Every 5th frame is a repeat

[pooksahib]

Hi. In 1972, Chuck Berry recorded a show for the BBC which broadcast an edited performance. The full show is on DVD and I did a frame-by-frame to see if it was interlaced. It looks progressive but what I noticed is that every 5th frame is a copy of the one before it. mediainfo tells me the DVD is 25fps PAL.

I'm hoping someone can tell me the reason for the repeating 5th frame and what I should do about it. Mnay thanks.

[Bernix]

You can try (without knowledge of anything) open in handbrake or vidcoder, select constant frame rate and not same as source but 25 fps. Audio settings set to auto passthrough. There is small chance, it will remove duplicates. Use as output mkv file, so you can after few seconds of encoding open videofile that is in progress and move frame by frame to see if it really removed duplicated frames.
Other way is to use Avisynth, but it is more advanced way (knowledge needed) If first don't help, you will have to use it anyway (IMHO).

And the reason is simple. 25fps + 5 = 30 somebody encoded it to 30fps so encoder has to duplicate each 5th frame to fit 30 frames to 1 second. If mediainfo say it is 25 fps, try it on Vob with movie. Somebody maybe it horrible spoiled. Mediainfo can't be always right, and your only hope is it is really 30fps video.

[jagabo]

Duplicating every 4th frame (ABCDD EFGHH...) is one way of converting 24 fps film to 30 fps video. But it makes no sense for a 25 fps PAL video. It would imply the source was 20 fps.

Sometimes when a PAL video is convert to NTSC every 6th frame is a duplicate (ABCDEE FGHIJJ...).

<edit>
There's a video at youtube which may have been made from the same DVD:

https://www.youtube.com/watch?v=YtrOr3WKmyY

It has the same duplicate frame pattern (and interlace comb artifacts). It looks like there's also a missing frame every 5 frames. That would imply someone started with a 25 fps video, converted to NTSC by duplicating every 5th frame, then decimated back to 25 fps by deleting one frame out of every 6 -- but deleting a unique frame, not the duplicate.

</edit>

[pooksahib]

Interesting stuff and my thanks to you both for taking the time. But I get the impression there's little I can do other than leave the DVD 'as is'...

[LigH.de]

StainlessS translated "exotisches Interlacing" (originally in German by scharfis_brain) into English and published the result on his MediaFire account (see DATA folder). This may help you understanding some of the perversions happening in video studios...

A simple duplication should be reverted by SelectEvery / SelectRangeEvery in AviSynth. But if there is still combing, then better check if IVTC solves this issue correctly. I believe TIVTC (in companion with TFM) has an adaptive mode which reacts quite reliably even to Telecine pattern shifts.

[jagabo]

I get the impression there's little I can do other than leave the DVD 'as is'...

You should upload a sample of the source. It may be less damaged than the youtube video and it may be possible to fix it.

[darkknight145]

Somebody maybe it horrible spoiled. Mediainfo can't be always right, and your only hope is it is really 30fps video.

The BBC had a habit of discarding old programmes after they had aired thinking no-one would want to watch it again and because of the amount of space these films/video tapes took up in the archive. Hence the original Film or PAL video has been lost to time, the video you have has been reclaimed from someone in the US and reconverted back to PAL.

[pooksahib]

Yes, it's criminal the way that so many BBC programmes were not kept. Apparently they would use the same tape more than once (like we did in the days of VHS...) Anyway, here's a bit of the DVD for inspection. Many thanks.

[poisondeathray]

For single frame duplicates, that are actually a missing unique frame, you can use filldrops , filldrops2, or related functions to interpolate over the duplicate . Basically they make a determination based on conditionalfilter if a frame is same or not, if it's detected as a duplicate it will use mvtools2 to replace the frame with a motion interpolated one . If you decide on using antialiaser or vinverse or something to reduce the comb artifacts, do that before filldrops

[poisondeathray]

e.g I used qtgmc in progressive mode to smooth over the aliasing and combing issues. If it's too strong (too smoothing for your tastes), another strategy would be to use a softer antialiser like maa2 with vinverse before filldrops. Also boosted the shadows slightly with hdragc.

There are times you need to tweak the threshold value for filldrops' conditonal filter, but 1.3 is usually good starting point. And "interpolated" frames is never perfect - beware of edge morphing artifacts, but it's usually good enough on this type of footage to make it smoother

mpeg2source()
filldrops2()
qtgmc(inputtype=2, sharpness=0.5)
hdragc(max_gain=0.5)



function filldrops2 (clip c)
{
even = c
super_even=MSuper(even,pel=2)
vfe=manalyse(super_even,truemotion=true,isb=false, delta=1)
vbe=manalyse(super_even,truemotion=true,isb=true,delta=1)
filldrops_e = mflowinter(even,super_even,vbe,vfe,time=50)

ConditionalFilter(even, filldrops_e, even, "YDifferenceFromPrevious()", "lessthan", "1.3")

}

And there is a station logo in the upper right that looks blurred out, I wonder where it was originally from ?

[pooksahib]

That was quick work, poisondeathray. And very good, too. I'm still amazed at the number of filters etc that can be used...

The unblurred logo appears on another DVD of the truncated show. It's not a logo I recognise but, when Chuck speaks some French, Finnish subtitles pop up! Interestingly, this other DVD is properly progressive with no repeating frames.

As far as picture quality goes, I've decided to keep my own recording of the show broadcast by the BBC and I'll use your script for the extra songs. I haven't tried it yet so I may be back if I don't get it to work. Many thanks for your assistance.

[jagabo]

I did a variation of where I resized the video to 720x480 before the rest of the processing.

Mpeg2Source("Chuck sample.d2v", CPU2="ooooxx", Info=3) 
AssumeTFF()
Spline36Resize(720,480)
QTGMC()
Merge(SelectEven(), SelectOdd())
FillDrops2()

The result had less residual combing and halos but was also less sharp.

[pooksahib]

Thanks for taking the time, jagabo.