Any good reason NOT to deinterlace the HD restoration of MP Flying Circus?

[RasterEyes]

I just bought the HD restoration of Monty Python's Flying Circus... and wow! What a huge improvement over the DVD collection I got as a Christmas present many long years ago.

One surprise for me, however, is that after the creators of these discs did such a masterful restoration job: Why would they leave so much of the video clearly interlaced? To my eyes, at least, the additional step of deinterlacing using HandBrake makes the great restoration even better.

I used these settings:

Detelecine: Default
Interlace Detection: Default
Deinterlace: Yadif/Default

Would they leave the interlacing intact because they consider artifacts caused by deinterlacing worse?

Perhaps motivated by an archivist's desire to preserve more of the original experience?

[poisondeathray]

Blu-Ray does not support 1920x1080p50, so that's probably the main reason

"50p" is only supported at 1280x720 for BD. Maybe they like the "Full HD" marketing term (when it really isn't "Full HD" if interlaced - it's really 1920x540 fields)

Maybe in a few years they will release it on UHD-BD which supports 50p at higher resolutions, and they hope fans like you will buy it again

[RasterEyes]

Blu-Ray does not support 1920x1080p50, so that's probably the main reason...

My US-release version of MPFC is 29.97fps, not 25fps, so that wouldn't be a good reason to keep interlacing in this case.

The restoration was done using a mix of film where film was available and salvageable, and videotape sources. You'd think, MPFC being a British show, the videotapes would be at 25fps. But if I step through many scenes a frame at a time, every single frame is interlaced and also different from the previous frame while playing at 29.97fps.

I'd think that there would have to be some repeated frames because of this, or signs that a 25fps source had been slowed down to 24fps and processed with 3:2 pulldown like movie film (a common way to convert PAL-50 to ~30fps). It seems oddly, however, like a lot of the original videotape was ~30fps.

The segments converted from film (like the animated intro) show a mix of solid frames and interlaced frames, just like you get with telecine conversion of 24fps film to ~30fps.

[poisondeathray]

My US-release version of MPFC is 29.97fps, not 25fps, so that wouldn't be a good reason to keep interlacing in this case.

The restoration was done using a mix of film where film was available and salvageable, and videotape sources. You'd think, MPFC being a British show, the videotapes would be at 25fps. But if I step through many scenes a frame at a time, every single frame is interlaced and also different from the previous frame while playing at 29.97fps.

I'd think that there would have to be some repeated frames because of this, or signs that a 25fps source had been slowed down to 24fps and processed with 3:2 pulldown like movie film (a common way to convert PAL-50 to ~30fps). It seems oddly, however, like a lot of the original videotape was ~30fps.

Perhaps, but that observation does not necessarily indicate anything specific - because you're looking at 2 fields. It could be field shifted , or a field blended conversion, or a motion interpolated conversion, etc...

Examine it more closely by looking the individual fields (or double rate deinterlaced to 59.94) then stepping through and examine the pattern. Make particular note of blends, or motion interpolation artifacts

The segments converted from film (like the animated intro) show a mix of solid frames and interlaced frames, just like you get with telecine conversion of 24fps film to ~30fps.

That would be a good reason not to "deinterlace" those sections - you would want to IVTC (inverse telecine). Deinterlace would degrade the quality , resulting in about 1/2 effective resolution. It would be similar to deinterlacing a telecined NTSC film source, instead of IVTCing - deinterlacing in that situation destructive instead of returning the full quality film frames


BTW, the same earlier logic applies for 59.94p and "NTSC" area releases - BD does not support 1920x1080p59.94

And if you have a decent TV (assuming that's what you're watching it on) , another option is to leave it interlaced and let the TV figure it out. Most have various cadence detection and deinterlacing options . Some have good deinterlacers, but some do not. One reason you might opt to go a software deinterlacing route (assuming deinterlacing was the correct treatment for some sections) - is if your TV does a bad job. But "yadif/default" in handbrake would be considered a "not very good job" either; at least compared to more complex deinterlacers

[RasterEyes]

[QUOTE=poisondeathray;2763084]

Perhaps, but that observation does not necessarily indicate anything specific - because you're looking at 2 fields. It could be field shifted , or a field blended conversion, or a motion interpolated conversion, etc...

Examine it more closely by looking the individual fields (or double rate deinterlaced to 59.94) then stepping through and examine the pattern. Make particular note of blends, or motion interpolation artifacts

I'll have to figure out how to do that, since I don't offhand know of any software I have or device that will present a single field at a time.

Frame rate : 29.970 (30000/1001) FPS
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : MBAFF
Scan type, store method : Interleaved fields
Scan order : Top Field First

I checked to see if the video is actually encoded as interlaced (it is) rather that having progressive full frames composed of previously-interlaced material.

While the video I'm looking at shows the impact of having been interlaced at one time, it's not an interlaced signal

The segments converted from film (like the animated intro) show a mix of solid frames and interlaced frames, just like you get with telecine conversion of 24fps film to ~30fps.

That would be a good reason not to "deinterlace" those sections - you would want to IVTC (inverse telecine). Deinterlace would degrade the quality , resulting in about 1/2 effective resolution. It would be similar to deinterlacing a telecined NTSC film source, instead of IVTCing - deinterlacing in that situation destructive instead of returning the full quality film frames

I also had the detelecine option switched on along with deinterlacing.

And if you have a decent TV (assuming that's what you're watching it on) , another option is to leave it interlaced and let the TV figure it out. Most have various cadence detection and deinterlacing options . Some have good deinterlacers, but some do not. One reason you might opt to go a software deinterlacing route (assuming deinterlacing was the correct treatment for some sections) - is if your TV does a bad job. But "yadif/default" in handbrake would be considered a "not very good job" either; at least compared to more complex deinterlacers

I have yet to view any of this material anywhere than at my computer. I don't know what features my most upscale viewing option (a DLA-NZ8 projector) has for handling interlacing. It has very rarely had to cope with such a signal. I do remember watching an old copy of Shakespeare in Love and discovering, to my chagrin, that the disc was interlaced 29.97 Hz video. No trouble noticing or disliking that, and I quickly replaced that disc with a later release of the movie as progressive 24fps.

At any rate, as greatly improved as this video is from the crappy DVD quality previously available, it's still not big-screen cinematic experience stuff. But I do know that, shown on my computer screen at least, the video looks much better to me after the detelecine/deinterlace processing I've done. Perhaps if I knew what to look for, I'd see new problems. (Is ignorance bliss in this case? )

I imagine that my video software (VLC typically) is blending two fields into a single frame when I pause interlaced video, a frame which won't obviously look interlaced when the two fields are representing a single progressive frame.

But "yadif/default" in handbrake would be considered a "not very good job" either; at least compared to more complex deinterlacers

What has a better reputation?

[poisondeathray]

I'll have to figure out how to do that, since I don't offhand know of any software I have or device that will present a single field at a time.

You could use avisynth, with AssumeTFF().SeparateFields() - but there is a bit of a learning curve

If you double rate deinterlace to 59.94p - essentially you are viewing fields that have been interpolated to frames - so you could indirectly examine the field pattern that way.

VLC doesn't work correctly for me, but it works ok for some people - it has a bob or yadif 2x.

Frame rate : 29.970 (30000/1001) FPS
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : MBAFF
Scan type, store method : Interleaved fields
Scan order : Top Field First

I checked to see if the video is actually encoded as interlaced (it is) rather that having progressive full frames composed of previously-interlaced material.

If you refer only to mediainfo - note that mediainfo only tells you how it was encoded, not necessarily anything about the actual content ,or relationship of the fields ( present or past generation)

I also had the detelecine option switched on along with deinterlacing.

If it's a common pattern, HB should be ok. Maybe 3:2:3:2:2 ? I've seen BBC use that for 25p content for NTSC region releases. The benefit is the duration and audio is the same, no slowdown , no fiddling with pitch

But I do know that, shown on my computer screen at least, the video looks much better to me after the detelecine/deinterlace processing I've done.

Yes, and that's what matters

Perhaps if I knew what to look for, I'd see new problems. (Is ignorance bliss in this case? )

Unfortunately that sometimes is the case... Instead of simply enjoying the feature, you see little problems

I imagine that my video software (VLC typically) is blending two fields into a single frame when I pause interlaced video, a frame which won't obviously look interlaced when the two fields are representing a single progressive frame.

Theoretically it should depend on the VLC settings. (It doesn't work correctly for me, but apparently does for some people) . If you right click video=>deinterlace (on/off/automatic) , and the video=>deinterlace mode (options there).

But "yadif/default" in handbrake would be considered a "not very good job" either; at least compared to more complex deinterlacers

What has a better reputation?

QTGMC . The main strengths are smoothness / fluidity , fewer artifacts, "calmness" in terms of less flickering. The main negatives are slower to process, requires many dependencies (requires avisynth or vapoursynth, plus several plugins), can be oversmoothing/too much denoising and oversharp when used at the default settings

The main problem with yadif is the deinterlacing artifacts , "marching ants", buzzing lines. The main strength is it's fast and easy to implement (it's bundled with many software).

[jagabo]

You might want to use Bob instead of Yadif(2x) in VLC. Bob replaces the missing lines from data interpolated from the lines above and below (of the same field) so there's no possibility of cross-contamination. Yadif(2x) will use lines from the alternate field in non-moving parts of the picture (and often makes mistakes). Picture quality is worse with Bob but when you're just examining fields to determine the nature of your video that doesn't matter. You could also use Discard but that looks even worse (lines are simply duplicated).

[DB83]

Just curious.

Does your Box Set have the 'Network' label ? - This was the British company that did the restoration and released the DVD/BDs in Europe. I did not realise, if so, they also released in the US.


Now here is the oddity. Mediainfo is somewhat notorious for getting things wrong and I only have the disks of the first series to check*. But the report states :-


1920*1080 25fps (NTSC) Interlaced. It is the NTSC bit that puzzles me.


* Network went out of business a few year ago which is a crying shame since they did some fine work (and for me it is hard to track down single series sets since I would not want to buy a full box set already owning series 1). So there is zilch chance of a UHD under that label.


The box set on Amazon.com does show the Network label but if yours shows something else I would question its authenticity given the different frame rates.

[RasterEyes]

Just curious.

Does your Box Set have the 'Network' label ? - This was the British company that did the restoration and released the DVD/BDs in Europe. I did not realise, if so, they also released in the US.

If you mean Netwerk, then yes.

Now here is the oddity. Mediainfo is somewhat notorious for getting things wrong and I only have the disks of the first series to check*. But the report states :-


1920*1080 25fps (NTSC) Interlaced. It is the NTSC bit that puzzles me.

That would puzzle me too, especially since I've never seen mediainfo label anything NTSC, even at 29.97fps, and especially not at 25fps.

But then again, I've only used mediainfo to examine MKV files. If this is some sort of assessment of a whole Blu-ray disc, I've never tried that.

[DB83]

Thanks. (Pedantic comment) On the disks, I have several from this source, it is neither an 'e' or an 'o'. It looks more like an '0' with a line through the center.

https://filmscoremonthly.com/board/posts.cfm?threadID=151322&forumID=1&archive=0#:~:te...ks%20as%20well.

[RasterEyes]

Thanks. (Pedantic comment) On the disks, I have several from this source, it is neither an 'e' or an 'o'. It looks more like an '0' with a line through the center.

Sorry, I didn't mean to be pendantic (although I am damn good at that), "network" is just such a generic word I wanted to clarify we were talking the same thing. Yep. Same logo.