Why interlaced HDTV?

In article , JC
writes

I'm not sure how simple I have to make this but with a progressive
standard it's easy to refresh the screen as many or as few times as
you like without introducing additional artifacts.

I am not sure how simple I need to make this, but with *any* video
standard it is *impossible* to refresh the screen multiple times for
each field (i) or frame (p) *without* introducing additional artefacts!

With interlace, any
change from the native refresh leads to artifacts. A progressive
broadcast on a modern pixel based screen can lead to the highest field
rate with no flicker or other artifacts.

Rubbish - see the other post for an explanation of this.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's ****ed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

In article , JC
writes

Imagine taking two separate photographs at half resolution, a short
time apart of something that's moving and then trying to merge them to
get double resolution. The moving object will be in a different
position on the second field to the first. Now imagine rapidly
flicking between these two images or showing them merged, before
moving on to the next two.

With a progressive system you just reshow the same FRAME multiple
times or in the case of a memory type display such as LCD, just change
the pixels as required.

Have you ever actually tried that and *looked* at the result? I doubt
it, because if you did you would notice that it produces just the same
motion deficiency that you are being so critical of interlace to
progressive conversion of.

If you show that same progressive frame several times then motion
produces the perception of several images of the object in the scene
that is moving. For example, 50Hz-720p on a 100Hz display with the
simple frame repetition that you describe, will produce double images of
moving objects. This is well known and was a major limitation of early
and current cheap 100Hz sets (although you, no doubt, attribute it to
the interlaced source) which is why better 100Hz sets use motion
compensation algorithms to fix it.

If you want to see why this happens, try drawing a graph of screen
position along the direction of motion versus time. An object moving
across the field of view at a constant rate will be represented by a
straight line on this plot. A 50Hz camera sees the position of the
object every 20mS on the time axis - so mark dots along the line
representing the position that the camera sees. Now redraw those dots
in the same positions some time later, representing the object position
on the display. If your display is 50Hz then each camera dot produces a
corresponding display dot - and your brain joins the dots up to create
what appears to be continuous motion again.

If you display repeated frames at 100Hz, then each dot from the camera
is displayed twice in the same position but 10mS apart on the time axis.
Now natural real life objects don't exhibit regular stop-go motion at
10mS intervals, they move and your eye tracks the motion to keep the
object of interest on the fovea. Your eye muscles simply can't keep up
with that stop-go motion - and your brain tells it that it doesn't need
to. You naturally misinterpret this display as *two* objects moving
smoothly across the display, one behind the other by a distance
corresponding to the 10mS time gap between display repeats - the same
effect as if you had an interlace image reconstructed in a frame store
and displayed progressively.

You can use the same approach to understand that motion blur on LCDs is
fundamental to the continuous display, not just a consequence of LCD
time constant. Just replicate each camera sample with a line extending
the full 20mS on the display. Now look at how your eye interprets the
continuous motion - as your eye tracks the motion of the object across
the screen all that resolution that you are so keen to capture at 50Hz
on the progressive camera is lost in a blur corresponding to the
distance the object moves in 20mS! That is why "memory type displays"
as you call them are just garbage for reproduction of motion - they
trade motion for resolution all the time.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's ****ed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

In article , JC
writes

My point was that with a progressive standard it's easier to reshow
the same frame multiple times for higher refresh rates without motion
or other artifacts

No it isn't - whether on interlace or on progressive it makes no
difference. Multiple displays of the same information in the same
position *always* introduces motion artefacts. It does it on interlaced
material and it does it on progressive material - the artefact is a
consequence of corrupting the temporal flow of information by displaying
each temporal sample multiple times.

--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's ****ed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

In article , JC
writes

720p will only have the same apparent temporal resolution as 1080i if
the 1080i source and display are none progressive. The 1080i field
resolution will only be 540 lines though compared to 720 with 720p.
1080i provides half the FIELD rate of 720p.

No it doesn't. It has half the *FRAME* rate - and that doesn't
correspond to half the temporal resolution since both systems sample the
scene at 50Hz. The advantage of 50Hz progressive over 50Hz interlace is
that the progressive system is capable of capturing information which
requires *both* the full spatial resolution *and* the full temporal
resolution *simultaneously*. At the very most, that can only represent
25% of the total information, since both systems have the same
resolution on static images and both have the same temporal resolution
of information up to half the spatial bandwidth. In practice, it is a
lot less than this 25% upper limit.

The above assumes a comparison of interlace and progressive systems of
the same line standard. However that isn't the topic under debate - we
are comparing 720p to 1080i, the latter having 2.25x as much static
spatial resolution as the former. Losing that upper limit of 25% of the
entire information of the 1080p signal to achieve 1080i due to motion is
still 70% more motion information than is possible to reproduce on 720p.

So all those sports stations that have adopted 720p for it's better
motion rendering have got things the wrong way round?

No, but your reason why they have adopted it is wrong. Progressive
doesn't provide better motion rendering over interlace. Progressive
doesn't provide better slow motion rendering over interlace either.
However, progressive does provide better and simpler electronic zoom
than interlace.

What crap - interlace sources have been the format of choice for
movement for over half a century.

Because it's a convenient form of lossey compression in the analogue
world. It's the technology of the 1930's.

Being old doesn't establish inferiority. On the contrary, experience
shows that there is a higher likelihood that youth lacks superiority in
almost all cases.

We've moved on from there.

"Moving on" is not progress when you don't know if you are moving in the
right direction!

Don't give up on "the old ways of doing things" until you understand
what they offered - your posts suggest that you are in favour of
progressive because it is new, not because you can demonstrate any
advantage it has over the "old fashioned" approach. I have yet to see
you mention the one clear advantage that progressive does offer over
interlace!

1080p production converts nicely to all current and proposed
interlaced and progressive standards. Choosing a progressive format
for HD broadcast is a natural progression of this. I don't follow how
this is in any way less backwards compatible than any other new HD
standard.

Because 1080p to 1080i does not produce a loss in *either* temporal or
spatial resolution. It results in a loss in the overlap between
temporal and spatial resolution - where *both* the full temporal and
spatial resolution are required simultaneously. 1080p to 720p results
in a spatial resolution loss, which is greater than simply the ratio of
the line standards - even with sinc interpolation to get the 720p
samples.

But progressive is NOT a move forward with technology in itself.
particularly when the option is between 720p and 1080i. Both formats
provide similar vertical resolution in the presence of motion, but the
interlaced option provides much higher horizontal resolution in all

You keep repeating this but it doesn't make it true.

1080i certainly does have much higher horizontal resolution than 720p -
that doesn't need repeated to make it true, but it may do to make you
recognise it.

720p has twice
the number of FRAMES as 1080i.

But that is NOT the same as having twice the temporal resolution, which
is what you keep (erroneously!) claiming! "You keep repeating this but
that doesn't make it true."

25 frames displayed as 50 half fields. In future both camera and
display will be natively progressive. This will show the true limits
of a 25 FRAME interlaced system.

Nobody is arguing that when the same line standard is used a progressive
system is superior to an interlaced system. The question is whether
720p is superior to 1080i - and the fact that one of those systems is
progressive and the other interlaced is not sufficient to establish its
superiority.

1080p production allows easy conversion to all lower standards. To fix
on 1080i as a broadcast standard would be to immediately throw away
half the information

You keep saying that, but it doesn't make it true! Progressive to
interlace conversion only throws half the information away when the
scene contains maximum entropy. Since the maximum entropy condition
means minimum useful information content, it is obvious that no real
video images fall into that category. Furthermore, as entropy
increases, either the bitrate of the transmitted signal must increase or
the information loss in the codec increases. Hence the main advantage
you claim for progressive is the first thing that gets lost in real
world digital transmissions of the progressive signal!

while introducing significant display artifacts.

As already established, it doesn't introduce any display artefacts if
done correctly.

720p may not have the static resolution of 1080i but wipes the floor
with it in all other respects.

You saying that doesn't make it true. Next time you look at a video
image make an estimate of how much of the scene is differentially moving
at any time. Panning makes no difference in perceived resolution
whether interlaced or progressive.


I really agree that we should be looking to 1080p but as production is
moving in that area anyway it would appear logical to adopt the
broadcast standard that's the closest technical match.

Which is 1080i - simply drop every other line in alternate fields: no
interpolative downsampling (with the consequential loss of resolution
inherent in all interpolation techniques) required.

...and then try and reconstruct them in the displays memory from the
two time different half fields

You really didn't follow the earlier discussion did you? Reconstructing
an interlace signal in memory for progressive display is THE WRONG THING
TO DO!! You don't *NEED* to do the wrong thing!!

1080i is 25 fields per second. I do not believe this to be
enough for a modern broadcast system.

You have not provided any evidence in support of this belief, whilst
over a century of the motion picture industry indicates that the
opposite is true. Even Imax runs at 24fps - and neither of the HD
systems under discussion come close to meeting that in performance.
ImaxHD runs at 48fps, but that is even further out of the reach of HDTV!


With 720p you have a down sampling of the source to from 1080 to 720
at the broadcaster

Losing vertical resolution in the process - unlike sampling a 1080i
field from a 1080p frame, getting a 720p frame is not an integer spatial
division!

So how does a 540 line half field system beat a 720 line full field
system when the two half fields are taken effectively taken at
different points in time, resulting in major interlace artifacts when
combined to create 1080 line full frame?

Quite simple really. Each of the 540 line fields is an integer division
of the original 1080p frames - no interpolation. 720p requires a line
division of 1.5 in every frame - can only be achieved with
interpolation, resulting in a final resolution which is inferior to what
720p native signals can achieve.

As already explained, your infatuation with recombining the interlaced
signals, which have different time data, is an unnecessary and erroneous
step. If you display it *as* two interlaced fields (which can be
achieved even on a progressive screen) the interlace artefacts you are
prattling on about simply don't arise!

Half the frames are removed and those that remain are split in to two
time separated fields. This not only throws away half the available
information

NO IT DOESN'T!!
You clearly have trouble understanding the difference between
information contained in the image and available information space which
can only be utilised by a maximum entropy image. Useful video signals
are not maximum entropy! The whole raison d'etre *for* interlace is
that it *doesn't* throw half the available information away, because
much of the video signal is redundant information.

but introduces significant interlace artifacts on the what
remains.

What are these interlace artifacts you are wittering on about? You have
only described artefacts that are a consequence of badly displaying an
interlace image on a progressive display - that is not an interlace
artefact, it is a stupidity artefact!

As i've said. Interlace was a 1930's

As you will one day discover, being old does not make it inferior!


You can't create information that's not there.

There is no need to create information that is not there. That is how
interlace systems have been successfully displayed for half a century
without any of the artefacts you are wittering about.

If the scene is static they'll have the same information. Where there
is the slightest movement they will differ. I regularly see this on
material that has been captured from conventional TV.

Yes, if you try to capture a still frame or reconstruct a series of
still frames to display progressively then you will get motion
artefacts. You should, by now, realise that this is not the correct
thing to even attempt. The interlaced fields represent different time
samples of the image and *MUST* be displayed in the correct time
sequence, just as progressive frames must be displayed in the correct
time sequence. Any deviation from that will result in the artefacts
that you consider to be restricted only to interlace.

When the interlaced material is converted back to progressive 1080i,

Why would interlace material require conversion to 1080i, it already is
1080i. I assume you mean 1080p, but why would you want to do this when
you know it combines two different time samples. You have to display
the time samples independently, trying to combine them is the cause of
the artefacts that you blame on the interlace format!

effectively contains half the temporal resolution of 720p. Yes in the
original 1080i signal there are 50 time separate half fields,
bloody hell, where did the other half fields go. They are full fields,
each being half of a frame from different time samples.

but when
merged to create the 1080 line picture, this is lost and causes a
reduction of visual resolution.

It isn't lost - it is misused, and being misused it is hardly surprising
that it creates visual errors. Display them as interlaced fields
retaining the temporal reference. This can be done on a progressive
display.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's ****ed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

In article , Jc wrote:
That's right. That's the way its been for the past 70 years and despite
the fact that you can't easily derive a full resolution still image from
it (something television was never designed for), it shows movement far
more naturally than a full resolution non-interlaced signal using the
same bandwidth.

But deriving a full resolution image is what you need to do on any
large progressive display. The common interlaced standards have half
the frame rate of the progressive standards, this worse for movement
in this case.

They may have half the *frame* rate, but if they have 50Hz *field* rate,
this should be enough to give the impression of the equivalent frame rate
but without requiring twice the bandwidth to transmit all the lines 50 times
every second. Just update the lines that have changed every field, and the
ones in between them every other field, and the eye averages them out and
gives the impression of full vertical resolution *and* 50Hz temporal
resolution (twice as good as cine film), as it always has done.

The only occasion where this fails is on edges of moving objects, which will
be slightly blurred because they don't have the full vertical resolution,
but since moving objects look blurred in real life anyway, this appears
perfectly natural.

If your display is incapable of updating its lines in this way, then it
hasn't been properly designed as a contender for replacement of the humble
CRT, which has always been able to update lines in any manner you please.

Rod.

In article , Kennedy McEwen wrote:
720p will only have the same apparent temporal resolution as 1080i if
the 1080i source and display are none progressive. The 1080i field
resolution will only be 540 lines though compared to 720 with 720p.
1080i provides half the FIELD rate of 720p.

No it doesn't. It has half the *FRAME* rate - and that doesn't
correspond to half the temporal resolution since both systems sample the
scene at 50Hz. The advantage of 50Hz progressive over 50Hz interlace is
that the progressive system is capable of capturing information which
requires *both* the full spatial resolution *and* the full temporal
resolution *simultaneously*.

The disadvantage is of course that, all other things being equal, a 50Hz
"progressive" system scanning *all* the lines 50 times per second would
output twice as many lines per second as an interlaced system scanning only
*half* the lines 50 times per second. In other words, it would require
twice the bandwidth to transmit it and twice the storage capacity to record
it. You get what you pay for. It would be easier to derive better printouts
of still images from such a system (if you think that's what television is
for), but it probably wouldn't look noticeably different on moving ones.

Rod.

In article , Jc wrote:
Maybe we're talking at cross-purposes, but I feel I'm struggling to
simplify something too. *Flicker* and *intermittency* (or "jerkiness" of
moving objects) are not the same thing. Flicker in a display can be
absolutely eliminated, but the rate at which pictures are updated cannot
be increased beyond what was properly sampled by the camera, and
maintained throughout the system.

I understand the broadcast world be moving to progressive production
technology. That's ideally 1080p50 with 50 full frames per second.

I hope you're right. I won't hold my breath waiting for it, but if it really
happens it will be an improvement on what we have now.

Rod.

In article
. co.uk.invalid, Alan
Pemberton wrote:
So, certainly, when all displays are linear
we'll still use a non-linear transmission format, and we'll
still call it gamma, but it won't be an "odd situation" in that
there will still be a sound technical reason for doing so.

Perhaps there will even be an industry standard value for it.

Might I suggest that the one we've already got would be a good choice?

I know it's different from the "official" figure of 2.8+/-0.3 specified
for PAL, but all the test charts I've seen seem to be designed for 2.2.

Rod.

In article
. co.uk.invalid, Alan
Pemberton writes
wrote:

So, certainly, when all displays are linear
we'll still use a non-linear transmission format, and we'll
still call it gamma, but it won't be an "odd situation" in that
there will still be a sound technical reason for doing so.

Perhaps there will even be an industry standard value for it.

Only if there is an industry standard contrast ratio. :-)
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's ****ed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

In article , Roderick
Stewart writes
In article , Kennedy McEwen wrote:
720p will only have the same apparent temporal resolution as 1080i if
the 1080i source and display are none progressive. The 1080i field
resolution will only be 540 lines though compared to 720 with 720p.
1080i provides half the FIELD rate of 720p.

No it doesn't. It has half the *FRAME* rate - and that doesn't
correspond to half the temporal resolution since both systems sample the
scene at 50Hz. The advantage of 50Hz progressive over 50Hz interlace is
that the progressive system is capable of capturing information which
requires *both* the full spatial resolution *and* the full temporal
resolution *simultaneously*.

The disadvantage is of course that, all other things being equal, a 50Hz
"progressive" system scanning *all* the lines 50 times per second would
output twice as many lines per second as an interlaced system scanning only
*half* the lines 50 times per second. In other words, it would require
twice the bandwidth to transmit it and twice the storage capacity to record
it.

Precisely - you need twice the bandwidth and storage but only get a much
smaller improvement in information content, because most of the
information in a typical video image (ie. not a synthetic high entropy
test sequence) is redundant.

You get what you pay for.

Yup, twice as much cost in almost every sense for an incremental
advantage, assuming the same line standard.

It would be easier to derive better printouts
of still images from such a system (if you think that's what television is
for),

But it isn't really, is it (unless you are the Met Police trying to
recover legislative quality images from a camera on a tube train)?

but it probably wouldn't look noticeably different on moving ones.

Oh there is no question that progressive looks better that interlace on
moving images - just not sufficiently better to justify the trade down
from a 1080 to a 720 line standard.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's ****ed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)