NTSC DVD Encoding Anomaly

[Christheminecraft47]

I have 2 DVDs I'm going to discuss today.
SpongeBob SquarePants: Complete 2nd Season (Disc 3), and SpongeBob SquarePants: Christmas
Both are Region 1, NTSC DVDs

One issue with the "Season 2" DVD is that most of the episodes have incorrect color composition. For Example, in the episode "Procrastination", on the "Christmas" DVD, SpongeBob looks normal. On the "Season 2" DVD, SpongeBob looks rather pale and doesn't accurately depict how it was originally aired, unlike the "Christmas" DVD.

Another issue here, which I would like to talk about, is the encoding.
I am going to use "Avidemux" and the raw VOB files from the discs to demonstrate this scenario.
All of the frames on the "Season 2" DVD are marked as "TFF" or "Top Field First", which is to be expected from a DVD source.
However, on "Christmas" DVD, some frames are "TFF", most frames are "FRM" which is just a progressive frame, which doesn't seem normal to me.
Not only that, but the playback on the "Christmas" DVD is choppy compared to the "Season 2" DVD, whereas it is rather consistent.

I would rather preserve copies with the correct color composition so I could deinterlace them myself, but I can't work with such wonky scenarios. I will leave some images below so you guys can take a look at this phenomenon. If you can, please let me know what is going on, why is this happening, what can I do to counteract this, etc.? If you guys need me to upload anything else helpful, please let me know. Thanks!


[Attachment 92684 - Click to enlarge]

[Attachment 92685 - Click to enlarge]

[Sharc]

Best guess: Your NTSC DVD is probably standard 2:3 telecined. So one should inverse-telecine it to obtain the progressive frames rather than just copy and pack into an .mkv container.
Look for a filter in your GUI (Avidemux) called 'IVTC' or 'detelecine' or similar (I don't use Avidemux).
Or upload short snippets (few seconds of a scene with motion) of your .vobs under question so someone may take a closer look.

[Christheminecraft47]

Keep in mind that the animation is 23.976fps. However, the car scene and the transition at the end are true 29.97i.
Pay close attention to SpongeBob, and the motion of the car and the transition after the scene when comparing the videos below.
I used OBS to capture VLC at 720x540@59.94fps for the best results. You may need to download the videos if they don't play here.
I have labelled the videos for easy comparison.

https://forum.videohelp.com/attachment.php?attachmentid=92688&stc=1&d=1781367003
https://forum.videohelp.com/attachment.php?attachmentid=92689&stc=1&d=1781367003
https://forum.videohelp.com/attachment.php?attachmentid=92690&stc=1&d=1781367023
https://forum.videohelp.com/attachment.php?attachmentid=92691&stc=1&d=1781367023

[jagabo]

All of the frames on the "Season 2" DVD are marked as "TFF" or "Top Field First", which is to be expected from a DVD source.

It's most common but not always the case.

However, on "Christmas" DVD, some frames are "TFF", most frames are "FRM" which is just a progressive frame, which doesn't seem normal to me.

Mixed hard and soft telecine is not uncommon.

Not only that, but the playback on the "Christmas" DVD is choppy compared to the "Season 2" DVD, whereas it is rather consistent.

Cartoons often have shots sped up or slowed down after being telecined, to better match the sound track.

Getting perfect motion probably won't be possible. And to come close will probably require going through each episode frame by frame and making manual adjustments.

I would rather preserve copies with the correct color composition

The colors are a bit tricky. There is a slight brightness difference between the left and right images. The left image is more saturated overall -- except for the yellows. The right image has really pumped up yellows. I suspect the left image is closer to correct. But here's the left and right images after adjusting the left image to more closely match the right:

[Attachment 92693 - Click to enlarge]

[Christheminecraft47]

Cartoons often have shots sped up or slowed down after being telecined, to better match the sound track.

Getting perfect motion probably won't be possible. And to come close will probably require going through each episode frame by frame and making manual adjustments.

Sure, but did you take a look at the car scene and transitions I mentioned earlier (look at the reply from earlier)? On the "Season 2" DVD, it looks as expected. Interlaced, and everything. On the "Christmas" DVD, it is rather butchered compared to the "Season 2" DVD, and I don't think it would make any sense as to why this would be the case. Pay close attention to all the videos in terms of motion and see if you can spot the difference.

[Sharc]

I used OBS to capture ...

"Captures" from DVD? Creates some doubts here. Are these screen captures? or DVD rips? As said before for getting better advice you should upload snippets of the original .vobs from your DVDs, not OBS "captured" (or processed) .mkv stuff. Maybe you should revisit your "capture" process.

[Christheminecraft47]

"Captures" from DVD? Creates some doubts here. Are these screen captures? or DVD rips? As said before for getting better advice you should upload snippets of the original .vobs from your DVDs, not OBS "captured" (or processed) .mkv stuff. Maybe you should revisit your "capture" process.

VTS_06_02.VOB, "Christmas" DVD, Timestamp = 00:06:51 - 00:07:12
VTS_01_02.VOB, "Season 2" DVD, Timestamp = 00:00:38 - 00:01:00

Here are the original VOB files from the respective DVDs and the timestamps that correspond to what I mentioned in this post. These VOB files are directly from their respective Discs, so you can't get any more proof than this.

https://drive.google.com/drive/folders/1hhf54Mg6yGnWaVzCRXEX5DvgbUee71y-?usp=sharing

[Christheminecraft47]

The link has been updated to allow access for everyone without the need for requests, so you guys should be able to access the VOBs without trouble now.

[Sharc]

The VTS_06_2.VOB is standard 2:3 telecine, not interlaced. IVTC (=inverse telecine) it to get progressive frames at 23.976 fps, with smooth and steady jerk-free motion of the car scene, for example. Nothing unusual.
In Avisynth:

LWLibavVideoSource("VTS_06_2.VOB")
TFM().TDecimate()

[Sharc]

... and the VTS_01_2.VOB seems to be a mix of TFF true interlaced (with occasional frame drops) like for the car scene, and 2:3 hard telecine for the rest. A bit messy IMO. It's propably best to view it using double-rate deinterlacing

LWLibavVideoSource("VTS_01_2.VOB")
AssumeTFF()
QTGMC() #or a deinterlacer of your choice

And yes, the colors look a bit off/pale ....

[Christheminecraft47]

I have deinterlaced and color corrected VTS_01_2.VOB, and the results are pretty spot-on.
The colors look a lot closer to the "Christmas" DVD, and the car scene is fully intact without any hiccups. Nonetheless, let me know what you think.

I used VirtualDub2 to deinterlace using Bob 2x TFF. Then, I used Avidemux's color balance filter to simply increase the saturation of the highlight up to full blast.

[jagabo]

Yes, there's a mix of 12p with duplicates/pulldown, 24p with pulldown, 30p with pulldown, 30i (60 different fields per second video), some missing frames (shots that were sped up by decimating frames), etc. The easiest thing to do is use QTGMC to make 59.94 frames per second. That will play like the original DVD on a 60Hz display. Otherwise you could treat the different frame rate sections separately and produce variable frame rate video. But that will be a lot of work. You could also decimate everything down to 24p, leaving the 30p and 30i sections jerky.

Oh, this makes VTS_01_2 colors match VTS_06_2 more closely:

# v1 is VTS_01_2.VOB

v1 = v1.ColorYUV(gain_y=-5, cont_u=-35, cont_v=-35)
hmask = v1.MaskHS(StartHue=160, EndHue=170, MinSat=30, MaxSat=80).BilinearResize(v1.width, v1.height)
v1 = Overlay(v1, v1.Tweak(hue=-2, sat=1.7), mask=hmask)

As noted earlier VTS_01_2 is generally more saturated. Except for yellows, which are much more saturated.

[Sharc]

..... Nonetheless, let me know what you think.
I used VirtualDub2 to deinterlace using Bob 2x TFF.

Your version of post#11 suffers from many ugly blends and residual interlacing artifacts. Worth to learn Avisynth and apply a decent deinterlacer. Or I think there are better deinterlacers available in Vdub than what you have been using.

(Or maybe it got messed up with your subsequent color corrections in Avidemux - I don't know)

[Christheminecraft47]

Your version of post#11 suffers from many ugly blends and residual interlacing artifacts. Worth to learn Avisynth and apply a decent deinterlacer. Or I think there are better deinterlacers available in Vdub than what you have been using.

(Or maybe it got messed up with your subsequent color corrections in Avidemux - I don't know)

That's because it was trying to "deinterlace" the animated sequences, which are 23.967p material. When it comes to 29.97i material, like the car scene, it is flawless. When it comes to mixed content like this, it's best to just deinterlace it this way. Plus, you don't really notice it when you're actually watching it.

[Sharc]

Your version of post#11 suffers from many ugly blends and residual interlacing artifacts. Worth to learn Avisynth and apply a decent deinterlacer. Or I think there are better deinterlacers available in Vdub than what you have been using.

(Or maybe it got messed up with your subsequent color corrections in Avidemux - I don't know)

That's because it was trying to "deinterlace" the animated sequences, which are 23.967p material. When it comes to 29.97i material, like the car scene, it is flawless. When it comes to mixed content like this, it's best to just deinterlace it this way.

As I wrote use a decent deinterlacer and you won't get this blended crap and still get the car scene right. I gave you the Avisynth script before. It's up to you of course what you believe is right for you, but at some point your processing is questionable IMO - unless you prefer blended frames and jaggies ....

[Christheminecraft47]

LWLibavVideoSource("VTS_01_2.VOB")
AssumeTFF()
QTGMC() #or a deinterlacer of your choice

How would something like QTGMC compare to Bob 2x?

[Sharc]

I will post a comparison when I am back in an hour or so

[Sharc]

So here a side-by-side comparison. Step through the frames and you will see the blends and jaggies (e.g. on black lines) on the left picture.
Note that I didn't apply a color correction for the QTGMC version as it is irrelevant for this comparison.
As I said, it is also a matter of personal preference whether one prefers or hates blends. So no criticism, just a hint.

[Christheminecraft47]

As I said, it is also a matter of personal preference whether one prefers or hates blends. So no criticism, just a hint.

Yes, I do appreciate the hint, but respectfully, I'm going to stick to Bob 2x. Yes, it can cause jaggedness, but that's just because of how the algorithm works and how interlaced content works. We could argue about this all day, but I value my time, and I'm sure you do as well.

And with the color correction settings, the best way to go would be to deinterlace and color correct the episodes from the "Season 2" DVD, because it has everything the way it was originally produced (albeit the color comp, but no big deal ), rather than being screwy like the "Christmas" DVD. I am aware that these episodes are on other compilation DVDs made by Nickelodeon/Paramount, but I suspect they suffer the same fate as the "Christmas" DVD. Doesn't make sense why they would make the frames all weird, but they probably don't care because they suck (lol).

So, after the results we've seen here, I rest my case in this scenario. However, I do have another case of an "Anomaly" to share.


Exhibit B: In the 2003 (Region 1, NTSC) DVD release of "The Brave Little Toaster (1987)", we have some weird frame encoding going on again.
Using Avidemux, we can see at the start of the film, the very first frame is "FRM", which is progressive. Next frame, it's "BFF", which is bottom field first, then back to "FRM", then it goes to "TFF" (Top Field First), then "FRM". And this pattern repeats for the ENTIRE MOVIE, making it difficult to IVTC, since it is dependent on one specific field order (doesn't let me autodetect) and it is film material (24fps), and makes playback noticeably choppy at certain points. If anyone has a clue about this conundrum, please let me know. Thanks!


[Attachment 92722 - Click to enlarge]

[jagabo]

Exhibit B: In the 2003 (Region 1, NTSC) DVD release of "The Brave Little Toaster (1987)", we have some weird frame encoding going on again.
Using Avidemux, we can see at the start of the film, the very first frame is "FRM", which is progressive. Next frame, it's "BFF", which is bottom field first, then back to "FRM", then it goes to "TFF" (Top Field First), then "FRM". And this pattern repeats for the ENTIRE MOVIE, making it difficult to IVTC, since it is dependent on one specific field order (doesn't let me autodetect) and it is film material (24fps), and makes playback noticeably choppy at certain points. If anyone has a clue about this conundrum, please let me know. Thanks!

The "fix" is trivial in AviSynth:

Mpeg2Source("filename.d2v") # be sure to use DgIndex in video -> Honor pulldown flags mode
TFM()
TDecimate()

The output is 24p. Of course the 30i and 30p sections will be jerky from the missing frames.

[Christheminecraft47]

The output is 24p. Of course the 30i and 30p sections will be jerky from the missing frames.

Would the 24p output be jerky because of those frames?

[Sharc]

Yes, it can cause jaggedness, but that's just because of how the algorithm works and how interlaced content works. We could argue about this all day, but I value my time, and I'm sure you do as well.

Really? Not to argue, but you have been asking for comments on your deinterlaced version of post#11, and I won't waste any more time other than perhaps recommending to review what (true) interlace, progressive, hard-/soft telecine and pulldown really mean and how to deal with it to gain a better understanding of your "NTSC anomalies" and issues like motion jerkyness, blends, jaggies and other strange "conundrums". Sometimes not straightforward especially with mixed content and anime.
Have fun!

[Sharc]

Exhibit B: In the 2003 (Region 1, NTSC) DVD release of "The Brave Little Toaster (1987)", we have some weird frame encoding going on again.
Using Avidemux, we can see at the start of the film, the very first frame is "FRM", which is progressive. Next frame, it's "BFF", which is bottom field first, then back to "FRM", then it goes to "TFF" (Top Field First), then "FRM". And this pattern repeats for the ENTIRE MOVIE, making it difficult to IVTC, since it is dependent on one specific field order (doesn't let me autodetect) and it is film material (24fps), and makes playback noticeably choppy at certain points. If anyone has a clue about this conundrum, please let me know. Thanks!

The "fix" is trivial in AviSynth:

Mpeg2Source("filename.d2v") # be sure to use DgIndex in video -> Honor pulldown flags mode
TFM()
TDecimate()

The output is 24p. Of course the 30i and 30p sections will be jerky from the missing frames.

Yes. I think same result to be expected using the Libav source filter like

LWLibavVideoSource("VTS_01_2.VOB")
TFM()
TDecimate()

[jagabo]

Yes, LSMASH (LWLibavVIdeoSource) performs soft pulldown by default so it should give the same result.

[hello_hello]

I''d be going the variable frame rate route myself. I created these with Avisynth, but if you don't use it, Handbrake should be able to encode them as VFR, and it'd probably have a decent de-intelacer.
The one you want to colour correct would probably have to be encoded as VFR with HandBrake, outputting a lossless file you could then color adjust and re-encode with Avidemux.

I could only encode 18 minutes of VTS_06_2.VOB for some reason. I didn't try very hard to work out why. Maybe the vob file became corrupted at some stage.

[Christheminecraft47]

I''d be going the variable frame rate route myself. I created these with Avisynth, but if you don't use it, Handbrake should be able to encode them as VFR, and it'd probably have a decent de-intelacer.
The one you want to colour correct would probably have to be encoded as VFR with HandBrake, outputting a lossless file you could then color adjust and re-encode with Avidemux.

I could only encode 18 minutes of VTS_06_2.VOB for some reason. I didn't try very hard to work out why. Maybe the vob file became corrupted at some stage.

Ok, first of all, I don't use Avisynth or Handbrake
Secondly, I would rather not deal with VFR because of the result it has shown. When I throw both into Avidemux, they report 25fps on the dot. The smooth motion in VTS_01_2's car scene is butchered because of this, for one.
Besides, in case you haven't read, I've already devised a solution before, and I have moved onto another case of encoding issues. Thus, I won't be using VFR because what I can do with what I have works efficiently when it comes to the SpongeBob Season 2 Material. But, thanks for the tip.

[Christheminecraft47]

Yes. I think same result to be expected using the Libav source filter like

LWLibavVideoSource("VTS_01_2.VOB")
TFM()
TDecimate()

Ok, if the video would still be choppy, then I do have an idea that could possibly work. I could just simply record it directly from my DVD player since it's able to play the movie much smoother, because it would be better at handling such encoding.

[jagabo]

I took a closer look at the two videos. There _is_ an issue with the the motion within the VTS_01_2 file. Because of the odd pulldown pattern the 12 fps character animation often results in more jerkiness. Normally the 12 fps character motion would result in motion (after IVTC) at every other frame, a sequence like A A B B C C D D -- every other frame is a duplicate. But in this file you often get a sequence like A B B B C D D D -- two unique frames, and two duplicates. You can see it in the attached clip if you step through frame by frame.

[Christheminecraft47]

I took a closer look at the two videos. There _is_ an issue with the the motion within the VTS_01_2 file. Because of the odd pulldown pattern the 12 fps character animation often results in more jerkiness. Normally the 12 fps character motion would result in motion (after IVTC) at every other frame, a sequence like A A B B C C D D -- every other frame is a duplicate. But in this file you often get a sequence like A B B B C D D D -- two unique frames, and two duplicates. You can see it in the attached clip if you step through frame by frame.

Yeah, that does seem to be the case. But, it when you actually watch it, it's negligible because it's not that noticeable.

Plus, since VTS_01_2 is encoded as interlaced and has the car scene untouched in its original 29.97i format, I would just simply deinterlace VTS_01_2, and it would also be negligible for the same reason. Plus, it would be easier to just deinterlace the content from the "Season 2" set (Not just VTS_01_2) and color correct them.

[hello_hello]

Ok, first of all, I don't use Avisynth or Handbrake

It's probably lucky then that Avidemux is doing the job perfectly.

Secondly, I would rather not deal with VFR because of the result it has shown. When I throw both into Avidemux, they report 25fps on the dot. The smooth motion in VTS_01_2's car scene is butchered because of this, for one.

Reverse placebo effect? For VFR video, editors often decode at the average frame rate, or they just make something up. In this case it seems to be the latter, because if you open those files with MediaInfo, it'll report an average frame rate of 24.108 fps for VTS_01_2, and 24.621 fps for the other. I even stepped through a fair section of the car crash one frame at a time and your earlier sample, which you de-interlaced to 60fps, has exactly twice the number of frames as the same section in my sample, which was deinterlaced to 30fps. Can 60fps look smother than 30fps? Sometimes, but it's definitely not butchered.

Besides, in case you haven't read, I've already devised a solution before, and I have moved onto another case of encoding issues. Thus, I won't be using VFR because what I can do with what I have works efficiently when it comes to the SpongeBob Season 2 Material. But, thanks for the tip.

I'll try to live with the disappointment.
Each to their own. I just thought I'd suggest another option. Have fun!