Predicting the Future, Interlace or Deinterlace?

[ses]

The question is, will players continue to have deinterlacing functions in the next 10 years or so? Or will everything gravitate towards progresive and suddenly in 15 years, I can't find a player to play the interlaced DVD home videos that I have filmed of my family, or all the interlaced VHS recordings that I have converted to DVD.
Before I really get into this project and make the wrong choice, what does the future hold?
Should I deinterlace my AVI's or leave them as they are?

[communist]

I can't find a player to play the interlaced DVD home videos that I have filmed of my family, or all the interlaced VHS recordings that I have converted to DVD.
I dont get it. Do you mean hardware or software DVD players cant play your interlaced material? Interlacing will be around for much longer because pretty much all the content produced for TV since the beginning is interlaced and will continue to be for quite some time.

[ses]

What I mean is the following: New TV's are progressive scan, they are not built based on interlace broadcasting.
Right now, a standard DVD player that you would buy for you living room will deinterlace interlaced content, but will they continue to do that down the line, or will interlaced material be one day, unplayable?

[buzzqw]

a good question

rephrased: would my interlaced dv, encoded as interlaced mpeg played flawlessy in new/future dvd player ?

BHH

[Blue_MiSfit]

dvd player


If it's really a DVD player, then yes it most certainly will. Interlacing is a part of the DVD spec, so if it's a DVD player it HAS to support interlaced video correctly!

Interlacing is here to stay, 1080i is a very popular HD format, and 480i / 576i are still dominant!

It's funny to me, but I see no way of moving away from this trend.

~MiSfit

[RickA]

A good question ses, but as the others have said I do not think you need to be as concerned on the viability of the format dissappearing as you are. Odds are in ten years time you may be thinking about re-burning your discs to 'refresh' their shelf life. Or possibly even converting them to some other new fangled means of media storage / playback devices by then. We could all be watching 3D holographic projections, who knows? :-]

Cheers,
R

[Mug Funky]

though deinterlacing is popular today (i call it a fad, but others in the industry have different ideas... to each their own), there's a massive legacy of both interlaced content and equipment that works with interlaced video.

and though increasingly content is being shot in progressive modes, there'll always be live broadcasts which will pretty much always be interlaced (even lay people will notice their sports being jerky as hell if they flicked the football onto progressive).

also, every mainstream video standard except the very much obsolete VCD fully supports interlace.

i think if in the future interlace is abandoned in some TVs, it'll simply be to save the cost of 1 chip, rather than any form of "progress".

[wonkey_monkey]

I really wish they'd taken the opportunity to do away with interlacing when HD came along. 720p50/60 (with 1080p24/25/30 for films) ought to be enough for anybody ;) Especially now everyone's getting plasma/LCD TVs which have to do heavy processing just to display these pictures, the first step of which is probably to deinterlace... *roll eyes*

David

[Ranguvar]

Most HD-DVD and Blu-ray films are progressive, but for (HD)TV, I see interlacing staying for a while.

But yeah, to be a DVD player it has to support interlacing.

And, I agree with david. Deinterlacing is a pain for us nerds, and I'm sure the lower-end TVs use a rather quick-and-sloppy deinterlacing.

[miamicanes]

There's a good reason why camcorders that can do "1080i60" can/will be able to be bought for only slightly more than the cost of a GOOD 480p60 camcorder -- mechanically and electronically, they're almost identical inside.

I'm simplifying a bit, but most consumer-grade camcorders have CCDs with approximately twice as many green elements as blue or red elements. Normally, clusters of 2x2 (maybe more, depending on the physical resolution of the CCD and the desired output resolution & aspect ratio) sensors, consisting of 2 green, 1 red, and 1 blue, would be read to generate each pixel of a 480p60 or 540p50 frame. For the sake of argument, pretend the CCD really has a physical resolution of ~1440 x 1080 elements (720 green elements per "row", alternating with a blue sensors on odd lines, and red sensors on even lines), and normally , and supports both 50 & 60hz progressive. In normal use, it uses a 4x4 matrix consisting of two green (one per line), one red, and one blue to discern a single pixel. But I think you can already see the marketing department's opportunity for mischief...

Suppose that instead of demosaicing the 4x4 grid to a single pixel, we rationalize that MPEG normally throws away up to 3/4 the color data anyway. Hmmm... 4x4 matrix, two green, one red, one blue. Count each sensor as a horizontal-resolution pixel, count each row as a "scanline", and suddenly the marketing department has just turned a high-quality 540p60 prosumer camcorder into a ghetto-fabulous "1080i60 HDcam". The quality will be nowhere near as good as "real" 1080i60, but they'll be cheap, and clueless consumers will ask why they should pay $2400 for a Panasonic camcorder that "only" does 720p60 when they can get a HanTeq HDcam at WalMart that does "1080i" for $499. The 720p60 camera will blow the quasi-1080 camera out of the water (because it will do 720p60 the way a formerly high-end camcorder did 480p60 -- demosaicing a 4x4 grid for each pixel, and will have a sufficiently-beefy storage system to allow a much higher bitrate than the cheap 1080i cameras), but in the great race to the bottom, the ersatz-1080i cameras will "win" at Wal Mart.

[NerdWithNoLife]

I have a question which is kind of on topic with this old thread. Hey, it's a new year right? So looking to the future is appropriate.

Is there any reason plasma TV's are incapable of displaying interlacing? I know they have a better refresh rate than LCD's, so what if... an interlaced picture was shown every 60th of a second? Not like this:
You ask why progressive displays have to deinterlace and can't just render the interlaced fields one by one skipping the odd and even lines alternately. I guess it is possible to build displays that work that way but the displays available today don't. Also since the persistence of the light emitting material in plasma displays is much shorter then in CRTs the result will appear flickering and inconvenient to watch.
But rather like DoubleWeave on AviSynth. So every 60th of a second instead of showing a bobbed field, or only half of the lines, it shows the previous field mixed with the new field - in a new frame, redrawing the entire frame. The drawback I can see is this would involve a bit of interpolation as the scan lines are meant to overlap a bit. But that interpolation should be totally possible. A plasma monitor should be able to keep up in refresh time, right? One of the things that really bothers me about LCD's is refresh time. Right now I have an HD CRT, and I'm waiting for something to come along that's an acceptable replacement to me.

I wanted to ask here, because there are other forums where people don't truly understand interlacing, or at least not on the level that the folks here do.

[LoRd_MuldeR]

Because LCD's work completely different than CRT's. The CRT screen has an electron beam, that moves over the screen and "draws" line after line. At the end of the screen the beam start over from the beginning. But on each run the beam will skip every second line. In one run it "draws" only the odd lines, in the next run only the even lines, then the odd lines again and so on. Hence the interlacing! If the human eye wasn't that slow, you wouldn't even see a picture on a CRT screen, but a colored dot that is moving very fast. The eye is too slow to see that, but the picture doesn't look perfectly stable. Hence the flickering! Now a LCD screen works completely different. There is a constant backlight, which is sending out light all the time. The liquid crystals in the LCD screen work like valves. The let more or less light slip through from the backlight for each individual pixel. Unfortunately the liquid crystals are relatively slow! For example a pixel can't move from "black" to "white" and back to "black" (or vice versa) very fast. That's also why fast motion tends to smear on a LCD screen! So I guess your idea wouldn't work because of the slowness of the LCD screens. Showing a line for exactly 1/60 of a second and then "deactivate" that line for the next 1/60 second just isn't possible. Also I think on a CRT screen there is some "glow" effect, caused by the phosphor. When the electron beam has passed a certain line, it will dim slowly by-and-by. That's why interlacing looks "okay" on a CRT screen. But on a LCD that kind of effect is missing, which is why you can't have the "CRT style" look. At least that's why I think why your proposed method isn't used in reality ...

(Just one thought: Maybe a plasma screen would be more suitable for that method than a LCD screen ???)

[NerdWithNoLife]

But my question was regarding plasma, not LCD. You are describing the reason LCD is totally unappealing to me. But if our eyes were fast enough, we wouldn't just see a dot - because a CRT has a phosphorescent screen that holds the brightness long enough for the other lines to be drawn before the beam comes back to that spot. My TV has three electron guns, one for each third (top middle bottom).

Though the picture is drawn by the beam in the fashion you describe, the modern video camera has a CCD which images the entire field at once, in the same instant. So even if the beam takes a 60th of a second to scan each line of the field, each of those lines are supposed to be from the same instant in time. Which is why plasma could possibly be even more accurate, in that respect. Or if plasma isn't up to the task, perhaps SED will be (if it ever does come to fruition).

[LoRd_MuldeR]

But my question was regarding plasma, not LCD.

Sees like I missed that part :o

But if our eyes were fast enough, we wouldn't just see a dot - because a CRT has a phosphorescent screen that holds the brightness long enough for the other lines to be drawn before the beam comes back to that spot

I think the truth is in the middle. The phosphorescent is fading slowly. Which is an effect you can hardly "simulate" on a LCD screen.

So even if the beam takes a 60th of a second to scan each line of the field, each of those lines are supposed to be from the same instant in time. Which is why plasma could possibly be even more accurate, in that respect. Or if plasma isn't up to the task, perhaps SED will be (if it ever does come to fruition).

Yes, that's why I speculated that Plasma is more suitable for your proposed method. But I don't know enough about Plasma in detail to answer this :)

Maybe it's the "dimming" effect that would be missing on a Plasma screen to make Interlacing look acceptable...

[2Bdecided]

So every 60th of a second instead of showing a bobbed field, or only half of the lines, it shows the previous field mixed with the new field - in a new frame, redrawing the entire frame.But that frame would include (and therefore you would see) combing every time something moved!

btw, you can drive LCDs, plasmas, or anything with a progressive signal that just contains the active lines of an interlaced signal, and black for the missing lines. Of course it halves the perceived brightness, though you can compensate for that a little. It works OK, but the flicker is objectionable. However, the motion portrayal and lack of deinterlacing artefacts matches a CRT. No plasmas or LCDs do this, but it's easy to do in AVIsynth. If playing from a PC, the effect breaks down if the video frames aren't synchronised to your monitor's refresh rate perfectly.

Cheers,
David.

[NerdWithNoLife]

A CRT with interlaced video shows combing every time something moves too.

Keep in mind what the CRT actually shows during interlacing. There is another "frame" in between what an LCD would show with interlaced video. What I mean is, first frame 1 field 1 is drawn. But then frame 1 field 2 is drawn. At that point the viewer sees both fields of frame 1 - and the combing. An LCD would show this (if it didn't deinterlace). But a 60th of a second later, frame 2 field 1 is drawn. It appears at the same time as frame 1 field 2, with combing. An LCD would not display this instant in time. That's why interlacing looks so bad on a PC. It skips half of what the viewer would see on a CRT.

To see the difference, play back an interlaced video a frame at a time on your PC. It doesn't look right. But now add DoubleWeave, and watch what happens when you step through the video. You still see the combing, but it makes sense. When shown in this order, that's exactly what the brain does: it makes sense of the video.

So I'm saying, there's effectively no difference between showing doubleweaved combed frames every 60th of a second, and interlacing with a new field drawn every 60th of a second. And there's no reason to black out the other field when a new one is shown. The old on is supposed to be shown at the same time as the new one. That's how interlacing works.

[JonE]

There is another reason why I think interlacing may live on a long time.

The only reason it existed in the first place was it meant you could halve the bandwidth needed for a given number of lines whilst avoiding flicker.

The same still applies in the digital domain, at least for broadcast material. Less bandwidth = more channels = more revenue.

My TV has three electron guns, one for each third (top middle bottom).

Er .. you mean red, green and blue.

TTFN,
Jon

[LoRd_MuldeR]

A CRT with interlaced video shows combing every time something moves too.

In fact it doesn't show any combing, because it never displays two fields at the same time.

First it shows the "odd" field for 1/60 of a second and then the "even" field for 1/60 of a second (or vice versa).

But the LCD would show both fields together for 1/30 of second, which leads to the "combing" effect.

The same still applies in the digital domain, at least for broadcast material. Less bandwidth = more channels = more revenue.

No, I think 30p costs less than 30i, because progressive video can be compressed more efficient than interlaced. Also 30p is enough, except for sports maybe.

BTW: German HDTV was sending with 720p at 50p, which is far enough as very few sources are good enough to be worth 1080p (be it 25i, 25p or 50p). And no nasty interlacing!

Also I noticed that most broadcasts were 25p frame-doubled to 50p anyway :scared:

[NerdWithNoLife]

In fact it doesn't show any combing, because it never displays two fields at the same time.

First it shows the "odd" field for 1/60 of a second and then the "even" field for 1/60 of a second (or vice versa).
It draws a field every 60th of a second, but the previous field is still visible while it does, because of the phosphors that retain the brightness.
No, I think 30p costs less than 30i, because progressive video can be compressed more efficient than interlaced. Also 30p is enough, except for sports maybe.
I agree. The bandwidth savings made a lot of sense in the old days when the signal had to be continuous and analogue, but that advantage doesn't translate the same way into digital. In the future there's no reason we can't still have progressive frames, but more per second to handle motion better. No reason except monitors that can't keep up in refresh time.

[LoRd_MuldeR]

It draws a field every 60th of a second, but the previous field is still visible while it does, because of the phosphors that retain the brightness.

Yes, the phosphor retains the light for a limited amount of time. But it's fading continuously. It holds the light long enough to give the impression of a "complete" image (instead of a moving dot), but it also holds the light short enough that never two consecutive fields are visible at the same time. If the latter wasn't case, then a CRT would have the same kind of visible "combing" artifacts on all moving objects that a LCD has with none-deinterlaced (weaved) interlaced contents...

[NerdWithNoLife]

Ok, granted. A 60fps doubleweaved video may not be an exact representation of interlacing, but perhaps a 120Hz monitor could add another in between frame to simulate fading of one of the fields. So there would be a new field, next interpolation to the effect of that field fading, then the next new field with the previous one fading even more, and so on. I just wish we could put more effort into simulating interlacing, rather than avoiding and removing it with loss of motion and nasty artifacts.

But the previous field is meant to be at least partially visible while the new one is drawn (and not completely replaced by it), or else there wouldn't be odd/even fields, and bobbing issues.

[2Bdecided]

A CRT with interlaced video shows combing every time something moves too.No it doesn't - the decay is exponential - there's something left by the time the next field is drawn, but not much.

(There's even a little left after that, as can be seen by displaying a very bright white dot moving on a black background, viewed in the darkened room. It leave a visible trail on the display, but it's really feint. There's effectively nothing left from one field to the next with "normal" content and viewing conditions).

Keep in mind what the CRT actually shows during interlacing. There is another "frame" in between what an LCD would show with interlaced video. What I mean is, first frame 1 field 1 is drawn. But then frame 1 field 2 is drawn. At that point the viewer sees both fields of frame 1 - and the combing. An LCD would show this (if it didn't deinterlace). But a 60th of a second later, frame 2 field 1 is drawn. It appears at the same time as frame 1 field 2, with combing. An LCD would not display this instant in time. That's why interlacing looks so bad on a PC. It skips half of what the viewer would see on a CRT.Interlacing looks bad on a PC because is show a combed frame, not because of this non-existant persistence of fields that you'd just invented!

To see the difference, play back an interlaced video a frame at a time on your PC. It doesn't look right. But now add DoubleWeave, and watch what happens when you step through the video. You still see the combing, but it makes sense. When shown in this order, that's exactly what the brain does: it makes sense of the video.No, completely wrong.

So I'm saying, there's effectively no difference between showing doubleweaved combed frames every 60th of a second, and interlacing with a new field drawn every 60th of a second. And there's no reason to black out the other field when a new one is shown. The old on is supposed to be shown at the same time as the new one. That's how interlacing works.No it's not.

The reason bob deinterlacing on a display looks worse than a native interlaced CRT is because it reduces the resolution vertically. Your eyes/brain are capable of doing some of the advanced processing within mcbob / tgmc simply by looking at the screen, and tracking moving objects with your eyes.


btw, you can see combing on a CRT - when showing progressive content on an interlaced CRT, sitting reasonably close, and tracking the object's movement with your eyes.

Cheers,
David.

[NerdWithNoLife]

No it doesn't - the decay is exponential - there's something left by the time the next field is drawn, but not much.
Could that exponential decay be simulated?

The reason bob deinterlacing on a display looks worse than a native interlaced CRT is because it reduces the resolution vertically. Your eyes/brain are capable of doing some of the advanced processing within mcbob / tgmc simply by looking at the screen, and tracking moving objects with your eyes.
But the image will bob up and down since one field is supposed to be slightly higher than the other. Like what you see after SeparateFields(). It's not solely a reduced resolution issue, is it? Isn't that the whole theory between even and odd lines?

btw, you can see combing on a CRT - when showing progressive content on an interlaced CRT, sitting reasonably close, and tracking the object's movement with your eyes.
Vs.

No it doesn't [show combing on an interlaced CRT] - the decay is exponential - there's something left by the time the next field is drawn, but not much.
Uh - so I wasn't entirely off that there is combing while interlacing is properly displayed. Assuming we had a plasma 120Hz monitor hooked up to a PC, could there be a way to create a video stream - from true interlaced footage - for it that could come close to the look of interlaced video? If there was, some people would prefer that to deinterlacing.

[LoRd_MuldeR]

With that script you can test your method:

i = MPEG2Source("Interlace.d2v")

s = i.SeparateFields()
b = BlankClip(length=5000, width=s.Width, height=s.Height, fps=25, color=$000000, pixel_type="YV12")

e = s.SelectEven()
o = s.SelectOdd()

x = Weave(Interleave(e, b))
y = Weave(Interleave(o, b))

Interleave(x,y)

It does work, but "great" is something else. Note that any resizing while playback would destroy the effect.

(Maybe one could weave with a darkened version of the previous field instead of black video to make it look better. Not sure how to do that best)

[NerdWithNoLife]

MPEG2Source("Interlace.d2v")

e = SeparateFields().SelectEven()
o = SeparateFields().SelectOdd()

eDim1 = e.Tweak(bright=-4)
eDim2 = e.Tweak(bright=-16)
eDim3 = e.Tweak(bright=-32)
oDim1 = o.Tweak(bright=-4)
oDim2 = o.Tweak(bright=-16)
oDim3 = o.Tweak(bright=-32)

i1 = Weave( Interleave(e,oDim2) )
i2 = Weave( Interleave(eDim1,oDim3) )
i3 = Weave( Interleave(eDim2,o) ) .DoubleWeave().SelectOdd()
i4 = Weave( Interleave(eDim3,oDim1) ) .DoubleWeave().SelectOdd()

Interleave(i1,i2,i3,i4)
Now all I need it a 120Hz plasma... then I can tell just how dumb I am.

[LoRd_MuldeR]

Something like this could work:

s = MPEG2Source("Interlace.d2v")

e = s.Tweak(bright=16).SeparateFields()
o = s.Tweak(bright=-64).SeparateFields()

q = Interleave(e,o)
r = q.Trim(1,q.FrameCount)

return Weave(r)

[NerdWithNoLife]

I'm just curious to see whether a plasma could come close to simulating interlacing. I'm not suggesting any movie be encoded this way, but for playback it would be interesting. But if nothing else, it could be used as a visual effect for the TV screen look. If 120Hz isn't enough to satisfy Nyquist's limit, perhaps if SED ever comes to market its insane refresh rate would be good enough to pull the concept off.

[lpcstr]

Wow, it's hard to believe that some people are worried that interlacing will go away. Interlacing is a plague that should have been eradicated a long time ago, and it's only now starting to show sines of going away. The sooner, the better.

[NerdWithNoLife]

I'd be happy to see 60 progressive frames per second take over, but in the mean time there are compatibility issues from the loads of footage that is (and was originally) interlaced. I prefer to watch my interlaced footage on an interlaced display; and I'd rather see progressive stuff on an interlaced monitor than interlaced stuff on a progressive monitor. Is there anything wrong with a monitor that could display both? It doesn't exist now, but someday it could.

[lpcstr]

I'd be happy to see 60 progressive frames per second take over, but in the mean time there are compatibility issues from the loads of footage that is (and was originally) interlaced. I prefer to watch my interlaced footage on an interlaced display; and I'd rather see progressive stuff on an interlaced monitor than interlaced stuff on a progressive monitor. Is there anything wrong with a monitor that could display both? It doesn't exist now, but someday it could.

Well, how long were you planning on holding on to that interlaced footage? I'd like to know so I can tell researchers and manufactures how long they have to hold back progress.

Seriously, are you going to hold on to your interlaced DVDs for 20 years? By the time interlacing is completely done away with there wont exist any DVD players to watch those disks anyway, so there's no problem.

[2Bdecided]

Well, how long were you planning on holding on to that interlaced footage?I don't know - how long do you think the world's archives will hold black & white footage? 4x3 footage? SD footage? etc etc etc!!!

What do you want? A big bonfire of all content that's not available as 1920x1080p24?!

Cheers,
David.

[lpcstr]

My point was that in the not so distant future, they won't be filming anything interlaced, and you'll have thrown out your old warn out DVDs, DVD players won't even exist, so there will be absolutely no need for any hardware to support interlacing.

Movies come from progressive film, so 24p content can be generated for all movies. A lot of television shows are filmed progressively too. 20 years from now if you're just dying to watch season 3 of Friends, you can handle watching a deinterlaced copy if it, or a copy with a bit of combing.

[scharfis_brain]

@LoRd_MuldeR:If the latter wasn't case, then a CRT would have the same kind of visible "combing" artifacts on all moving objects that a LCD has with none-deinterlaced (weaved) interlaced contents...
If you watch close enough you'll notice combing on every CRT-TV set (50HZ/100Hz no matter what!).
This is due to the persistence of vision in our eyes.

Btw.: 100 (120) Hz CRT TVs are also scanned interlaced. But their field frequency is 100 Hz. They won't display 100 Full images per second. So This also introduces new artifacts, but these are more difficult to track down.

[NerdWithNoLife]

Well, how long were you planning on holding on to that interlaced footage? I'd like to know so I can tell researchers and manufactures how long they have to hold back progress.
A refresh rate that can barely handle 24p is progress? What's wrong with having a lightning fast refresh time? Progressive footage is great, but many newer high tech sets couldn't even show 48p from IMAX. Edit: (based on refresh time alone; I know there are other issues).

[2Bdecided]

20 years from now if you're just dying to watch season 3 of Friends, you can handle watching a deinterlaced copy if it, or a copy with a bit of combing.Friends is progressive.

My camcorder footage isn't. That's the thing I'm most bothered about being able to watch in 20 years time.

All live sports events are either 576i SD or 1080i HD over here. Not that I watch sport, but I assume the market for it will be greater than zero in 20 years time ;)

There might be a day when people archive all their interlaced content onto progressive-only formats (like the way all composite originated content is now archived onto component-only formats), but we're not all there just yet. The best deinterlacers aren't fast enough, or "perfect" yet.

Cheers,
David.