New motion compensated deinterlace filter (TomsMoComp)

[Mug Funky]

hmm... nice idea (wtf? i have no clue :))

one nitpick - the effect of phosphorescence only lasts a few lines.

try take a photo of a TV screen at a shutter speed less than 1/50. it's rather like exponential decay in cooling, and i've never been able to move my eyeballs fast enough to notice this effect.

i'd love to see edge-direction and motion compensation together at last though. definitely.

(wow... this is an OLD thread! hehe)

[mf]

Originally posted by Mug Funky
one nitpick - the effect of phosphorescence only lasts a few lines.
Then explain to me how it comes that certain old monitor types that happen to have a slower phosphor will appear to flicker less than newer ones (same goes for TVs) at the same refresh rate? :) Observation has proven that the effect reaches beyond "a few lines". Also, I've noticed that modern TVs appear more easily "pixelized" - the jagged edge you see when an object on an interlaced screen is in motion. I've done some comparing in TV shops with a lot of screens displaying the same signal :).

[Mug Funky]

ah, yes. i thought about old green monochrome screens, but didn't bother to mention it. those were SLOOOOW.

my current tv has only about 100 lines of phosphorescence, and none of them are above half the intensity of the actual scan.

of course on an older tv there might be a more visible effect, but i've only ever noticed pixellation and apparent interlace artifacts when watching telecined NTSC (which i avoid like the plague for obvious reasons in PAL land :))

artifacts are easier to spot with less scanlines

[kassandro]

I can't believe in phosphorescence. If it were true, then you must see combs in the presence of motion, but I have never seen any on TV. When I watch CNBC on my digital satellite TV card, it is very difficult to read the ticker tape, because the numbers are moving fast and interlace makes them almost unreadable. On the other hand, if I watch the same program on TV the numbers are just sharp. There cannot be phosphorescence. I think the flicker is due to the whimpy 50 fps. The Americans with their 60 fps are much better of in that respect.

[mf]

Originally posted by kassandro
I can't believe in phosphorescence. If it were true, then you must see combs in the presence of motionWrong. The other way around. If the phosphor is too fast you see combs. Just put your computer monitor in interlaced mode (if you have an ATI card, nVidias don't support it). Most of the combs are invisible but some come through.I think the flicker is due to the whimpy 50 fps. The Americans with their 60 fps are much better of in that respect.Beside the point. I'm talking about that flickering gets worse or better at the same exact refresh rate with varying monitors (and thus phosphor types). If you want to test it for yourself, take a brand new LG or whatever monitor, put it at 60Hz, and then compare it with the same brand from a few years back (I've tested it with an Iiyama 21" from 1995). Same resolution. Same refresh rate. Less visible flickering.

(We are getting pretty off topic though - might want to resume discussion about deinterlacers before a mod feels the need to intervene)

[billou2k]

I dont think it is that offtopic. After we re taking about why we have to deinterlace. On the opposite I think it's intersting if we can make our mind on what really are the differences between interlaced tv playback and pc monitors and why it is so difficult to play interlaced material on a monitor.
If you have got/found technical articles about it that would be great to share here as well I think. That's intersting to read all your opinions.
As for the basic deinterlacing filters existing at the moment is there one that basically plays back a 25/30 fps video at 50/60fps and simply refreshes a field every 1/50th 1/60th of second between the lines of the previous field?
(just to see what it looks like)

[mf]

It is not so hard to display interlaced material on a monitor - you just need a proper videocard and a CPU beefy enough to handle the InterlacedResize() to your display resolution. I've tested it with telecined anime at a high refresh rate (100Hz) in interlaced mode, and even if the refresh rate doesn't match(!) the output looks smooth and not combed. Only sometimes does a little combing come through (at low motion) - which apparently doesn't change or go away by matching the refresh rate to the video framerate (60hz). I was amazed at the fact that it just looked smooth, as I've never seen telecined material on a real TV before (I live in PAL land so if anything's been telecined it's also been converted back to PAL causing choppiness).

[kassandro]

Originally posted by mf
Wrong. The other way around. If the phosphor is too fast you see combs.
Almost by definition, you can only see combs if two fields (with a time difference) are displayed simultaneously and this can only happen if phosphorescence lasts too long and not if it is too short. Probably, if you would have watched tv on these old green monochrome screens you would have seen combs. If phosphorescence would last longer than 1/50 of a second then you would begin to see combs for PAL clips. On the hand, if phosphorescence is too short, than you would see flicker. Ideally you would like to have phosphorescence a little bit less than 1/50 of a second and then to disappear instantly. Unfortunately, this is impossible, because the decline of phosphorescence is more or less exponential in time. That's why a high vertical frequency is so important to reduce flicker. The high priced 100 HZ tvs therefore display each field twice, to reduce flicker.

[kassandro]

Sorry I was quoting instead of editing!

[scharfis_brain]

If you want to see, how a TV work, try to simulate that with my scanlines-simuator
TVdeint


a normal 50Hz (60Hz) TV should work like this:
tvdeint(0.2).temporalsoften(some strong parameters)

a really cheap 100 (120) Hz TV works like this:
tvdeint(0).selectevery(1,0,0).temporalsoften(...)
(this TV will have reduced area-flicker but line-flicker is not gone)

a better 100/120 Hz TV uses complete bobbing-deinterlacing with thresholding to reduce all flicker

[kassandro]

No! The combs are just replaced by ghosts, which is typical for blur deinterlacers. On a TV you have neither ghosts nor combs. I think the human brain "remembers" the old field with a motion correction such that it fits to the new field, which is just displayed. From this point of view the sharp impression of interlaced tv is just an optical illusion, which you simply can't emulate on a computer screen.

I simply loaded the wrong script. Now the flicker is so strong, that nobody would watch such kind of tv. I have a two week old 19 inch TFT-display (DVI connected), though.

[MfA]

100/120 Hz tvs use motion compensated deinterlacing (of a slightly higher calibre than TomsMoComp ;).

[scharfis_brain]

@MfA: only the really expensive ones.
the cheap ones are only displaying every field twice (flickering on thin lines)
the better ones have thresholded deinterlacing (stairstepping)
and the best ones have motion-compensated deinterlacing and/or fieldmatching for PAL/NTSC-Film.

[Mug Funky]

interestingly enough, the REALLY cheap 100hz prog TVs actually display combing! just like a computer screen.

[Didée]

Well, I'm sitting here with a 100Hz TV of Phillips. It's the "biggest" model they offer without their "true motion" technology.

- On motion, some little ghosting is present - must be combing, but in fact I cannot spot it as such.

- scrolling text is 100% combed, and often just un-readable.

Now, is a device for € 1200.-- really THAT CHEAP? I had expected a little more of it!

- Didée

[scharfis_brain]

hehe, saw that on a sony, too

that must be some kind of deinterlacer, that doubleweaves lowmotion-areas and bobs high motion.

scrolling text has maybe not enogh motion for getting bobbed.

this means that probably a too high threshold for motion dectection is choosen.

Aren't there any hidden menues or options to customisze the deinterlacing of yout Philips TV?


btw. I really dislike 100 Hz TVs because they cannot reproduce the input-signal as it was.
Most of them also apply some kind of image-(dis)-enhancement, which ofversharps edges, flattens detailed areas, smears motion.

I like my good old pure analogue 50 Hz-TV with a slow Phosphor, which makes the 50Hz flicker not that noticeable...

[avih]

or, just get a BIG (and good) progressive monitor, and use dscaler ;)

OT: the only issue i have with dscaler is that i can't sync the monitor refresh 100% to the video signal frequency (or a multiple of it).

i actually think that dscaler CAN do better in this department (use some software schmidt-trigger to 'move on' to the next frame sync). otherwise, every once in a while i get about a full sec of 'matching refresh rate to video freq' jitter...

i wouldn't mind even dropping/dupping a field once in a while instead of a jitter for a full second. i just want it smooth ;)

[Leak]

Originally posted by avih
OT: the only issue i have with dscaler is that i can't sync the monitor refresh 100% to the video signal frequency (or a multiple of it).

i actually think that dscaler CAN do better in this department (use some software schmidt-trigger to 'move on' to the next frame sync. otherwise, every once in a while i get about a full sec of 'matching refresh rate to video freq' jitter...

I haven't used DScaler for some time since I hardly watch TV anymore (and if I do, I do it on our 16:9 set), but I guess at least version 5 should be able to fix that using Reclock since it's using DirectShow all the way...

Of course, version 5 also got it's own MPEG2/DVD decoder so it'll be perfect for watching DVDs, but alas it's still some way off... :(

np: Latyrx - Latyrx (The Album)

[avih]

well, reclock adjusts the VIDEO (through the audio) playback speed to match the monitor's refresh rate. but with TV playback (or an external video signal) you just can't change its rate. it's a broadcast. all you can do is either change the monitor's refresh rate in few ppm (part per million) to match the video (probably depends on the capabilities of the display card, i know my old voodoo3 could do that), or dup/drop fields (was frames) once in a while to get back into sync. mind u, i'll assume a monitor refresh a double of the field rate (i.e. pal 50Hz, monitor 100Hz). so by adjusting using 1/100 of a sec, it could be hardly noticable, but much smoother.

[Leak]

Originally posted by avih
well, reclock adjusts the VIDEO (through the audio) playback speed to match the monitor's refresh rate. but with TV playback (or an external video signal) you just can't change its rate. it's a broadcast.

Ugh, you're right - what was I thinking? *bangs head on desk*

np: The Orb - Rose Tinted (Dal Vivo A Roma)

[TEB]

Hi. when using .info in avisynth 2.5.6 it still says Parity: top field first after using this deinterlacer.. sHouldnt avisynth report progressive?

[Wilbert]

AviSynth keeps track of whether a given clip is field-based or frame-based. If the clip is field-based it also keeps track of the parity of each field (that is, whether it's the top or the bottom field of a frame). If the clip is frame-based it keeps track of the dominant field in each frame (that is, which field in the frame comes first when they're separated).
http://www.avisynth.org/AssumeFieldFirst

[TEB]

thx for the response but i still dont get it...:( I thought the point to deinterlacing was to make the frame progressive, hence one shouldnt need to set that avisytht variable? Or am i missing something major here?

[Wilbert]

Currently (v2.5x and older versions), AviSynth has no interlaced flag which can be used for interlaced video. There is a field-based flag, but contrary to what you might expect, this flag is not related to interlaced video. In fact, all video (progressive or interlaced) is frame-based, unless you use AviSynth filters to change that.
http://www.avisynth.org/AdvancedTopics