48 replies to this topic

[alanh]

Hi,
I am watching the tennis and have noticed the following;

• The line markings have smooth edges even when the camera is panned. This is not possible with interlaced scanning image capture.

When the progressive image is reread to convert the image to interlaced, both fields represent the same point in time. As a result the display on a flat screen TV is progressive at a frame rate of a multiple of 25 frame/s. Each frame may be displayed twice or 4 times as identical frames or with motion compensation.

• Non jerky movement of the ball. There is a bit of lag behind the ball movement due to the low light level with the roof closed. There is no jagged edges on the moving ball either.
  
• Slow motion replay for normal motion is smooth because more images are captured per second. This means less missed motion between captures.
  
• The prerecorded package used whilst the roof was being closed contains very smooth very slow replays shows that they are also using the 100 frame/s digital camera.

Pity it is shown at a lower data rate.  This is evidenced by the grass being blured when the camera is moving and then becomes sharp enough to see the blades of grass.

The original is in HD for the UK market, pity that 7 did not transmit it on 7 Mate. I suppose the higher satellite costs from the UK to Melbourne would have prevented this.

AlanH

Edited by alanh, 09 July 2012 - 02:47 AM.

[Malich]

alanh, on 09 July 2012 - 02:44 AM, said:

• Slow motion replay for normal motion is smooth because more images are captured per second. This means less missed motion between captures.
  
• The prerecorded package used whilst the roof was being closed contains very smooth very slow replays shows that they are also using the 100 frame/s digital camera.

As an interesting aside, you might be surprised by what's possible using motion interpolation software these days.

The video in the link was filmed with a Canon 60D at 60p, with motion interpolation done in Twixtor to slow it down to 1~2% of actual (equivalent to shooting at 300~600fps). Shutter speed is the key in order to prevent blur - for that video, shutter speeds of 1/2000 - 1/4000 were used. If you look carefully, there's only one shot where blur was masked in to hide some warping.

[alanh]

Motion interpolation is an "educated" guess. The high capture frame rate is the real thing. That's why movie industry's leading producers want to increase the frame rate to 48 or 60 frame/s.

You cannot use a shutter speed of 1/2000  - 1/4000 when there is no supplementary lighting, and the light is coming though dense clouds and a clear roof. After all it was heavily raining.

Alanh

[Malich]

I wasn't suggesting that's what they used at Wimbledon, or that motion interpolation wasn't an educated guess - just pointing out that modern motion interpolation techniques, available even to amateurs, can produce results that are near-indistinguishable from filming at high frame rates.

For the rest, it seems you didn't take the time to look at the video in my link - a US$1000 consumer DSLR was used to record the video at 60fps progressive, at shutter speeds varying from 1/2000 to 1/4000, in daylight with no supplementary lighting. From that 60p video, smooth slow motion equivalent to filming at 300~600fps was "faked".

Edit: Shutter speeds that fast are not necessary for motion interpolation anyway, except to minimise blur. As you yourself said, "There is a bit of lag behind the ball movement due to the low light level with the roof closed", so the same effect in low light / with longer shutter speeds happens regardless of whether the action is captured by high frame rate cameras or simulated by motion interpolation.

Edited by Malich, 09 July 2012 - 11:28 AM.

[Craig]

alanh, on 09 July 2012 - 02:44 AM, said:

When the progressive image is reread to convert the image to interlaced, both fields represent the same point in time. As a result the display on a flat screen TV is progressive at a frame rate of a multiple of 25 frame/s. Each frame may be displayed twice or 4 times as identical frames or with motion compensation.
AlanH

When talking about 50i and 50p, they're actually both the same IN TERMS OF MOTION.  They both have 50 unique images per second.  It's just that the progressive system has full vertical resolution with each frame, as opposed to 'half' vertical resolution with each video field in interlace. (ODD lines in field 1, EVEN lines in field 2).  With sport or other native 50i material, slow down an interlaced video signal, and you'll see each video field is a unique picture, with elements [like a ball] in different positions between fields.  So to correctly de-interlace 50i, you need to go to 50p.

With film it's different, as each pair of fields IS from the SAME frame of film, so you can weave the two video fields together to form one Progressive frame at only 25p.  Both fields are sourced from the same point in time (ie the one film frame).

25 Frame progressive video is really 'half frame rate', which gives the juddery vision and motion artefacts associated with the 'film effect'.  Half the motion/temporal 'smoothness' of true 50i and 50p.

[DrP]

Whatever you do, don't mention to alanh that modern cameras do not actually take interlaced pictures.  They take sequential full frame pictures and then internally scale and interlace them to produce interlaced content.

[MLXXX]

alanh, on 09 July 2012 - 02:44 AM, said:

Hi, I am watching the tennis and have noticed the following;

A more reliable approach is to capture the broadcast stream and inspect individual frames in it to see whether the odd and even fields extracted from the camera source are offset in capture time rather than being part of a single (segmented) progressive frame.

It is remarkable what adaptive de-interlacing algorithms in home TV sets are capable of these days.

It is implausible that a 25i transmission of live sport would be at 25 SF. Tennis is too rapid a sport to be captured adequately at 25p. I suggest, alanh, that you analyse the transport stream and see whether there is any comb pattern in the odd and even fields, relative to each other, in non-static scenes.

Edited by MLXXX, 09 July 2012 - 04:49 PM.

[alanh]

I don't have to because the camera original is progressively scanned at 50 frame/s just like they did at the World Cup Soccer in South Africa.
There is no odd and even fields coming from the camera.

I don't need to examine any data streams, because I have seen enough interlaced images to tell the difference on the screen. They are not using the even lines from one progressive frame and the odd progressive frames otherwise the usual interlaced image problems reappear.

If you look at the movement of the ball there was no jerkyness.

Use your eyes not some theory you have gained from the internet.

AlanH

[MLXXX]

** SIGH **   Alanh, have you forgotten the interlaced frames I extracted from the 2010 World Cup soccer for your benefit? These clearly showed timing displacement between the Odd and Even fields. You thought those were progressive too.

I suggest it is not good enough for you to glance at a modern TV screen and because it lacks visible interlacing artifacts shout "Eureka, it's progressive segmented frame!" and create a thread on this forum.  Why you continue to believe that modern adaptive deinterlacers don't work effectively, mystifies me.

I watched the Wimbledon gentlemen's final last night on channel 7 in Brisbane (was very very pleased for Federer  though it was a pity the Scot missed out yet again),  but for my eyes the coverage had the usual smoothness of 25i [I have the motion interpolation feature on my plasma TV permanently switched off]. It did not have the telltale cinema look of 25 SF.

Edited by MLXXX, 09 July 2012 - 07:11 PM.

[alanh]

MLXXX
Please explain why there is no interlace aliases on the images on the tennis.

The objective is to get the best quality pictures on the screen.

AlanH

[nbound]

Before this degenerates into another of the usual discussions... Anyone got some caps of Wimbledon that can settle it?

[MLXXX]

alanh, on 09 July 2012 - 07:31 PM, said:

The objective is to get the best quality pictures on the screen.

Alanh, as explained to you umpteen times in the past, the better quality for sport is 50 motion updates a second, rather than 25. If the better quality for tennis were 25p then camera sources for tennis would simply be adjusted for 25p.  Sometimes television sports interviews are done at 25p; but I've never -- to my knowledge -- seen sports action done at 25p with live TV coverage.

Although I recorded some of last night's final it was to a dedicated USB drive attached to the TV and there is copy protection stopping any transfer to a pc.

Perhaps someone else can assist alanh in this subject area by extracting a raw MPEG-2 frame from an accessible recording of the recent Wimbledon DVB-t broadcasting in Australia. Care needs to be taken as a lot of software will automatically deinterlace. We need a raw frame or two, that includes a moving object. Left to right movement produces an obvious mice teeth pattern in the raw frame. (Modern deinterlacing algorithms recognise this pattern and synthesize a smooth picture.)

[alanh]

MLXXX,
if you wish to think of our SD system to contain 288 lines 50 field/s there is a problem. The interlace and rapid movement causes the edges of rapidly moving objects to become jagged. There was none of that.

I have not disputed that a TV system which is 50p from camera to display is better, however this is not currently possible without DVB-T2 and MPEG-4 compression.

In the meantime progressively scanned images in the camera are better provided both fields come from the one frame.

You are yet to answer my question.

AlanH

[MLXXX]

It has been answered multiple times before, alanh. Modern deinterlacers are able to extract, and where necessary synthesize, a remarkably alias-free, and detailed, picture from interlaced input.

In terms of a 720x576 pixel SD picture, each horizontal line receives a fully faithful update each 1/25th second, odd lines receiving their update 1/50th second out of phase with the even lines.

[The second paragraph is true to a first approximation. There are certain special techniques available with encoding codecs for interlaced content for frames that contain static material.]

Edited by MLXXX, 09 July 2012 - 10:55 PM.

[alanh]

MLXXX,
So you are proposing to convert a 50 frame/s progressive image in the camera so that they can produce smooth slow mo replays, and then use only the odd lines in the first frame and then the even lines to give the same effect as a 25 frame/s interlaced image. It always shows jagged edges on near vertical lines in the image. These jaggies move as the camera moves. I have seen this millions of times. If the original image uses each alternate frame of the 50 frame/a progressive image each transmitted frame represents one point in time. This will produce smooth edges. If the motion compensation is applied to this image it means that the additional frames shown will have smooth edges and the motion compensation will reduce the jerkyness. I am not using one of these de-interlacers and I don't need it because there is no jagged edges in the picture.

In addition a 50 frame/s full HD camera is exposed for 14 ms in every 20 ms. A 25 frame/s full HD progressive camera is exposed for 27.9 ms every 40 ms. This assumes no shutter in the camera. The reduction in exposure time minimises the blurring in the camera.

The telecast does not contain any whip pans which is more likely to causy jerky movement not only due to a low frame rate but also because of insufficient data rate.

This discussion has already been conducted on this forum in great detail in the past so I have no intention of repeating it.

So despite what you think the tennis was received as a 25 frame/s signal but sent in interlaced order. It is no where as bad as you claim.

This discussion is going the same way as the continuous full carrier in AM, you would not accept that, so I assume you won't accept this.The movie industry has been using a frame rate of 24 frame/s for a century and continues to do so. The movie screen is considerably larger than most TVs and this will make the jerkyness more obvious.



AlanH

[Malich]

alanh, on 09 July 2012 - 10:03 PM, said:

I have not disputed that a TV system which is 50p from camera to display is better, however this is not currently possible without DVB-T2 and MPEG-4 compression.

Alan, once again you've over-reached reality in an attempt to push your particular hobbyhorse re: the magical properties of DVB-T2 & MPEG-4.

It's statements like this that make people think you speak nothing but ****. In the future, please stop and think before you make similarly stupid statements.

[MLXXX]

Alanh, your inability to accept the easily measured fact that at the bottom of a 50Hz 99.9% sine wave modulation the RF energy output by the AM transmitter remains low for many cycles of the medium wave carrier [and in modern transmitter design that the input power at that point in the modulation is quite low] is your own failure of understanding. I don't think it assists your case to raise that matter again here.

____________

Yes years ago a lot of us noticed "jaggies".  But your continual references to jaggies in 2012 in my opinion rejects the capacity of modern day deinterlacing processing to synthesize smooth lines.

In typical alanh fashion, you provide no raw image capture from an MPEG stream to support your contention each pair of fields was from one capture time (i.e. from one progressive frame) in 7's recent terrestrial broadcasting of Wimbledon. You provide opinion and assumption, but no evidence.

My recording made last night on a dedicated USB hard drive reveals fluid motion, consistent with normal 25i. If I could extract frames and upload them I would.  But then it is not I who started this thread...


To illustrate what can be done if making an effort to investigate, I remind you of an old post of mine, following my analysis of the fields making up frames from the 2010 FIFA Word Cup:

MLXXX, on 17 July 2010 - 11:17 PM, said:

Alanh,
as they say, a picture speaks a thousand words!

I recorded the opening match of the world cup soccer in the evening of 11 June 2010 Eastern Australian time. The recording is currently taking up 14.5GB on a hard drive. I extracted four consecutive frames using VideoRedo TV suite H264. I then cropped out a 300x300 pixel section to highlight a moving object. It appears to be a microphone and/or camera suspended by a cable.

The software enables a capture of any frame in the source. If the source is 50p, there are 50 selectable frames per second. If the source is 50i, there are 25 selectable frames per second. The software options include deinterlacing set to on or off.

The frames are as follows:

To illustrate interlacing.
World Cup opening match 11/6/2010, Johannesburg, South Africa v Mexico
As the teams were entering playing field, 16 min 14 sec before start of play.
Captured Brisbane, 3D trial, MPEG-4 AVC, 1920x1080i50

Full Frame 00 - deinterlacing off: Saved as jpeg

Full Frame 00 - deinterlacing on: Saved as jpeg

Full Frame 03 - deinterlacing off: Saved as jpeg

Full Frame 03 - deinterlacing on: Saved as jpeg

Uncompressed montage of 300x300 crops from frames 00, 01, 02, 03 with deinterlacing off or on (beginning 150 pixels from top of frame, 500 pixels from left side of frame):

You will note that the frames with deinterlacing off are slightly sharper, but reveal a comb effect with the moving object. The teeth of the comb extend one pixel vertically and about 6 pixels across. There has been no rescaling of the original 1920x1080 side by side frame. The comb teeth would look twice as long if rescaled by a factor of 2 to give the correct aspect ratio.

Edited by MLXXX, 10 July 2012 - 04:57 AM.

[DrP]

The problem here is that you are all wrong and alanh is right.  Forget that it's been demonstrated to alanh that aliasing / staircasing is aparent in progressive video.  Forget that it's been demonstrated to alanh that modern cameras do not actually take interlaced pictures.  Forget that it's been demonstrated that an interlaced image does not necessarily display aliasing / staircasing when correctly filtered.  Forget it all.  Just accept that you are wrong and alanh is right.  That's all there is to it.

[Malich]

Maybe he could arrange some sort of media release, or get the minister to make a statement to that effect?

[alanh]

MLXXX,
If you wish to prove your case on the tennis then produce the images. No interlaced scanning was used in the transmission of the tennis pictures. Only transmission in the interlaced order.
I did not make the comments on the Soccer about aliases as I have about the tennis. It could have been done a different way!'

As for AM and FM you choose to ignore all of the RF engineers at your own peril. You have never used the equipment for various levels of modulation I have including using Bessel function figures to produce no carrier in FM. This is a standard technique to prove the level of deviation.

The collective others, the high data rate produced by 1920 x 1080 x 50p requires improve compression which is already being used for Australian 3-D transmissions, however this is insufficient particularly 3-D 1920 x 1080 x 50p DVB-T2 is required.

No one wants to see the result of a huge compression rate using MPEG-2. People complain enough now with much lower compression rates.

If you all want to stay in the past it is your progative.

AlanH

Edited by alanh, 10 July 2012 - 12:49 PM.

[DrP]

The moon is not made of cheese.  Cheese has quite a strong odour and since you can not smell the moon it must not be made of cheese.

cf Witches are made of wood.

Edited by DrP, 10 July 2012 - 01:05 PM.

[MLXXX]

alanh, on 10 July 2012 - 12:48 PM, said:

MLXXX,
If you wish to prove your case on the tennis then produce the images. No interlaced scanning was used in the transmission of the tennis pictures. Only transmission in the interlaced order.

Just to be clear, alanh, are you confirming it is your belief that the recent Wimbledon tennis was broadcast in your Perth DVB-t region as 25 segmented frames per second; that is to say with both fields of an interlaced frame being extracted from one and the same source progressive frame; equivalent to a 25p transmission?

(I will not restart the debate on the modulated FM carrier frequency [that modern receiver design continually tracks with a PLL]. )

[alanh]

I wait to see if the Olympic games have the same clarity of line markers in the athletics. As far as the signals for the tennis go the transmissions to Brisbane are on the same network and use an identical signal.

I will not restart FM again however it has nothing to do with receiver design at all. It is the fundamental characteristics of FM. Spectrum Analysers to not demodulate the signal to display the power of each frequency.

AlanH

[MLXXX]

Ok, so alanh confirms his belief that the broadcast of the tennis he saw in Perth (I presume on the 7 network) was 25 segmented frames per second, equivalent to 25p.

My own observation was that the tennis I saw from the 7 network in Brisbane had the usual motion fluidity associated with true 25i, so I would think the fields would have been from different capture times (probably 1/50th second apart, the broadcast originating in Britain). I'd mention that the picture detail was on the poor side, even for SD, presumably because of a fairly low MPEG-2 bitrate; but there were not the aliases I have seen in the past in some tennis coverage, e.g. of the pattern of the tennis net.

I think that's a stalemate. Alanh has made a surprising claim, without any proof. I have queried that surprising claim, but I have no MPEG-2 transport stream frames that I can access for analysis.

[ckent]

In other "news", I've now started wondering how long ago SBS had actually started getting Tour de France in HD … because I remember in the bad old days of 576p50, the live pictures were actually better quality than the replays, and the interlacing effects were not evident on the live pictures.  At the time they didn't have the upscaling effects of 576i50 -> 576p50.

It was a marginal improvement but I think it was happening long before SBS ever announced they were broadcasting HD content.

CK.