To deinterlace or Not to deinterlace?

[shred]

I am capturing at 640x480 29.97(NTSC) and there are obvious signs of interlacing in high motion frames.

I will be converting to SVCD (480x480) using TMPGEnc.

I will be viewing the SVCD on a standard NTSC TV.

If I leave everything exactly as is, will I get the best (original) quality for TV viewing?

Should I deinterlace using the VirtualDub filter (Smart Deinterlace) and then set the "Encode Mode" to Interlace in TMPGEnc when converting? This doesn't seem to make much sense. It seems like a lot of extra converting to get basically the same end results as if I had left the original frames unmodified.

Thanks
shred

[Betamax]

SVCD IS Interlaced, so if you're capturing in Interlaced why would you then de-interlace??

TV-output always is Interlaced, so it's obvious you get better quality when you create material that's Interlaced too.
While the restrictions of VCD is non-Interlaced, you will have less quality with that (but it's hardly noticeble if the captured source is of high quality, like DV and full DV-resolution).

[banjazzer]

Leave as interlaced. However, you need to get the field order right. If the result is obviously jerky, the field order needs changing.

[dphirschler]

Leave interlaced unless it is film-based. For film-based, use inverse telecine and burn a 24fps film svcd. Your player will do the 3:2 pulldown on playback.


Darryl

[shred]

I thought it would make sense to leave it interlaced, so thank you all for confirming this.

Now, is there a good way to "Preserve" the original interlacing.

I am capturing w/ VirtualDub at 640x480 and converting w/ TMPGEnc.

Obviously I should not do any deinterlacing using either piece of software right? My question is about the Source and Encode settings in TMPGEnc.

I assume they should both be set for interlaced, but this does not seem to get the best results. It seems to cause some heavy pixelation effects in high motion sections of my video. Would setting both to non-interlaced "preserve" the source?

I am testing with a 30 second clip that I have encoded in every possible combination of interlace settings. I am using CBR at the highest bitrate setting as I am not worried about saving space at this point.

Again, I will be viewing the final SVCD on a standard NTSC TV.

Thanks again,

[SLK001]

No, it doesn't make sense to leave it interlaced - especially if you are making SVCDs. Encoding with interlaced video will cost you about 20 - 25% of your bit rate over non-interlaced video (since one frame out of every five is a frame composed out of previously encoded video, if you encode these, it will cost you "real" bps).

Capture your video.
Inverse Telecine it to FILM.
Encode as FILM.
Perform 3:2 pulldown.
Author the SVCD.

[banjazzer]

I can't agree that throwing away half of the video information makes any kind of sense.

[SLK001]

Well, it doesn't, but we're talking about removing the TELECINEing process, not separating fields and discarding them.

Take this typical NTSC telecined stream:
TOP_FIELD - A__B__B__C__D
BOT_FIELD - A__B__C__D__D

We have created a frame from two previous fields that already existed. If we encode this stream, we are wasteing the bits needed for one of these frames.

Now take this typical FILM stream:
TOP_FIELD - A__B__C__D
BOT_FIELD - A__B__C__D

If we just encode this one, then tell the DVD PLAYER to perform the 3:2 PULLDOWN by putting in the proper flags (they routinely do this for commercial DVD's), we have saved the bits that it would have taken to encode the first example.

"Interlace" is, I'll agree, a poor choice of words. All TV in the world (except some HDTVs) is interlaced, in that EVEN fields are transmitted, then ODD fields are transmitted, giving an "interlaced" screen. This interlacing really becomes visible when the signal is telecined as above.

[banjazzer]

NTSC has extra fields to NTSC(film). By inverse telecining you should be able to remove the extra padded fields. However, both before and after you are still dealing with interlaced material, so I cannot see any need to de-interlace.

I must add that I very rarely deal with NTSC material, so this is a theoretical rather than practical answer.

[SLK001]

If you encode as film, then your available bitrate per second is spent over 24 individual frames. If you encode already telecined, then you are spending the same bitrate per second over 30 individual frames, thus robbing 25% of the bits per second from each of the 24 frames to encode the last six. Obviously, you will suffer a quality hit on the original 24 frames.

[banjazzer]

I was suggesting inverse telecining before encoding, but leaving both source and object interlaced.

[vhelp]

Some anal'ness to take w/ a grain of salt, if you wish..

NOTE, all this dicussion below will assumed Capturing Cable/Satalite.

The short version.. NOT to de-Interlace.
However, the short, and not TOO anal.. MPEG-2 has support, in that it
has flags or fields that, when played via your DVD player, it will display
those fields/frames correctly for your TV screen, hence Interlaced.. where
as for MPEG-1, this support, or flag/fields in the MPEG-1 header or
chunks (whatever you want to call it) are NOT their, hence your issues
when you play an MPEG-1 at full frame ie, x480
Your TV is NOT getting the proper layout, by your DVD player, cause MPEG-1
does not have the Interlace support inside or embeded in the MPEG-1 file.

But, the better your:
* Source's "originator's qulaity" ie, broadcaster who aired clip (qlty it's in)
* Source quaity being aired finally
* your equipment to RECEIVE the aird source
* your capture equipment's ability and quality
* your skill in obtaining this broadcasters source, ie TV show or movie
* your level of skill to capture this source
* your ability to, and finally, encode this source (filtering, editing, etc)
* your Encoder's abilites, MPEG algorithem, and final quality etc.
* your skill in using the encoding, and then some (above)
* your authoring software/equipment
* your DVD players specs and abilities and quality attributes
* your TV ( size does count for some )
* your TV's method for handling Interlace video
* type of TV, ie widescreen etc.
* and, finally, your ..

..will reflect your final quality from an MPEG-2 encode w/ Interlace on. So..
given the above, the better your chances at obtaining good quality from an
MPEG-2 encoded WITH Interlace on. Now, with that in mind, lets move on..

De-Interlacing is another issue w/ respect to final quality.
The better your method of de-Interlacing, the better your final quality will
be OVER an encoded MPEG-2 clip while Interlaced turned on.
If not de-Interlaced properly, you'll have worse'r looking enocded vs. MPEG-2
encoded w/ Interlace on.

IVTC'ing goes hand-in-hand w/ de-Interlacing.
Source etikets will refelect needs for IVTC or not vs. De-Interlacing only,
instead. ie, if source is pure Interlaced (ie, evrey frame interlaced) the etikets
would NOT require IVTC, but rather, de-Interlace instead.

One of the main reasons why there are so many poor looking final encodes via
a so called good de-Interlace, is because those de-Interlaced method(s) do not
account for IVTC (or telecined) materials. Instead, these de-Interlace'ers
de-Interlace every field. Hence, a failed method. Hence, a poor quality
encode - no matter how high bitrate (subjectively speaking)
This is because, everyone here agrees (incorectly) that when you capture w/
both frames or x480, that your source will be Interlaced - NOT entirely true !!
It will be Interlaced IF:
A, source (broadcaster's source is Interlaced or IVTC)
B, source (broadcaster's source is Pure Interlaced)
C, NOT.. if source (broadcaster's source is non-Interlaced)

* IF Line C, then even if captured at x480 your final source would be of
.. NON-Interace nature. Encode w/ Non-Interlace !! Never Interlaced.
* IF Line B, then this capture IS going to have Interlace frames.
* IF Line A, then this capture IS going to have Interlaced and/or IVTC frames.

The only time when your final capture clip will NOT have Interlace frame
even though you capture at x480 is when Line C is met. Otherwise, 99% of
the time, you will have Interlaced frames. Line C will depend on your
broadcasters Source and/or HOW they air THEIR source materials or how they
receive their source (ie, from 3rd party etc)

I'll sort of repeat what I said above..
When you capture both fields/frame or x480, you need to look for certain
condition in your video.
1, are the frame PURE Interlaced, every frame is distorted looking ?? or,
2, are some frame clearn while others distorted ?? ie, 3 clear and 2 distored (IVTC)
3, broken or hybrid IVTC ie, 3,3,3,3,3,3,2,3,3,3,3,3,3,2 patterns or worse
than these even.
4, other types

with lines 1 and 2, you can pretty much bet you'll get good quality encodes
IF you encode perperly (that will require time and skill ect)
Line 1 would undoubtly require a GOOD de-Interlacing method, 29.970 fps, while
Line 2, would undoublty require a GOOD IVTC mehtod. as this is Film 23.976 fps
These methods take time to learn and master. I'm working on Line 3. These
are the tricker ones to manage or develope a good method for.

I will not go ANY further into the inners of:
* Interlace
* MPEG-2
* Broadcasters source being aired
* and HOW would be THE perfect method
.. cause their aint, yet
At the moment, I can't put them all into words just yet. So.. I'll wait
and leave it up to someone else that can better explain in words, what I'm
talking about above, and would love to "continue" talking about here.

Well, have a good evening all.
-vhelp

[SatStorm]

Vhelp, I asked you in the past, I asking you again today:
What is that " Satalite " you keep refering to your posts? I don't know anything "Satalite"
Or you mean Satellite? DVB /s(atellite) transmissions?

Anyway, the subject "Inverse Telecine" is a big one.
Technically speaking, the best way is using Vdub's ITC based on reconstructing from fields (adaptive). Another alternative is using TMPGenc's filter for this job. Most of the times, I prefer TMPGenc method an I'm ITCing with flickering priority (otherw may find motion priority more pleased. Test what you like most)

Both methods are slow, and true experts use Avisynth special scripts with CCE for this. I am not an expert, so I prefer the Vdub/TMPGenc solution. I never liked CCE or Avisynth to tell you the true. I always find myself work much easier with TMPGenc and Vdub. Whatever...

If you use the TMPGenc method, you keep the interlace source as is and you don't de-interlace.
If you choose the Vdub method, de-interlacing using the adaptive method in TMPGenc, may show better results. Expecially for HDTV, advance users find it better.
The final choice is your ofcourse....

[pijetro]

Wow, great info guys, but back to the original question..

Short answer, if you're capturing Blockbuster videos, then SLK001 is totally correct. If you're capturing TV signals, homemade movies, sattelite etc.......... then don't deinterlace.

As the people above are pointing to, rented VCR tapes are telecined. Inverse telecining is a space saver and gets rid of deinterlace :P .
For captured home videos, you leave alone, because Mpg2 supports the interlace, and the quality is maintained from the original source. :P [/b]

[Dragonsf]

But one question remains:I'm capturing through DV from VCR in EU (PAL 25fps).On some movies (esp. with horiz. panning), the interlace artefacts are still to be seen very prominent on the SVCD.But when I deinterlace (with VDup), these artefacts are completly gone.The overall inpression of the deinterlaced SVCD is not worse then the interlaced version.
How can you explain this effect?

[freak_in_cage_10k]

Hi, I have a question regarding interlacing, i was going to post it in a new topic but figured it would save peoples time if i put it in here:

i am using DVD2SVCD and CCE v2.5. Should i interlace the movie or not
? i am a bit of a newbee and so a simple yes or no would be preferred although any advice is welcome!!! I am encoding from a Pal DVD.

Thanks very much everyone!

[SatStorm]

@ Dragonsf: If you using TMPGenc, then I suspect you have to change the source aspect ratio to 4:3 display and the video arrange method to: Full Screen (keep aspect ratio 2). I have this effect only if I mess up the input/output aspects and TMPGenc try to resize/feet the screen on the output.
Overall, it is a aspect problem, try some combos untill you found what you have to use!

@freak_in_cage_10k
For PAL no
For NTSC no for normal TV, yes for HDTV

[banjazzer]

Almost all PAL DVDs are progressive (not interlaced). The main exceptions are DVDs compiled from material first shown on TV.

[Dragonsf]

No, I used ULEAD MF for converting.As the movie ( TV release of BTTF III especially) was in 4:3 and the end result was 4:3 and no cropping involved, I think aspect ratio was no issue.
The effect is always good be seen with fire/explosion/steam and such fast moving events.

[SatStorm]

@ banjazzer: I noticed exactly the opposite. ALL the PAL DVDs I know are interlace with only a few exceptions.
Most of them are manufacture in east Europe, and some in southern Europe (Italy, Greece...). All are interlace. And most PAL post in this forum state that the DVDs they use to backup are Interlace.
Even some NTSC DVDs I get from internet are interlace.
From which manufacture you refer to?
I am very interesting for progressive DVDs. They look excellent on cheap videoprojectors!


@Dragonsf: I can't help you, sorry. I only experience the problem you describe when I mess up TMPGenc, like cropping the top and the bottom of a capture and not use "Full Screen (aspect ratio 2)" when I encode.

[shred]

Wow, Lots of information... I love it.
Thanks so far, to everyone for offering their suggestions.

I will tell you what I have found so far while working with a few clips captured from my camcorder (Sony C-VHS). The source "looks" to be interlaced. All frames with any motion have the obvious interlacing/comb effect when viewed on my PC.

As stated before, I am capturing at 640x480 29.97 fps (NTSC). And I am encoding to mpg2 480x480 (SVCD). Burning with Nero. I am making my comparisons by viewing on my TV(Sony Trinitron 36")

If I DON'T deinterlace, I get quite a lot of distortion in high motion frames, no matter how I encode (Interlaced or Non-Interlaced) although I can tell that the frames are still interlaced because when I pause the picture on the screen it "shakes".

If I DO deinterlace and then encode as interlaced I get what seem to be good results, but the final product is no longer truly interlaced. When paused, I get a still picture on the TV.

For testing I a captured the same clip at 320 x 240 and encoded with no interlacing, and got good results too. I can't really tell the difference between this sample and the one just above.

I would like to "preserve" the original quality and interlacing if at all possible, but the picture seems to get very distorted/pixilated if I don't deinterlace. And don't forget, I am encoding with TMPGEnc using CBR at the highest possible rate.

Thanks again, to everyone

[banjazzer]

ALL the PAL DVDs I know are interlace with only a few exceptions.

OK. I'd better not generalise. All UK PAL DVDs with a few exceptions are progressive. 8)

[shred]

My Theory:
During compression TMPGEnc does not handle interlaced video very well.

Why:
When I encode an interlaced video I get LOTS of distortion that is very evident when I pause the video at a frame with lots of motion. I am not talking about the flicker effect, because I should expect that in a truly interlaced video (right?). I am talking about heavy pixelation type distortion. The distortion is MUCH less evident if I deinterlace first.

Now I am relatively new at using TMPGEnc, and there are a TON of settings, so maybe I am just doing something wrong.

Would it help to capture at 480x480 to begin with? That way TMPGEnc would not have to do any resizing.

Also, I have not tested with Telecineing. I thought since my source and destination are the same fps, that it would not be useful to me.

Thanks again,
shred

[pijetro]

I'm also capturing C-VHS. If that's the case, then forget about telecining, as you are capturing already at 29.97fps. If you capture at 480x480, Tmpgenc should work faster, because it has less to calculate, and the quality shouldn't differ from 640x480.

Two things I suspect. 1)Field order is wrong, and 2) unfortunately the bitrate for SVCD isn't always ideal for the C-VHS source. It's great for sources like DVD rips, and MiniDV stuff, but I learned that the hard way, and think a DVD burner is the only way to go.

Try encoding a 10 sec. clip at DVD bitrates, and see what a difference it makes.......It's incredible.

[Dragonsf]

As I got a DVD burner lately, I must agree.The difference between SVCD and DVD is overwhelming.Not in the sense, that a proper done SVCD is much worse when viewed on a TV, but all the worries about bitrate/deinterlace and such aren't necessary anymore.MY DVD-way is:
Capture with DVApp thru HWDV Bridge to avi, cut/edit bad frames and commercials, pass it thru VDup (logoaway/border masking) to tmpgenc and encode/author it with ULEAD MF.This will take about 12 hours to do for 90 min of footage, but most of it runs will I'm sleeping.

[freak_in_cage_10k]

Thanks a lot SatStorm and banjazzer, v much appreciated!

[shred]

Thanks for the info pijetro.

I will try to convert to DVD bitrate and see what you are talking about. Unfortunately I do not have a DVD burner (yet), so I wont be able to compare on the TV, but I will compare on my PC and see if the distortion I am seeing in my SVCD appears to be gone.

I am sure my field order is correct, as the playback is smooth, but it is very distorted in areas.

One more question:
If I pause a truly interlaced frame on my TV it should look as if the frame is "flickering" back and forth correct? The reason I ask is that normal VHS tapes don't do that, but maybe that's because they are only 210 lines resolution and are not interlaced.

Thanks again, and again,
shred

[SLK001]

The reason I ask is that normal VHS tapes don't do that, but maybe that's because they are only 210 lines resolution and are not interlaced.

Once more, you are confusing "interlaced" with "telecined". ALL TV's - both PAL and NTSC - (except for some HDTV modes) are interlaced. To view something on your TV screen, it will be sent interlaced - odd fields displayed, then the even fields. The persistance of a TV is set up to be long enough that the corresponding odd or even fields blend together (gives a smooth transition). On a computer screen, the persistance is much shorter, so the comb lines created by the telecine process are noticeable.

If you intend to encode a telecined source (interlaced) you need to use ALTERNATE SCAN instead of ZIGZAG scanning for your MPEG conversion. I don't use TMPG so I cannot comment on it's ability to do either one.

[pijetro]

You seem to be preoccupied with that pause button lol:

Like SLK001 said, the TV is constantly running its scan, so you shouldn't notice the flickering. Who cares, as long as when it's playing, it's giving you great results. It could be your player. Did you try it on a friends standalone?

[shred]

No, I think I DO understand interlacing.

Tell me if I am wrong.
Each frame is made up of two fields which are interlaced as the even and odd lines on your screen. As your TV displays each frame it shows the even lines first, then the odd (or vice-versa). Since the second field to be displayed is slightly "later" it gives the effect of 60 fps. It's not very hard to understand.

Now, if I view a single frame paused on my Player (APEX AD1100W), if it is interlaced, the two fields, even and odd, should be from slightly different moments in time. Therefore as the TV is displaying this paused image on the screen it is displaying the even, then odd, then even, then odd, etc., etc. fields from this one frame. Hence the flickering effect. That should be one way to tell if the video is really interlaced right?

You're right, pijetro, I don't care if the picture "flickers" while paused as long as it plays back smoothly. I am just trying to determine if the video I am looking at is truly interlaced or not. I prefer that it flickers when paused if this means I am getting a truly interlaced video when played back at normal speed, because this should give me the smoothest playback.

The reason I am so curious about this is that some of my tests DO NOT flicker when paused even though I DID NOT deinterlace the original video. I am taking the same avi that was captured at 640x480 and using different source and encode interlace settings in TMPGEnc but NO deinterlace filters. Sometimes the paused video flickers and sometimes it does not. What would cause this? Does TMPGEnc automatically deinterlace with certain settings.

I just though of something though. When I pause a DVD, I don't think it flickers. Hmmm. I will have to test tonight when I get home to be sure. Are DVD players smart enough to only display one field when paused to prevent this flicker effect if they know the video is interlaced?

Many Thanks.