fps film to 30 fps. Film has always looks good on TV. Part of that is because each frame, of the 24 fps, was photographed in approximately 1/48 of a second and scanned in a progressive manner and converted to interlace for the NTSC system. Telecines do not have an image that starts off as an interlace. Think about that! The film frame is progressivly scanned. Today some of the newer TV cameras for surveillance applications are progressively scanned and the image output is in an interlaced mode. When a video tape of this type of image is displayed on an interlaced display in the "still" or "pause" mode, the image because of no interlace when captured looks twice as sharp even if the subjects were moving.
With today's new DTV standard it seems obvious that 24 frame progressive should be of great interest to the broadcaster. The images would be free of aliasing. If used in the NTSC format however each field if read out as interlace may contain small area flicker due to only half the samples in the vertical direction being displayed per field. The human eye has to integrate out aliasing effects per field, to see the full resolution of the picture. This aliasing is the interline twitter we have all seen. The flicker perception of the human eye demands a refresh rate of at least 60 times per second to prevent flicker. The eye has to try and filter out the 30 cycle flicker between the two fields. In order to save bandwidth, it was decided to refresh alternately the odd lines and the even lines. This means that for a given signal bandwidth the number of pixels in an interlace standard will be twice the number of pixels of a progressively scanned system, resulting in good spatial resolution but poorer temporal resolution.
But interlace scanning creates some well-known artifacts, especially with moving pictures. In a television image (not from film), it is impossible to combine two fields to make one frame or picture for moving objects, as each field comes from a different moment in time. But with progressive scanning, the image has all of the pixels captured at one time. There is no difference between that of film or the image acquired by the progressive CCD.
Compression of interlaced signals is a lot more complex and performs worse than compression of progressively scanned signals, where the entire image is captured at one time, akin to a film camera with its approximate 1/48 second exposure time.
Those that think that the flicker would be intolerable have to bear in mind that 24 frame progressive transmitted as NTSC would be read out as interlaced and 3:2 formatting put in upon transmission at 30 fps. This would produce the 12 cycle judder that we have become accustomed to in film transferres to TV. In the DTV standard, the decoder in the receiver would display the image similar to film image but at 72 times per second to the display, thus eliminating any perceptible flicker. (In a motion picture theater, the film running at 24 frames per second is displayed twice for each frame. The eye sees 48 frames per second on the screen. This was done very early in the film world, more than 95 years ago, to reduce flicker.)
Now if you take this concept up to HDTV, at 1080 progressive at 24 fps, you have a lot less to compress than at 30 fps with a 2:1 interlace, and the resultant image would look much better displayed at 72 fps, because of the absence of flicker, than the way film is transferred with a 3:2 pull-down putting in a 12 cycle judder pattern. HDTV at 1080i 1125/30/2:1 just would not hold up to 1080p 1125/24/1:1 for transmission of film transferred to video. (The same would hold true for those that want to transfer film at a 25 fps as in PAL countries.)
Compression systems of today extract the
3:2 pull-down and work only on the 24 frames per second. This allows more
bits to make a better picture than a live TV broadcast where the encoder
has to be pushed to the wall because of the differences in the interlaced
It is still true that certain things will have artifacts due to violation of Nyquest's rule in the temporal mode, such as the "wagon wheels" in a western will probably still rotate backwards. But, hey, we've lived with that for an awful long time and it doesn't really detract from the story. NTSC images from film have always had that problem. It's true that if you film at 30 frames per second or higher these artifacts tend to disappear the higher you go but at the expense of a lot more data to go along with it.
My conclusion is that since prime time
will mostly come from film in the early days of DTV, then it makes a lot
of sense to work at a 24 frame per second capture rate. There would be
more bits available for the MPEG compliant compression encoder, even at
the same variable or fixed bit rate. The progressive images at 24 frames
per second would be flicker free and would be displayed at the rate of
72 times a second, just like a computer monitor, instead of interlace at
30 times per second for each field. This would also require less bandwidth
for the same quality of picture. So for the NTSC transmitters, the quality
would be about the same as what we now accept as standard. But for
the new DTV, a virtually flicker free image would be displayed on the new
DTV receivers, even though the image was captured at 24 fps. When DTV in
the future wants to go to MPEG's HP@HL (High Profile @ High Level) or more
commonly known as the HDTV standard it will not be necessary to go to 60
fps at the transmitter. The receiver will take care of that.
The URL for this page is http://www.tech-notes.tv/Jim/Articles/24_Frame.html
Please use this URL in all links or references to this page
© 1998-2001 by James A. Mendrala