Log in

View Full Version : Interlaced capture is normal?


DJ_TripleM
11th April 2003, 23:49
After solving some other problems I've captured my first frames in full PAL resolution using Virtualdub. However, the output I get is interlaced, as you can see here:

http://www.dj-triplem.com/urpicshere/fok/capture2.jpg

Trying to solve the problem, I activated Swap Fields, which gave me this:

http://www.dj-triplem.com/urpicshere/fok/capture3.jpg

Neither is very good output. I asked a friend of mine and he said that getting interlaced output is normal. Other programs deinterlace on the fly, Virtualdub leaves all filters off so you get 'raw' output. I don't know, that means everyone would get interlaced output when using Virtualdub and I'm sure I would have seen a thread or two about it in here?

I know I know, I'm a complete noob when it comes to capturing video, but I hope to find someone with the patience to solve this for me by tuesday (this is when I want to record an important program). Even if you don't make it by then, thanks for any help you can give me.

jggimi
12th April 2003, 04:32
Usually, analog broadcast television is interlaced. Film content in NTSC broadcasts may be Telecined, which produces the same sort of artifacts. To avoid it, you could possibly capture at a low enough resolution that you only get a single fields worth of video -- vertical resolution of 288 (PAL) or 240 (NTSC) -- not usually recommended. There may be some capture cards that de-interlace on the fly -- mine won't. :-)

Or, you can deinterlace, typically in post-processing. I capture without any filters, and de-interlace during my final encoding steps. I use Don Graft's AviSynth filter set (DeComb), which includes a pretty good Field Deinterlace.

For a lot more information, see www.doom9.org/ivtc-tut.htm -- this link was intended for DVD transcoding to MPEG-4, and is not related to video capture. But you may find the sections that describe interlacing, Telecining, NTSC->PAL conversions helpful. In addition, at the bottom, there are a bunch of links you may find helpful.

^^-+I4004+-^^
12th April 2003, 05:19
> to solve this for me by tuesday (this is when I want to record an important program).

no problem,capture like this (interlaced,it's normal)
and you'll deinterlace afterwards (or not if you aim
to make dvd or svcd...)

but if i saw correctly you had some funny
resoultion of xxx x 545?
(or you modified the image before putting it on web?)
if humanly possible,you'll use 576 for pal...

there are plenty of decent deinterlacers out there,so
now your only job is to make decent
interlaced tv capture.....
if rally and motoraces can be deinterlaced OK,then
other things can be too.....

DJ_TripleM
12th April 2003, 07:38
Originally posted by ^^-+I4004+-^^
no problem,capture like this (interlaced,it's normal)
and you'll deinterlace afterwards (or not if you aim
to make dvd or svcd...)

but if i saw correctly you had some funny
resoultion of xxx x 545?
(or you modified the image before putting it on web?)
if humanly possible,you'll use 576 for pal...Ok, I'll deinterlace afterwards. What do you mean 'not if you aim to make dvd or svcd'? Because I plan to make both a DivX version as well as a svcd..

About the resolution, I think that's my mistake. I cropped the image a little bit too much. I checked and my capture mode is 720 x 576, UYVY. I don't get any frame drop so I'm happy :)

Thanks ^^-+I4004+-^^ and jggimi for your help!

jggimi
12th April 2003, 14:26
...What do you mean 'not if you aim to make dvd or svcd'? ...Because televisions are designed to show interlaced material. If you deinterlace, either through blending or interpolation, you do degrade the image. The consensus is that SVCDs / DVDs should not be deinterlaced.

VCDs, on the other hand, are at a low enough resolution for a single field (288/240).

DJ_TripleM
12th April 2003, 17:05
Originally posted by jggimi
The consensus is that SVCDs / DVDs should not be deinterlaced.
So if I make a SVCD, I only have to resize to SVCD resolution and convert to MPEG-2? And I won't get these lines when I playback on my DVD player?

bb
12th April 2003, 17:11
Originally posted by jggimi
The consensus is that SVCDs / DVDs should not be deinterlaced.
It depends:

- If your source is not "truely" interlaced, but phase-shifted only, then the best solution is to restore the orginal progressive frame (e.g. Decomb's Telecide()), no matter if your target format is DivX, SVCD, or DVD. Hollywood movies are a good example.

- If your source is "truely" interlaced (that means you have a time difference between each field like when filming with a DV camcorder), then it's not possible to restore progressive frames without loss.
-- For SVCD you should always deinterlace, because the overall quality will come out significantly better, although MPEG-2 is used on SVCDs, which supports interlaced encoding.
-- For DVD you should not deinterlace. There's enough bitrate available to encode in interlaced mode, so that you can avoid the quality loss of the deinterlacer.

Just for completeness (as I wouldn't recommend to create VCDs to anybody): Even for VCDs you'll get better quality if you don't just drop one field, but capture in full PAL resolution, then deinterlace and downsize to 352x288 (I'd use a bilinear filter here, like SimpleResize). Of course this is more time consuming than simply capturing at 352x288, but you get much better quality (for what you can get on VCD). Because VCDs use MPEG-1 you there's no interlaced encoding available.

bb

^^-+I4004+-^^
12th April 2003, 19:39
>For SVCD you should always deinterlace, because the overall quality will come out significantly better, although MPEG-2 is used on SVCDs, which supports interlaced encoding

as i'm constantly repeating;
deinterlacing will make video
jerky:
it'll also blur image some more...

bb how do you react?

i would opt for using that mpeg2 possibility
of interlaced encoding.....
anyhow,make good capture,and then
see what results you like best;
deinterlaced(jerky and blurred somewhat)
or
interlaced encoding (smooth motion,but more blocks because of
lo bitrate)

pick what you like the best...
(i DO deinterlace all my stuff,but because i don't have dvdplayer
OR recorder...i would need both take real advantage of interlaced
mpeg2....SVCD is not teh real thing,but..the only one that gives interlaced on 1 cdr)

Ookami
12th April 2003, 19:48
It depends on what bitrate you use for SVCD. SVCD is not equal SVCD...

So, if you're always going for max bitrate and don't care how much you gonna put on one CD then I would vote for not deinterlacing.
Deinterlace only if there's some issue like blended fields etc. .

Deinterlace only if you want to make such stuff like 1 CD SVCD movie (no standard stuff).

>Just for completeness (as I wouldn't recommend to create VCDs to anybody): Even for VCDs you'll get better quality if you don't just drop one field, but capture in full PAL resolution, then deinterlace and downsize to 352x288 (I'd use a bilinear filter here, like SimpleResize). Of course this is more time consuming than simply capturing at 352x288, but you get much better quality (for what you can get on VCD). Because VCDs use MPEG-1 you there's no interlaced encoding available.

Well, of course, that's obvious. Agree with all of your last points.

jggimi
12th April 2003, 19:51
Well, my SVCDs have been left with interlaced MPEG-2 video. But then, they are all at maximum bitrate (2520kbps). Testing both with and without deinterlacing led me to that conclusion.

---

Your mileage may vary. Objects in mirror may be closer than they appear. Some settling of contents may occur during shipping. All models over 18. Slightly higher in California.

bb
14th April 2003, 07:07
Originally posted by ^^-+I4004+-^^
bb how do you react?

Well, you've given the answer yourself: Due to the bitrate being too low you'll get blocking artefacts when encoding interlaced. Interlaced encoding needs up to 30% more bitrate for the same quality.

@all: I'm talking about SVCDs, not XSVCDs; the bitrate (video + audio) is limited to a little more than 2.700 kbps. If you prefer interlaced mode, CVD @ 352x576 (PAL) may be an option. The lower resolution will compensate for the higher bitrate needed for interlaced mode. Or you may wish to encode XSVCDs at bitrates like 4.000 kbps. However, both formats may fail to play on certain DVD standalones.

bb

Ookami
14th April 2003, 16:36
@bb

Blocking when using interlaced mode? Always? I am aware of the fact that interlaced video will "eat" more bitrate, but that's normal. If the picture would not highly degrade when deinterlacing I doubt that anyone would leave it interlaced :) .

On my sources I've only got blocking when encoding stuff like Futurama, but in that case even when deinterlacing... I stopped doing NR, so my TMPGEnc Pro has even more problems with my analogue sources, but only with Futurama I had blocking. But, now I'm encoding to the DVD standard so I encode nowadays very seldomly to SVCD.

IIRC always stayed in the specs and never encoded any of this XVCD etc. stuff (a few test clips aside).

But, I never encoded material from a DV camera (filmed, shaky etc.). IIRC, mb1 wrote that such material is not watchable on even max. SVCD bitrates. Do you mean maybe this stuff? Or do you have experienced blocking with every source you got when encoding interlaced? What about clean DVD/DVB etc. interlaced sources?

If you were a newbie I would say that you are doing something wrong, but as you are who you are, I am a bit confused. Maybe your analogue TV has even more interferences than mine? Or my eyes are even worse than I think, so I don't even see the blockings? Would be really interested in what you encode and how/with what.

Cheers,

Mijo.

bb
15th April 2003, 07:13
Ookami,

when I encode DVDs it's almost always progressive material (PAL country). So I can't tell you about these very clean sources; I simply had no need to encode interlaced.

My DVB sources (DVB-C) are hollywood movies most of the time, usually progressive, too (I can't even remember that I had to phase-shift once; I guess the TV broadcasters prefer to encode progressively, too, in order to save bandwidth).

Where interlaced encoding gets interesting is for DV or analogue TV sources. For DV you have much influence on how good a deinterlacer can do its job: Don't move the cam too fast, especially in the vertical direction, and use a tripod whenever possible. The big problem occurs where the picture moves more than two lines vertically from one frame to the next. In this case the deinterlacer doesn't work correctly - even Tom Barry's motion based deinterlacer doesn't scan very far (I think only one or two pixels in the neigbourhood). Of course I know about the problems deinterlacers introduce, but I still prefer to deinterlace because of the bitrate saving, and DV material usually needs a lot more bitrate than e.g. DVD sources.

As you cite mb1 (I'd like to know the source, by the way), I guess he didn't say this in general. You can - and should - improve your source using various filters, especially filters like Convolution3D which can greatly improve compressibility. I said already that I prefer to encode progressively, but I'm different from most DV people because my source is progressive most of the time (my camcorder supports progressive scan). My SVCDs made from DV sources are pretty watchable.

I did a few tests on TV caps, the lastest was on "Mission Impossible 1". I encoded interlaced and progressive SVCDs and CVDs. The interlaced encodes had many blocks in high motion scenes (remember the battle on top of the train with the helicopter behind). The CVD was clearly better, though, but the interlaced encodes were clearly outperformed by their progressive counterparts. Even the progressive SVCD looked better than the interlaced CVD in spite of the higher resolution. CCE 2.50 was used for these tests (of course with correct settings like zigzag for progressive and alternate scan for interlaced source). I don't remember the bitrate chosen, but it was calculated by FitCD aiming for two SVCDs.

Regarding the degradation of quality I'd say it depends on which deinterlacing method you use. I usually recommend Decomb's FieldDeinterlace or trbarry's motion compensating filter.

Regarding interlaced encodes I prefer to use CCE or TMPGEnc, but not Procoder. Procoder is known to produce great looking results on low-bitrate interlaced encodes, but because of the method it uses (field-based encoding) the resulting video is not fully MPEG-2 compliant, as far as I know, and it won't play on every standalone.

bb

DJ_TripleM
15th April 2003, 11:02
Today is the day of the program that I want to record in as high quality as possible. I'm aiming to convert to 2 SVCD's. I'll record it from tv, we have PAL-G in the Netherlands.

If I understand correctly, I will record at full PAL resolution in the best possible quality. Maybe I can put a sample of the capture on my webspace so you guys can give me advice as to what's the best way to encode?

Thanks again for all your help so far :)

Ookami
15th April 2003, 21:19
@bb

Thank you for your long posting.

-mb1

http://www.mb1.de.vu/
http://www.8ung.at/mb_1/mb1svcd/analogtest.htm
http://www.8ung.at/mb_1/mb1svcd/dv-test1.htm

etc. That's quite old material and mb1 would probably say that it's too outdated anyway :) . Hope you can find that quote there...

IIRC, he posted that quote (from my old posting) on Ultimate Board, but I may even read it on his site. Hopefully, I have not something mixed up.

-SVCD interlaced or not (from analogue sources)

Still I don't get it that I'm not getting any blocks with most of my encoded material, and you always get when you leave it interlaced. Many times I even lowered the SVCD bitrate to a lower one than the max. without any blocking. I only use TMPGEnc Pro as it's the best MPEG2 encoder that I have a license for (the Premiere MPEG output is not even near etc.).

>Regarding the degradation of quality I'd say it depends on which deinterlacing method you use. I usually recommend Decomb's FieldDeinterlace or trbarry's motion compensating filter.

Yes, it depends. But there is always a degradation, so it', again, a degradation if you deinterlace. But, like always, you have to decide for yourself and compromise.

@DJ_TripleM

Why don't you try it for yourself? Encode 1 part interlaced and the other not... And then choose what you like most, like all people do :) .

Cheers,

Mijo.

bb
16th April 2003, 06:51
Originally posted by Ookami
-SVCD interlaced or not (from analogue sources)

Still I don't get it that I'm not getting any blocks with most of my encoded material, and you always get when you leave it interlaced. Many times I even lowered the SVCD bitrate to a lower one than the max. without any blocking. I only use TMPGEnc Pro as it's the best MPEG2 encoder that I have a license for (the Premiere MPEG output is not even near etc.).
Mijo,

blocks are not the only form of quality degradation. You can easily avoid them without changing the bitrate by lowering the resolution or smoothing the video. So you could e.g. use lower smoothing thresholds when encoding progressive video. Smart deinterlacers (not the SmartDeinterlacer only) blur only small regions of the video where the combs occur, whereas with stronger smoother settings you affect the whole video.

To illustrate my point this may help (you may disagree with the numbers):
- Progressive encoding: 5% - 10% quality loss due to deinterlacing
- Interlaced encoding: 30% quality loss due to MPEG-2 interlaced mode

If you ask me, interlaced modes should be banned. I don't like all the hassle you get with this outdated, old-fashioned mode.

Another point is playback on progressive devices:
E.g. if you want to play your SVCD on a computer monitor, there's no problem if it's progressive, but if it's interlaced, you'll get further quality degradation due to realtime deinterlacing during playback (or you live with the combing, which is even worse).

bb

Ookami
16th April 2003, 08:51
@bb

I am aware that blocking is not the only way of degradation and that NR filters can improve compressibility etc. I have done everything from applying so much filters that I've encoded on 0.5 FPS to stressing the encoders so much that you can only see blocks mixed with some colors ;-) . And the avoidance of blocking with NR and reducing the bitrate doesn't apply in this case as I, as stated, use only the in-spec SVCD stuff. Also, the only blur thing I use is the TMPGEnc Pro "soften block noise" (mostly on ~15) etc., no additional NR filters. After all, until last year I was primary a Divx ;-) /DivX/SBC/VCD etc. user, SVCD I started quite late compared to my dwelling on these boards, so deinterlacing was just normal. Ok, just to say where I come from.

Also, I'm aware how the different algos work (deinterlacers), but I still remember having a hard time to get the same results with most of them (back in the time i used the good old Area based one, a long time I even followed Hawk's hints from the very old capture guide). Not to mention Donald Grafts smart deinterlacer with about 1856954082606 (+/- 3,54 :D ) different combinations :) .

I agree that deinterlacing is not always a "bad" thing, even Avery Lee did put a rather crude deinterlacer in VD and it still can give satisfying results in many cases. It's like I said hundreds of times always a compromise between your tastes/time...

As far as the hassle, that's the main reason why I stopped doing NR, mostly. Many things that I've (many times, ruined, also) encoded with NR filters are compressed to max. 3 CD Divx, 3 CD SVCD etc. so if I look at the quality of the encode it's mostly just not worth it to tweak the filters for hours, to encode dozens of test clips etc. etc.... The end encode ruins the picture anyway, and it's not worth the hassle of doing NR. When I start to digitize my favorite material from (S)VHS/TV -> DVD-5 I will, again, use all your (and some of mine ;) ) knowledge to achieve the perfect encoding. With many of the end formats it's just not worth it, IMO. Altough, IMO, you can gain much more compressibility/quality if you use NR etc. on LQ formats like DivX -> 1 CD than doing a DVD encode, but... :)

All the other points you mention are valid and I agree with them (maybe not some numbers, but your probably more experienced with SVCD, anyway), I also have to say that my lazyness and the mentioned reasons are probably also one of the reasons why I stopped doing any changes of most of my material. Garbage in = Garbage out (ok, when I get married my wife will carry the garbage, if she's going to listen :D ).

As a end note to this long winded posting, I just want to say that your generalisation "interlaced encode = blocking/LQ" has made me to inquire further and if I'm going to do many SVCD test encodes these days, your to blame. :)

Ok, sorry for being so long but it seems its one of these days.

All the best and thank you for your nice explanations and postings.

Cheers,

Mijo.

bb
16th April 2003, 10:08
Originally posted by Ookami
if I'm going to do many SVCD test encodes these days, your to blame. :)
Please don't tell your girl friend - I've just had an argument with my wife about sitting in front of my computer instead of helping her in the garden... :D

bb

^^-+I4004+-^^
16th April 2003, 14:35
a few of my thoughts,as (by now) everybody is wondering "why is ivo keeping a low profile if interlacing is mentioned"....well,not anymore....

as far as i understood mpeg2 interlaced encoding i would think that 30% is a bit too much (don't believe loss is that big)_ mpeg2 encoder should ;
-> take interlaced frame,split it to 2 fields,and encode fields separately (from that moment,fields are separated all the way,only to be merged in tv-viewer's eye.....)...ie. it's encoding fields as an 16x8 blocks...(if i remember correctly...read that somewhere long time ago)

if it didn't used this trick,mpeg2 interlaced encodings might look much worse (those combs are pretty hard to encode)..it's probably less effective in this mode,but not 30% worse!also,it depends how do you watch it;i have done some svcd experiments with sport,and they look like shit on monitor,while are perfectly acceptable on tv-out from normal viewing distance....i can hardly see ANY blocking when i watch on tv-out (as i don't come to 20cm from the screen and search for blocks)..but i regard video shouldn't be watched on monitor anyhow......

for me svcd is such a waste of space....better do 480x576 (deinterlaced) mpeg4 of some sort....2500kbit? mpeg4 looks like a champ on that bitrate...and on much higher resolutons too...

>If you ask me, interlaced modes should be banned. I don't like all the hassle you get with this outdated, old-fashioned mode

now,this is true,but only because my tv-out has problems outputting interlaced content (hehe)......and now i'll pose some considerations on "interlaced or deinterlaced video" (you all know this by now...i hope) :
for still scenes,there's no gain in keeping them interlaced (you might even do deinterlacing of such scene,with with some vintage deinterlacer that knows nothing about detecting combing in particular parts of image etc.,and still it will look OK-> for still scenes interpolation deinterlacing will perfectly match original content,and there should be no loss........off course,newer concept of not touching this parts AT ALL is much better...)
and for motion scenes interesting thing happens; the effective resolution of interlaced system is split to half when hi-motion
occurs (every field carries unique information in that case..and every field has only 288 lines ie. resolution is halved,and they'll get mixed in your brain anyhow...simillar like deinterlacers mix them....fast motion gets blurred-end of story) so you see less details anyhow......then,the only reason to keep something interlaced would be to preserve fluid motion...after this,i ask myself,do i really need motion THAT fluid or not (for example i have done deinterlacing on car rally racing and on moto races...they don't jerk that much at all..infact,they look fine even though 1 field is actually interpolated and has completely wrong information)...and i don't do much sports anyhow....

also,consider this; very fast motion frames are blurred on film too ; so where's the big disadvantage of deinterlacing ? must admit i don't see it! the biggest destroyer of details in my mview must be resizing which acts like a decent blur filter the more you downsize....
you want the smoothest motion you can get?yeah,well i don't..(hehe)

after all this i must agree with bb that image is not that much destroyed after deinterlacing as some might think; i mean come on.... can you really tell the time from the wrist watch of a golfer swinging his arm to make a shot (hit?birdie?).....
i doubt it....
(also:it's not fun zooming into interlaced video..hehe)
and the progressive systems are coming wether we like it or not (my poor VHS collection...hehe)

/ivo (completely deinterlaced)

bb
16th April 2003, 15:19
Originally posted by ^^-+I4004+-^^
mpeg2 encoder should ;
-> take interlaced frame,split it to 2 fields,and encode fields separately (from that moment,fields are separated all the way,only to be merged in tv-viewer's eye.....)...ie. it's encoding fields as an 16x8 blocks...(if i remember correctly...read that somewhere long time ago)
Actually there are two ways of encoding interlaced MPEG-2. The method you outline above is supported by Canopus Procoder only as far as I know. All other encoders do not use this field-based approach (which isn't compatible with all DVD players by the way). Have you ever thought about what this weird "alternate scan" is all about? Well, if not, read it up :p
.but i regard video shouldn't be watched on monitor anyhow......
There will be more progressive TV sets in the future.
for me svcd is such a waste of space....better do 480x576 (deinterlaced) mpeg4 of some sort....2500kbit? mpeg4 looks like a champ on that bitrate...and on much higher resolutons too...You're getting off-topic here, I think. MPEG4 can't be played on most standalones.

for still scenes,there's no gain in keeping them interlaced (you might even do deinterlacing of such scene,with with some vintage deinterlacer that knows nothing about detecting combing in particular parts of image etc.,and still it will look OK-> for still scenes interpolation deinterlacing will perfectly match original content,and there should be no loss........off course,newer concept of not touching this parts AT ALL is much better...)
Still scenes don't need any deinterlacing at all. There are no combs to remove.

the only reason to keep something interlaced would be to preserve fluid motion...after this,i ask myself,do i really need motion THAT fluid or not
There are people out there who can tell the difference. I'm satisfied with 25fps, and most people are satisfied with 24fps in cinema.

Cheers,
bb

^^-+I4004+-^^
16th April 2003, 21:23
>Have you ever thought about what this weird "alternate scan" is all about? Well, if not, read it up

probably didn't found much info in CCE .pdf...,but yes,there should be some purpose for it...(heh)

>There will be more progressive TV sets in the future.

yes,i was thinking about small size of monitor screen

>Still scenes don't need any deinterlacing at all. There are no combs to remove.

yes,but i have posed such example to show that "crude old" deinterlacing won't destroy still scenes because in stil scenes the other field can be "calculated" with the interpolation in the nice way....(to show that still scenes are not a problem anyhow...for old or new deinterlacers (which automatically "shut-off" when detect no-combing still scene))


>There are people out there who can tell the difference. I'm satisfied with 25fps, and most people are satisfied with 24fps in cinema.

it has less to do with framerate and more to do with substitution of a field with something it's not (interpolated field is not the "real" field with another timeline information).ie. this is what makes deinterlaced video source jerk on camera pans....