What's the best deinterlacing method for VHS video for burning to DVD?

[DB83]

What about the 'AV-in NR' ? I guess that NR refers to Noise Reduction.

Even so, I would expect an external filter to be more effective. And, as stated, noise will impact the ability to compress and use lower bitrate and thus have more video on a dvd.


Might well do some experimenting of my own, with other software, later.

[Ferrari420]

If a DVD has progressive video on it, will it play in progressive with a DVD player connected to the TV via RCA?

[Ferrari420]

What about the 'AV-in NR' ? I guess that NR refers to Noise Reduction.

Even so, I would expect an external filter to be more effective. And, as stated, noise will impact the ability to compress and use lower bitrate and thus have more video on a dvd.


Might well do some experimenting of my own, with other software, later.

I tried MSU Denoiser and I decided not to use it because it makes the pixels around the denoised area wobble which is really distracting. There's an outline around people's faces with random pixels. And it ruined one video with streaks of blue and green flashing on the wall.

[Sharc]

What about the 'AV-in NR' ? I guess that NR refers to Noise Reduction.

Even so, I would expect an external filter to be more effective. And, as stated, noise will impact the ability to compress and use lower bitrate and thus have more video on a dvd.


Might well do some experimenting of my own, with other software, later.

I tried MSU Denoiser and I decided not to use it because it makes the pixels around the denoised area wobble which is really distracting. There's an outline around people's faces with random pixels. And it ruined one video with streaks of blue and green flashing on the wall.

Upload an example so someone may take a look. Maybe something went wrong. It shouldn't be that bad from your capture .....
(Anyway I would recommend to spend your time with learning Avisynth basics rather than fiddling too much with all kind tools.)

[Sharc]

If a DVD has progressive video on it, will it play in progressive with a DVD player connected to the TV via RCA?

By "RCA" you mean Composite I assume?
Usually the DVD player sends field by field to the TV. The point is that for progressive video on the DVD (PsF) the 2 fields are from the same moment in time which is typical for Film material. Your VHS capture is interlaced though with the fields presenting different moments in time. Keep it like that for DVD. Or forget mpeg2 DVD and deinterlace to 50fps and encode mpeg4 AVC (x264, .mp4 or .mkv).

[DB83]

As an experiment, I attach two slightly different versions in dvd form of the sample clip. 'Pushed' the bitrate so one could get approx 2 hours on a single-layer disk.

Which one is best ?

[Sharc]

As an experiment, I attach two slightly different versions in dvd form of the sample clip. 'Pushed' the bitrate so one could get approx 2 hours on a single-layer disk.

Which one is best ?

I had a quick look at DVD22 only. It's flawed IMHO.
- Field order is wrongly flagged (it's correctly encoded as TFF though)
- There is a problem with the Chroma. It advances with every second field only. Something went wrong. Which encoder did you use? It seems to have handled YV12 interlaced incorrectly. The chroma flaw is eye-catching.
- Bitrate is too low IMO. Blocking artifacts and smeared details. Maybe Half D1 (352x576) would have helped a bit.

Edit:
DVD21 suffers basically from the same deficiencies. It is flagged progressive, TFF .... The stream is interlaced TFF though.

[Ferrari420]

If a DVD has progressive video on it, will it play in progressive with a DVD player connected to the TV via RCA?

By "RCA" you mean Composite I assume?
Usually the DVD player sends field by field to the TV. The point is that for progressive video on the DVD (PsF) the 2 fields are from the same moment in time which is typical for Film material. Your VHS capture is interlaced though with the fields presenting different moments in time. Keep it like that for DVD. Or forget mpeg2 DVD and deinterlace to 50fps and encode mpeg4 AVC (x264, .mp4 or .mkv).

Yes composite. Will my progressive DVD play with two fields or convert back to interlaced and play back as interlaced because it's connected via composite? I don't trust hardware deinterlacing and the result is better for me when it's immediately deinterlaced, and there's no delay when playing back.

[DB83]

I would take a look at DVD21 since that reported somewhat suprisingly. (The 'experiment' was to 'force' BFF on DVD22)

BTW How do you determine field-order other than mediainfo which often does not even report it ? I always thought that avi defaulted to BFF whereas mpeg2 could be either.


And I believe both dvds used HCenc via avstodvd

[Sharc]

If a DVD has progressive video on it, will it play in progressive with a DVD player connected to the TV via RCA?

By "RCA" you mean Composite I assume?
Usually the DVD player sends field by field to the TV. The point is that for progressive video on the DVD (PsF) the 2 fields are from the same moment in time which is typical for Film material. Your VHS capture is interlaced though with the fields presenting different moments in time. Keep it like that for DVD. Or forget mpeg2 DVD and deinterlace to 50fps and encode mpeg4 AVC (x264, .mp4 or .mkv).

I don't trust hardware deinterlacing...

It's often better than mediocre or flawed SW deinterlacing.
Your TV or player will normally bob-deinterlace the interlaced DVD source, so you get 50fps temporal resolution.
When you deinterlace your video yourself you will end up with 25fps progressive on your DVD, because 50fps is not DVD compliant. Temporal resolution is lost, and you will get a 'stroboscopic' playback as has been shown before.

[Sharc]

I would take a look at DVD21 since that reported somewhat suprisingly. (The 'experiment' was to 'force' BFF on DVD22)

BTW How do you determine field-order other than mediainfo which often does not even report it ? I always thought that avi defaulted to BFF whereas mpeg2 could be either.


And I believe both dvds used HCenc via avstodvd

Even using GUIs you should do it right

How to determine the field order has been explained many times in this forum. I am just lazy to find the posts.
For video:

AssumeTFF()
separatefields()

Select a scene with motion. Step through the fields with your SWplayer. If the motion is continuous the field order is - as assumed - TFF. If the motion is jerky forward-backward-forward-backward ...... the field order is BFF. Don't trust MediaInfo in this respect.
You can do similar with VirtualDub.
(Telecined Film has a specific 3-2 pattern).

[Alwyn]

You can do similar with VirtualDub.

Open file, open filters, choose Deinterlace, select YADIF and Double Frame Rate, Top Field First. Step thru video. If you've chosen the incorrect field order, it will be obvious.

Analogue AVI is normally TFF. DV AVI is normally BFF.

[Sharc]

You can do similar with VirtualDub.

Open file, open filters, choose Deinterlace, select YADIF and Double Frame Rate, Top Field First. Step thru video. If you've chosen the incorrect field order, it will be obvious.

Analogue AVI is normally TFF. DV AVI is normally BFF.

And once we are at it:
Go to frame 41 and watch the back of the boy when you step through the video (right panel = bobbed picture in Vdub).
Do you see the white (decolored) line coming and going? This is the Chroma flaw I have been mentioning. The chroma does not advance synchronously with the luma at the fieldrate, but only at the framrate. May be the encoder, may be the settings, maybe the generated script, maybe the GUI ..... you name the culprit

[Alwyn]

Go to frame 41

That flaw's on your VOB, not on the original, @Sharc.

[Skiller]

Will my progressive DVD play with two fields or convert back to interlaced and play back as interlaced because it's connected via composite?

Yes, individual fields – no matter if the DVD is encoded progressive or interlaced.
This is because RCA/Composite is an analog continuous stream of fields; there exist no frames in Composite video.

However...

I don't trust hardware deinterlacing.

Unless your TV is really crappy or 15+ years old, it will likely be smart enough to detect that two consecutive fields are taken from the same point in time and switch the deinterlacing to weave (= none).
And even if not, the deinterlacing of TVs is not that bad, definitely not worse than Yadif. Being limited to 25i rather than 50p on DVD, and using RCA/Composite, you don't really have a choice. Pre-Deinterlacing does not help.

[Sharc]

Go to frame 41

That flaws on your VOB, not on the original, @Sharc.

Of course it's in the VOB. Did I say anything else? Btw. it's the VOB of DB83's encodes in his post #66 where he was asking for comments.
Ferrari's capture is ok, so it should just be processed and encoded correctly rather than being butchered by all means (base rate deinterlacing or interlaced with wrong field order flagging, inadequate noise filtering, faulty YV12 interlaced chroma processing, too low bitrate .....)
And Ferrari's video is 2hours + 29minutes. No chance with mpeg2 on a DVD5 720x576, even though using a decent encoder (like Canopus Procoder which is very good at preventing macroblocking) plus some noise filtering, up to 2 hours may be doable.

[DB83]

With the deepest of respect.

I use generally accepted s/w. What butchers it ? The s/w. The encoder. The play-back.


I did look for some feed-back on how to determine field-order (without resort to software) and I do find conflicting (as so often one finds with video) opinion.


Worry not. Later today i will post a sample that could be construed as TFF but is actually BFF (if general consensus is true and that is the core of the arguement here)

[Alwyn]

Of course it's in the VOB. Did I say anything else??

You didn't say what file you were looking at. There's mine, yours and 2xDB83s. We're not mind-readers. So, what exact file are you referring to? Is it DB83s 21?

so it should just be processed and encoded correctly rather than being butchered by all means.

That's a bit rich considering it's on your DVD and not mine. There is obviously a problem with your software workflow. Whatever process and encode you used doesn't appear to be very good if it produces that artifact.

[DB83]

Ok. So here is the last sample. A different clip and do tell me what the field-order is here.

[Sharc]

Of course it's in the VOB. Did I say anything else??

You didn't say what file you were looking at. There's mine, yours and 2xDB83s. We're not mind-readers. So, what exact file are you referring to? Is it DB83s 21?

DB83s 21 and 22. See also post #67

so it should just be processed and encoded correctly rather than being butchered by all means.

That's a bit rich considering it's on your DVD and not mine. There is obviously a problem with your software workflow. Whatever process and encode you used doesn't appear to be very good if it produces that artifact.

Maybe I missed something, but I have only seen your baserate deinterlaced choppy (I call it stroboscopic) 25fps variant of post#35. What you don't understand is that the YV12 4:2:0 interlaced chroma issue will not materialize for your progressive variant of course. If it exists in the interlaced variants something went wrong and should be corrected. 25fps baserate deinterlacing for home-video DVDs is no option for me.
But to each his own.

[Alwyn]

Ok. So here is the last sample. A different clip and do tell me what the field-order is here.

I reckon it's Progressive.

Why? Open it as I described in VDub and you will see, when stepping through:

1. There's no combing.

2. When one deinterlaces it, nothing changes except a still frame is inserted at every second frame. A deinterlaced video would give you movement on every frame (because a frame is created for every field).

[Alwyn]

If it exists in the interlaced variants something went wrong and should be corrected.

It's in yours.

[Sharc]

Ok. So here is the last sample. A different clip and do tell me what the field-order is here.

It's progressive video, means PsF = progressive segmented frames. Both fields from the same moment in time, typical for movie footage.
One can treat it as progressive for further processing.
Discussed and explained so many times ....

[Sharc]

If it exists in the interlaced variants something went wrong and should be corrected.

It's in yours.

I know, it's much less visible though, but it should be corrected as well.

Edit:
Here we go (maybe it was an issue with the beta release of DVDStyler. I reverted to the official release for this example)
Anyway I have no plans to investigate it in more detail as I am hardly doing any mpeg2/DVD stuff anymore.

[DB83]

Progressive ?

With the highest of respect. Do tell me that the program that created this is bad. But if it is so bad then noone should use it.


I actually made it easy for you inasmuch that I had used this sample before. For the video fed in to avstodvd was DV.avi (BFF). Nothing was changed.


In my mind, this kinda destroys any theory that you can tell if a video is TFF or BFF simply by checking it frame-by-frame. Not enough movement ? How much do you need ?


In essence I have read many comments that disagree with the method of determination. I just want a conclusive one.

[davexnet]

In my mind, this kinda destroys any theory that you can tell if a video is TFF or BFF simply by checking it frame-by-frame. Not enough movement ? How much do you need ?


In essence I have read many comments that disagree with the method of determination. I just want a conclusive one.

This is one method, using vdub by itself.
In vdub2, open the deinterlace filter and choose yadif, double frame rate (top or bottom).

Step through the video. If you see a backwards and forward motion where it should be smooth in one direction,
it's the wrong field order

sharc's dvd5.zip looks pretty good to me, however at that bitrate, the OP would probably have to go to a third disk
for 2 hours 30 minutes

[Sharc]

In essence I have read many comments that disagree with the method of determination. I just want a conclusive one.

You may have read, but did you understand? I really wonder that members who have been here for decades are still struggling with this basic video knowledge and testing methods
Don't trust what GUIs and tools report. That's a key learning. Thus stepping through the fields and trusting your eyes is really the only reliable method to draw conclusions about progressive, PsF, interlaced, field order, telecine, phase- or field shifted progressive video, orphaned fields etc.
In most cases it is easy to draw the right conclusion. There are cases though which let's one scratch one's head a bit more ....

[DB83]

Ok my friend. I never claimed to be an expert on all things video related (I am sure that an Australian will join in on this one)

But pls admit you got it wrong on this one or expand on it rather than palm it off on the tools. Or is a DV source NOT BFF ?


TBH when I started with video the software I used did not use the terms TFF or BFF but rather Field A and Field B. Was that easier to understand or not ?

[Sharc]

Ok my friend. I never claimed to be an expert on all things video related (I am sure that an Australian will join in on this one)

But pls admit you got it wrong on this one or expand on it rather than palm it off on the tools. Or is a DV source NOT BFF ?


TBH when I started with video the software I used did not use the terms TFF or BFF but rather Field A and Field B. Was that easier to understand or not ?

I am no expert either.
Field A, field B, top field, bottom field don't say anything about the temporal sequence (cadence). TFF for example says that the top field (or field A if you prefer) should be played before the other field (Bottom field or field B) in order to preserve the motion continuity.
AviSynth uses 'Top' and 'Bottom', 'Even' and 'Odd' in its syntax.

Going off topic I am afraid.

[DB83]

Ok. I'll concede on this one (apart from the DV comment)

I just went back in my time machine to the laste 90's when I was using Ulead capture software. It only supported mpeg1/2 or DV and they refered to these as Field B and Field A. So here we have a complete 'contradition' if one expected A to be on top of B . And that, initself, has probably clouded my thought process.


And it seems that many comments only relate to personal experience since I now understand that there is (apart from DV) no hard and fast rule - it all comes down to hardware etc.


Apologies to all for the off-topic diversion.