Jump to content

720p V 1080i: Whats Better For Fast-action Sport?


Recommended Posts

Like I said, I want to see a greater verity of h.264 content before I come to any conclusions.

I can't find any 720p/50 h264 content. The stuff that I've seen so far is all 24fps. Even the Planet Earth Series is 24fps (according to the places I looked). Is there any true 720p/50 stuff or is it just upconverted from 720p/24? Even Microsoft's selection of HD WMV content is 24fps.

Link to post
Share on other sites
  • Replies 131
  • Created
  • Last Reply

Top Posters In This Topic

All,

As I have stated elswhere in this forum, all DVDs both SD and all HD formats are recorded at 24 or 25 frame/s. This is to minimise the data required and matches most of the program material which is ex 24 frame/s film.

So it is the player which decides the output frame/field rate, for all HD formats the rest of the signal specifications are identical between the Americas and the rest of the world. This is not true of broadcast signals.

As far as the portrail of motion is concerned, all cameras should have a shutter including electronic cameras. The amount of time the shutter is open is determined by the amount of light available and the setting of the camera apeture.

For example in still photography if you wish to freeze the motion of a moving car across the screen you use a very short time the apeture is closed eg 1/1000 second (1 ms). This needs full daylight particularly if the car is at a distance and a telephoto lens is required. For for motion pictures at 24 frame/s, you need to expose the CCD imaging device for 1 ms every 42.67 ms. This means that only 2.4 % of the available light is used.

The longer the time the shutter is open the more the image moves on the CCD imaging device causing blurring. The only exception to this is to use interlaced scanning which will halve the time between exposures thus using twice as much of the available light, at the expense of blurring the horizontal edges in the picture.

If this is done then all images will be sharp but the jerkyness of the motion will be determined by the speed of motion across the screen, the compression system used and the data rate available. To get these sharp images particulaly on horizontal edges you need to used progressive scanning in the camera.

At the display end the frame rate of under 31 frame/s produces intolerable flicker, so the display must either display each frame twice or store & display the static image and alter the image only where motion is occuring. The last method requires a memory address to control a pixel, so that scanning in the display is not required. This would mean that RAM memory and a Digital to analog converter would be required for each pixel instead of a dynamic memory. Obviously at this point this is not possible with current technology.

AlanH

Link to post
Share on other sites
As I have stated elswhere in this forum, all DVDs both SD and all HD formats are recorded at 24 or 25 frame/s.

If I understand your post correctly, then everything is recorded at 24 or 25 fps, but interlace format like 1080i which produces 50 or 60 fields per second would produce better motion portrayal (assuming one is using the best deinterlacer and scaler) than progressive format. I feel like an idiot for even suggesting 720p is better for fast action sport since every progressive frame is duplicated to avoid flicker. Which begs the question, why care about 1080p/50-60 when it's not really 50 or 60 original fps?

Link to post
Share on other sites


The 720p & 576p specifications are a FULL 50 frames per second. If the broadcaster decides to only show content which is 25fps and show every frame twice then that's their choice. Jeez, why do people have to make something so simple into something so complicated?

Link to post
Share on other sites
The 720p & 576p specifications are a FULL 50 frames per second. If the broadcaster decides to only show content which is 25fps and show every frame twice then that's their choice. Jeez, why do people have to make something so simple into something so complicated?

I don't know what channel 9 (90) run there 1080i at for the Rugby League, but it is magic on my Hitachi 42" 8800 plasma. Channel 7 (70) had the Rugby Union on last night and there was flicker in the background on the grass. Channel 7 need to lift their game.

Link to post
Share on other sites
Why is it that 720p and 1080i look so much better on a 1080p display. Is it more pixels that are filled that give the seamless pic or their handling of the signal.

Since 1080i is the same resolution as 1080p, you need a 1080p display to show it at full resolution. Lower resolution displays must downscale.

720p and even 576i/p can look better on a 1080p display if you sit very close, as the pixels on the 1080 display are smaller and less visible.

Link to post
Share on other sites
The 720p & 576p specifications are a FULL 50 frames per second. If the broadcaster decides to only show content which is 25fps and show every frame twice then that's their choice. Jeez, why do people have to make something so simple into something so complicated?

The only 576p I have seen in Oz has been bob deinterlaced 576i from SD video cameras.

True 576p does not seem to be used for anything, and even if it was, few people would pick it over 576i.

Link to post
Share on other sites


If I understand your post correctly, then everything is recorded at 24 or 25 fps, but interlace format like 1080i which produces 50 or 60 fields per second would produce better motion portrayal (assuming one is using the best deinterlacer and scaler) than progressive format. I feel like an idiot for even suggesting 720p is better for fast action sport since every progressive frame is duplicated to avoid flicker. Which begs the question, why care about 1080p/50-60 when it's not really 50 or 60 original fps?

Only film and some 1080 video is shot at 24/25 progressive frames per second.

Most SD and HD video cameras run interlaced and therefore have 50 motion updates per second for much smoother motion.

If you ever saw sport shot at 25fps, you would be very unimpressed, interlaced formats are MUCH better for sport unless the progressive format delivers 50 or 60 frames per second.

True 720p video from a video camera does not repeat frames.

Video does not flicker only displays do, and most digital displays cannot flicker at any frame rate.

Only CRT displays require high refresh rates to avoid flicker. Digital displays do not refresh they update, which is very different.

If a still image is displayed on a CRT it must be constantly redrawn (refreshed) at a fast rate to avoid flicker, but when most digital displays are showing a still image, no refresh is required as no pixel data has changed and the pixels stay set at their last state until they are required to change to a new value, hence no flicker. Repeating frames on a digital display therefore serves no purpose.

Link to post
Share on other sites
The 720p & 576p specifications are a FULL 50 frames per second. If the broadcaster decides to only show content which is 25fps and show every frame twice then that's their choice. Jeez, why do people have to make something so simple into something so complicated?

Exactly.

Useful software to check out whether a 50p transmission is really 50p or just (25x2)p is VideReDo Plus.

Free to Air

This afternoon, ABC HD showed Art Museums of the World - Musee D'Orsay: Renoir. This was "filmed" in Europe by a Japanese production company. Aunty broadcast this at 50 distinct frames per second. I have verified this by stepping through parts of an off-air recording of it, frame by frame.

Tonight's Ch 7 HD news bulletin was at 50 distinct frames per second most of the time, even reports from journalists in the field. Archive material (old black and white film) came in pairs of [almost] the same frame, which is to be expected. Towards the end of the bulletin, a segment of last night's Rugby match in Brisbane against Wales was shown. This started out in a jittery fashion basically at 50 unique frames but with some frames repeated [as decoded and reconstructed by VideoReDo Plus]; though at all times the ticker at the bottom of the screen scrolled at a steady 50 unique fps. After a few seconds the feed stabilized and everything was at 50 unique frames a second.

So congrats to AuntyHD and 7HD for providing us with some genuinely 50p, 50p material.

American material

For those without access to their own off-air recorded transport streams of genuine progressive material, there's a clip of some 59.94fps basketball referred to here. The file, clip.ts, is 25MB and is at a genuine 59.94 unique frames per second, in a 1280x720 format. [it appears to be from a cable or satellite service.] I've sampled some of this frame by frame, and have noticed that the players move quite some significant distance, even in 1/59.94 of a second!

As already noted in this thread, how much better [or worse] progressive transmission at around 50 or 60 unique fps is to view, compared with its interlaced equivalent in conjunction with an adaptive deinterlacing algorithm at the viewing end, is another question, partly dependent on the data rate available for the encoded transmission.

Link to post
Share on other sites
So congrats to 7HD for providing us with some genuinely 50p, 50p material.

This is a bit off topic...but i dont think Channel 7 deserve congratulations for ANYTHING they do with broadcasting, let alone their bastardized HD efforts. hehe :blink:

Link to post
Share on other sites

Re 7, I was surprised to find their in the field journalists are capturing video at a genuine 50fps. Given 7's attitude to HD I thought they might have settled for 720x576 i

Link to post
Share on other sites


American material

For those without access to their own off-air recorded transport streams of genuine progressive material, there's a clip of some 59.94fps basketball referred to here. The file, clip.ts, is 25MB and is at a genuine 59.94 unique frames were second, in a 1280x720 format. [it appears to be from a cable or satellite service.] I've sampled some of this frame by frame, and have noticed that the players move quite some significant distance, even in 1/59.94 of a second!

As already noted in this thread, how much better [or worse] progressive transmission at around 50 or 60 unique fps is to view, compared with its interlaced equivalent in conjunction with an adaptive deinterlacing algorithm at the viewing end, is another question, partly dependent on the data rate available for the encoded transmission.

Thanks for the link MLXXX. The clip is relatively short but should be good enough for me to analyse. Just have to figure out how to play it on my TV. Playing it on my pc seems pretty clear and sharp but would like to see it on my 37" LCD TV.

Since true 720p/50-60 footage actually does exist, I'll just have to try and keep hunting for some more sample clips. For some reason it's not a commonly uploaded format (most of it is 24fps and not worth downloading since I want to compare true 720p/50-60 and 1080i/50-60 and then draw my own conclusion).

Link to post
Share on other sites
As far as the portrail of motion is concerned, all cameras should have a shutter including electronic cameras. The amount of time the shutter is open is determined by the amount of light available and the setting of the camera apeture.

For example in still photography if you wish to freeze the motion of a moving car across the screen you use a very short time the apeture is closed eg 1/1000 second (1 ms). This needs full daylight particularly if the car is at a distance and a telephoto lens is required. For for motion pictures at 24 frame/s, you need to expose the CCD imaging device for 1 ms every 42.67 ms. This means that only 2.4 % of the available light is used.

The longer the time the shutter is open the more the image moves on the CCD imaging device causing blurring. The only exception to this is to use interlaced scanning which will halve the time between exposures thus using twice as much of the available light, at the expense of blurring the horizontal edges in the picture.

If this is done then all images will be sharp but the jerkyness of the motion will be determined by the speed of motion across the screen, the compression system used and the data rate available. To get these sharp images particulaly on horizontal edges you need to used progressive scanning in the camera.

Yes, shutter speed is often an extremely significant limitation with respect to motion unless a lot of light is available, but I can’t see how it has any effect on motion smoothness, which is strictly dependant on frame/field rate.

Video compression at restricted data rates absolutely destroys resolution in fast motion, to the point that interlacing is irrelevant when it comes to motion blur.

High data rate true 1080i with high quality deinterlacing provides extremely sharp and detailed images with no discernable loss of image sharpness on horizontal edges.

True 1080i from a video camera usually looks more impressive then 1080p from film, and certainly has vastly superior motion smoothness.

1080p 50/60 would be a little better if it where available, but it just isn’t and wont be in the foreseeable future.

At the display end the frame rate of under 31 frame/s produces intolerable flicker, so the display must either display each frame twice or store & display the static image and alter the image only where motion is occuring. The last method requires a memory address to control a pixel, so that scanning in the display is not required. This would mean that RAM memory and a Digital to analog converter would be required for each pixel instead of a dynamic memory. Obviously at this point this is not possible with current technology.

This does not make sense, and I’m surprises you posted it. Frame rate and flicker are in depended of each other. Most digital displays are completely incapable of flicker at any frame rate. A still image displayed on an LCD for example with look identical at 1Hz, 10Hz, 100Hz or 1000Hz refresh rates.

Since the data stored in the frame buffer is identical to what is already displayed, no pixels change at the refresh interval (no update occurs).

Link to post
Share on other sites
Exactly.

Useful software to check out whether a 50p transmission is really 50p or just (25x2)p is VideReDo Plus.

Free to Air

This afternoon, ABC HD showed Art Museums of the World - Musee D'Orsay: Renoir. This was "filmed" in Europe by a Japanese production company. Aunty broadcast this at 50 distinct frames per second. I have verified this by stepping through parts of an off-air recording of it, frame by frame. On the other hand, Ch 7 HD broadcasts House, produced in Los Angeles, at only 25 unique frames per second.

Tonight's Ch 7 HD news bulletin was at 50 distinct frames per second most of the time, even reports from journalists in the field. Archive material (old black and white film) came in pairs of [almost] the same frame, which is to be expected. Towards the end of the bulletin, a segment of last night's Rugby match in Brisbane against Wales was shown. This started out in a jittery fashion basically at 50 unique frames but with some frames repeated [as decoded and reconstructed by VideoReDo Plus]; though at all times the ticker at the bottom of the screen scrolled at a steady 50 unique fps. After a few seconds the feed stabilized and everything was at 50 unique frames a second.

So congrats to AuntyHD and 7HD for providing us with some genuinely 50p, 50p material.

American material

For those without access to their own off-air recorded transport streams of genuine progressive material, there's a clip of some 59.94fps basketball referred to here. The file, clip.ts, is 25MB and is at a genuine 59.94 unique frames per second, in a 1280x720 format. [it appears to be from a cable or satellite service.] I've sampled some of this frame by frame, and have noticed that the players move quite some significant distance, even in 1/59.94 of a second!

As already noted in this thread, how much better [or worse] progressive transmission at around 50 or 60 unique fps is to view, compared with its interlaced equivalent in conjunction with an adaptive deinterlacing algorithm at the viewing end, is another question, partly dependent on the data rate available for the encoded transmission.

Hang on mate; are you certain the footage you saw was true 576p?

It was more likely 576i bob deinterlaced to 576p, which provides 50 different and distinct frames per second, just like 576i deinterlaced at the consumer end.

I would be VERY surprised if anything is shot in 576p 50.

720p 50 or 1080i 50 downscaled to 576p 50 is MUCH more likely.

Link to post
Share on other sites


Hang on mate; are you certain the footage you saw was true 576p?

It was more likely 576i bob deinterlaced to 576p, which provides 50 different and distinct frames per second, just like 576i deinterlaced at the consumer end.

I would be VERY surprised if anything is shot in 576p 50.

720p 50 or 1080i 50 downscaled to 576p 50 is MUCH more likely.

The first thing I needed to do was deactivate the de-interlacing of VideoReDo Plus for playback. Whilst the de-interlacer cleaned up the herringbone or comb effect on 1080i material, it played havoc with the 59.94p and 50p material, combining adjacent horizontal lines and halving the vertical resolution. I note that the de-interlacer is probably of a basic design, as VideoReDo is not marketed as a viewer.

When the de-interlacer was switched off, vertical resolution was restored with the progressive material.

Whilst I would not swear on a stack of bibles, as far as I could tell when examining individual saved frames (using PaintShopPro), there was as much detail vertically as horizontally. I could see no sign of compilation from an interlaced source, even where there was rapid panning.

So I will say that the Ch2HD Art Museums of the World (1280x720p 50) program, and most segments of the Ch7HD news, were at 50 unique full frames per second and not artificially compiled from 50i. Similarly the American 59.94 basketball clip appears to be genuinely at that rate, with no duplication of frames or compilation from a similar interlaced source.

As you say, it seems more likely the various video cameras (studio, field, etc) in use by Ch7 would operate in a more recognised format than the 576p which Ch 7 has adopted for its so-called high-definition transmissions.

You raise an interesting point re 1080i 50. That could indeed be bobbed and rescaled down to a reasonably close approximation to 576p. I could not say that was not happening with the Ch7HD transmissions.

The important point though is that there is no suggestion (not in what I have seen) of the 720x576p having been manufactured by starting with 720x576i and attempting a sophisticated interlace exercise at the broadcast end so as to generate a pseudo 720x576p.

Of course people at Ch7 are probably smirking at our musings. They would know what is being broadcast (well at least some of them would I'm sure). I note that the test transmissions of a few months ago in a higher resolution format seem to have ceased.

I also note that some Ch7HD programs are telecast with repeated frames, e.g. Stargate Atlantis, such that there are only 25 unique frames per second. This is not immediately noticeable when viewing but is quite clearly so when examining the frames with appropriate software. The repeated frames are not quite identical to each other as decoded with an MPEG codec, but for practical purposes are the same.

Link to post
Share on other sites
American material

For those without access to their own off-air recorded transport streams of genuine progressive material, there's a clip of some 59.94fps basketball referred to here. The file, clip.ts, is 25MB and is at a genuine 59.94 unique frames per second, in a 1280x720 format. [it appears to be from a cable or satellite service.] I've sampled some of this frame by frame, and have noticed that the players move quite some significant distance, even in 1/59.94 of a second!

As already noted in this thread, how much better [or worse] progressive transmission at around 50 or 60 unique fps is to view, compared with its interlaced equivalent in conjunction with an adaptive deinterlacing algorithm at the viewing end, is another question, partly dependent on the data rate available for the encoded transmission.

Alright i had a look at that clip and my conclusions were this:

Motion looked very smooth, you notice that right away that it seems 'cleaner' some how in the movement.

But the encode is terrible in all other aspects. Pixellation surrounding every player, that follows them like a kind of digital "ghosting". Image seemed too sharp also...to the point of harshness. Very unpleasing to the eye in almost every respect except the motion.

If that is 720p...i think i prefer 10's 1080i.

Link to post
Share on other sites

Yes that basketball 720p clip is pretty poor, apart from its smoothness. But I've just had another look at the 720p program I happened to record yesterday afternoon from Ch2HD (Museum program). It opens with some fast zooms and pans and has a solid reality to it. Best PQ I've seen from Ch2HD.

Link to post
Share on other sites
Whilst I would not swear on a stack of bibles, as far as I could tell when examining individual saved frames (using PaintShopPro), there was as much detail vertically as horizontally. I could see no sign of compilation from an interlaced source, even where there was rapid panning.

Simple bob deinterlacing involves no “compilation”, there will be no sign of interlacing.

If the source was 1080i 50, there will be no visible loss of vertical detail when converting to 576p, and there will be truly 50 distinct frames per second.

So I will say that the Ch2HD Art Museums of the World (1280x720p 50) program, and most segments of the the Ch7HD news, were at 50 unique full frames per second and not artificially compiled from 50i.

Although it is possible the Art Museums program was shot in 720p, it’s highly unlikely.

True 720p video cameras are rare, most are really 1080 cameras that can down convert to 720p for output.

720p is a poor choice of video format for that kind of program; 1080i would be much more appropriate due to its superior detail.

If you are going to shoot HD video in an art museum, you want as much detail as you can capture, fast motion is the least of your worries.

The important point though is that there is no suggestion (not in what I have seen) of the 720x576p having been manufactured by starting with 720x576i and attempting a sophisticated interlace exercise at the broadcast end so as to generate a pseudo 720x576p.

I have not seen the particular programs you are referring too, but I have seen plenty of bob deinterlaced content from 576i source on channel 7, and I have seen no evidence of any “sophisticated” deinterlacing used at the broadcast end.

The only real give away to bob deinterlacing is some vertical jitter on bright horizontal edges.

Bob deinterlacing never results in combing or other typical interlacing artifacts and always provides 50 real frames per second if the source was a 576i 50 or 1080i 50 video camera.

If the source was 1080i 50, it’s very unlikely you would be able to see any vertical jitter and resolution should be just as good as genuine 576p, again with 50 real and distinct frames per second.

I also note that some Ch7HD programs are telecast with repeated frames, e.g. Stargate Atlantis, such that there are only 25 unique frames per second. This is not immediately noticeable when viewing but is quite clearly so when examining the frames with appropriate software. The repeated frames are not quite identical to each other asdecoded with an MPEG codec, but for practical purposes are the same. House on Ch10HD has similar repeated frames, but of course in a higher resolution pixel format.

All film or 24fps content will have frames repeated to generate 576p 50 or 720p 50.

A way around this terribly bandwidth wasting situation is to use frame repeat flags in the video stream and not transmit the repeated frames. The decoder reads the flags and repeats the frames it is told to, so there is no need to transmit repeat frames.

If a flag system was in use, the repeat frames would be pixel for pixel identical, which would be easy to check in a photo editing program. From what you said they are not, so wasteful repeat frames are being transmitted, so half the transmission bandwidth is being wasted when film source is being shown. What a great idea that is.

Here is a link to an old post of mine discussing blur issues with 720p video in real world situations

http://www.dtvforum.info/index.php?s=&...st&p=446098

Link to post
Share on other sites
Yes progressive formats are more efficient to compress, but not that much more efficient.

... Even using the full 20Mb (approx) available bandwidth for TV transmission with h.264, 1080p 50 will be bit rate starved.

I have read that for similar image quality 1080i compresses ~70:1 and 720p compresses ~200:1. Unfortunately I can't track down the reference ... It isn't really a fair comparison anyway as 1080i has more horizontal detail to cope with. The better comparison is compression of 1080i50/60 vs 1080p50/60 using MPEG2. Such a comparison I haven't found yet; but I did find this ...

People in the EBU have said that with MPEG4 1080p50 should end up only 20% bigger than 1080i (ref 1, ref 2), but in some cases users found 1080p50 @ 8Mb/s as good as 720p50 @ 6-8 Mb/s. There are several interesting caveats in their test methods (e.g. they used 2160p50 source material downscaled and in some tests the oversampling involved skewed the results), but they also stated:

1080p/50 provides more information in the spatio-temporal domain and encoders can conduct the compression more efficiently. However, at the lower bitrates (i.e. <10...13 Mbit/s) the 1080p/50 encoder becomes more overloaded with information, depending on the content, and this information overload appears to become the dominant factor affecting quality.

To me, this implies that 1080p @ 12-15Mb/s should generally be OK (especially as the codecs mature).

Papers like the one you linked are based on the use of 1080i CRT displays and predated advanced deinterlacing systems and 1080p displays. We have moved past that stage.

Two things:

a> Yes we have moved on, that was kind of my point. Interlacing was used in the 1930s to deal with problems encountered with the technology of that era. By the time the US FCC established the Advisory Committee on Advanced Television Service in 1987, one problem had gone (phosphor persistence); and by the time the prototype systems were developed MPEG2 had arrived to help address bandwidth.

b> Given that you have quite rightly pointed out on many occasions (at least six times already this year) that it is the majority of displays that are still failing to do the deinterlacing properly, I don't believe that you can really claim that CRT-era (de)interlacing problems are behind us.

The best deinterlacers are still only available in stand alone video processors or PC’s fitted with the best of the new video cards.
The better pixel adaptive systems are normally only available in the higher end PC video cards and stand alone video processors, and the very advanced motion compensated systems are strictly in the realm of high end video processors costing big money. ... A lot of the criticism of 1080i is out of date and does not take into consideration the advancements in deinterlacing technology. When you see what can be done with true 1080i and high quality deinterlacing, it’s hard to believe people would find 1080p 24 preferable. ... People buying cheap 1080p displays should be aware that deinterlacing performance on those sets is very likely to be second rate. These sets will be outperformed by 768p displays that have better deinterlacing for 1080i.
Most people have never seen what 1080i is capable of, as very few if any digital displays handle 1080i properly.

High end PC video cards and stand alone video scalers can do a significantly better job.

The people that chose 1080i as the standard for HD where not as stupid as people think, its just that digital displays that can handle it properly have been very slow coming.
... 1080p is just much easier to deal with. ... For 1080i video source the big issue is deinterlacing quality, and a lot of displays fall short in that area.
All [1080p] film source that has been encoded as 1080i can be simply deinterlaced ... The only problem has been that almost all displays could not perform this simple operation, although PC’s have been doing it in software for years without issue.
h.264 compression is not the savior many would think. I have the same content in 1080i 60 18Mbit Mpeg2 from US pay TV and 1080i 50 10Mbit h.264 from European pay Sat, and the Mpeg2 version is way better then the h.264 version.

The h.264 version looks fine when there is no movement, but on scene changes and fast motion it breaks up intolerably, where the higher data rate Mpeg2 version does not.

The h.264 compressed version also suffers from significant posterization, a problem I have noticed on most h.264 content and other Mpeg4 systems, even at quite high data rates.

Mpeg2 may be old, but given sufficient data rates it performs outstandingly well.

I have a large collection of HD content and in my experience there is not substitute for data rate. Fancy compression systems offer better performance at very low data rates, but picture quality is significantly compromised.

Those are very good points, and they show that advanced compression methods need to mature more before HDTV v.2 could be rolled out (the problem of always waiting for the Next Big Thing aside :-).

_____________________

One thing I should make clear is that I believe that the whole HDTV design was f**ked up by the Advanced Television Systems Committee (which, don't forget, was a private sector organisation -- it isn't just government committees that turn horses into camels). Yes they did some very very fancy design work but they obviously could never reconcile the progressive-scanners vs we've-got-to-out-do-MUSE-with-a-bigger-raster camps. The result -- a "standard" with something for everyone.

Supposedly, 1920 comes from the standard-definition 4:3 horizontal pixel count of 720 getting reworked for 16:9 and the resolution doubled (i.e. 720 x 3/4 x 16/9 x 2 = 1920 pixels/line). I'd have expected criteria to be used such as:

  • at least 4x the resolution (to beat MUSE :-)
  • simple fractional upscaling of SD content
  • convergence with computers (admittedly something elements of the TV industry were dead against (p.13))
  • easy bandwidth handling (late 1980s technology could handle at least 1.5Gb/s, but not 3Gb/s (ref))
  • scaleable for basic digital cinema (full colour + panoramic widescreen).

A solution such as 1536x864p (4:2:2 YUV) in a 1750x900/60Hz frame or a 2100x900/50Hz frame (and "dual-link" extended to 2048x864p (4:4:4)) pretty much hits all those criteria. Yeah, the raster is smaller than 1920x1080, but the ATSC wanted the biggest and the best, despite the fact that it resulted in a decade of criticism about how damn hard it was to make decodesr that handled all the variations and 1080i displays that were affordable for end users.

Adrian

Link to post
Share on other sites
A way around this terribly bandwidth wasting situation is to use frame repeat flags in the video stream and not transmit the repeated frames. The decoder reads the flags and repeats the frames it is told to, so there is no need to transmit repeat frames.

If a flag system was in use, the repeat frames would be pixel for pixel identical, which would be easy to check in a photo editing program. From what you said they are not, so wasteful repeat frames are being transmitted, so half the transmission bandwidth is being wasted when film source is being shown. What a great idea that is.

The precise processes involved in MPEG2 encoding are quite complex. My understanding is that if there are only slight changes between frames, the MPEG encoding of the second frame need only represent those changes. There would appear to be a loss of efficiency compared with use of a zero change frames [or flagged identical frames], but not as severe a loss as a halving of efficiency.

Here is a link to an old post of mine discussing blur issues with 720p video in real world situations

http://www.dtvforum.info/index.php?s=&...st&p=446098

As I recall, a few months ago I asked Owen to advise whether the pictures he had referred to in that post were keyframes (I) or calculated (P or B), and he advised they were keyframes he had extracted from the source.

Examination of frames other than I-frames can be misleading, as the MPEG codec can introduce blur in computing a frame within the available bandwith constraints for the encoding. Posterization is a well known problem with limited bandwith MPEG encoding. It affects those parts of the image that are changing in fast irregular ways.

An option in VideoReDo Plus allows it to advance to the next I-frame at the press of the right arrow key. I did that with the brief basketball clip I referred to last night, hoping to be able to find a frame of the basketball in mid flight. Unfortunately the I-frames came when the ball was either being held or travelling slowly. In any event it is a poor quality clip: even the I-frames don't contain much detail.

If I have the time and inclination I might examine some of the off-air Ch 7HD sports footage. If I can find an I frame that includes fast motion, and that frame is not blurred it will tell us that exposure time can be short enough not to cause blurring. This would be a different result to what Owen found in the clip he commented on in the post he has provided the link to.

And if further examination of the Ch7HD frame revealed good vertical resolution, it would suggest that the frame was not generated by a bob of alternating even and odd lines of a 576i original. I note that such examination would be unlikely to prove that the source was merely 1080i converted to a good approximation to 576p. I think only the Ch7 engineers could tell us that! Within the MPEG2 bandwidth constraints in use, the video from a 1080i camera downconverted to 576p may be indistinguishable from the video from a genuine 576p camera (should such cameras even exist in the Australian broadcasting system).

Link to post
Share on other sites
The precise processes involved in MPEG2 encoding are quite complex. My understanding is that if there are only slight changes between frames, the MPEG encoding of the second frame need only represent those changes. There would appear to be a loss of efficiency compared with use of a zero change frames [or flagged identical frames], but not as severe a loss as a halving of efficiency.

Yes you are quite right there mate, the Mpeg compression system should not devote too many bits to a repeated frame, I forgot about that.

Link to post
Share on other sites
...

People in the EBU have said that with MPEG4 1080p50 should end up only 20% bigger than 1080i (ref 1, ref 2), but in some cases users found 1080p50 @ 8Mb/s as good as 720p50 @ 6-8 Mb/s. There are several interesting caveats in their test methods (e.g. they used 2160p50 source material downscaled and in some tests the oversampling involved skewed the results), but they also stated: ...

Thanks Adrian. That first reference in particular makes fascinating reading. More experimental method than you can poke a stick at. :blink:

Link to post
Share on other sites
Although it is possible the Art Museums program was shot in 720p, it’s highly unlikely.

True 720p video cameras are rare, most are really 1080 cameras that can down convert to 720p for output.

720p is a poor choice of video format for that kind of program; 1080i would be much more appropriate due to its superior detail.

http://www.dtvforum.info/index.php?s=&...st&p=446098

This program was shot in 1080i60 (or at least finished in i). They then made a 60-50 digital conversion, without decent motion compensation. The Aust copy was 1080i50

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Recently Browsing   0 members

    No registered users viewing this page.





×
×
  • Create New...