Script to detect duplicate frames?

[kieranvyas]

I gave some edited footage to an expert to look over (as I was experiencing some problems) and he gave me this response:

"I had a look at your clip today and I noticed that it has an odd field pattern of mostly progressive 50 frames per second but sections of duplicated frames. Something like:

2:2:2:2:2:2:2:2:2:2:2:2:2:4:2:4:2:4:2:2:2

So I split it into fields and ran it through a script that detects duplicate frames. I've attached the first script. It shows frame numbers on the left followed by a number with 5 decimal digits, a value of around 2 indicates it is the same frame as the one before it. However this is not exact."

(I have attached the script to this post)

Now unfortunately he is totally unavailable to contact. I am hoping to pay someone else to fix the problem but won't be able to tell if they've fixed it without a broadcast monitor - which I don't own.

Can anyone tell me how to reproduce a test like this and how to read the results to find out if my video has an 'odd field pattern' or duplicate frames? Thanks!

[poisondeathray]

The best test is your eyes . Either look at the fields individually or "bob" them to frames

Scripts are not always reliable - false positives, false negatives. Compression "noise" can confound metrics and results (e.g. a true duplicate will be missed because of temporal compression differences)

Duplicate frames can be "normal" when you intermix sections for example like you did with 25p and interlaced. They can be normal even in interlaced content (e.g. when you have a static shot) .

[kieranvyas]

The best test is your eyes . Either look at the fields individually or "bob" them to frames

Scripts are not always reliable - false positives, false negatives. Compression "noise" can confound metrics and results (e.g. a true duplicate will be missed because of temporal compression differences)

Duplicate frames can be "normal" when you intermix sections for example like you did with 25p and interlaced. They can be normal even in interlaced content (e.g. when you have a static shot) .

The problem is, the problem that I'm consistently getting is these horrible interlace lines on objects in motion. But this is not visible on my monitor or TV. However, it is very visible on a broadcast monitor that I was able to test it on once.

[poisondeathray]

You should know the drill by now...

Post an unprocessed sample that demonstrates the problem

Broadcast monitor is only as reliable as the setup - if it wasn't setup properly you might be getting incorrect results

[manono]

Can anyone tell me how to reproduce a test like this and how to read the results to find out if my video has an 'odd field pattern' or duplicate frames?

an mfile.txt is created when running an analysis pass in Virtual Dub using Donald Graft's MultiDecimate filter in an AviSynth script:

http://f0003555.xsph.ru/docs/avisynth/russian/externalfilters/multidecimate.htm

Unfortunately, Donald Graft's site with all his filters seems to be gone now. I'll include it here in case you'd like to try. After creating that mfile.txt, you can then remove those duplicate frames. But, as pdr suggests you might not need MultiDecimate and might be able to remove the dupes with a more simple decimation filter (TDecimate). But a sample is needed to be sure.

[kieranvyas]

The attachment didn't work in this post for some reason, see the post below.

[kieranvyas]

You should know the drill by now...

Post an unprocessed sample that demonstrates the problem

Broadcast monitor is only as reliable as the setup - if it wasn't setup properly you might be getting incorrect results

Uncompressed clip of a particularly bad section attached to this post


The editor I am thinking about hiring said: "I think the issue might be as simple as going back to the beginning and working in a different timeline."

Source footage is: 1920x1080 progressive 50fps MPEG-4 Quicktime

[poisondeathray]

Obvious cadence issues there .

Do you have the corresponding original, original clip ? You said it was AVCHD converted in FCP to Prores/Mov ? Or do you have the MOV ?

[kieranvyas]

Obvious cadence issues there .

Do you have the corresponding original, original clip ? You said it was AVCHD converted in FCP to Prores/Mov ? Or do you have the MOV ?

Here is the original. This clip is particularly bad for causing the horizontal interlace lines through objects in motion (when encoded to DVD format), would you be able to tell me if its the edited uncompressed file, or the original source footage that might be causing this problem?

EDIT: I have just realised, the first half of 'worst(short)' is slowed down in premiere pro. The second half is the clip played normally. Hopefully that won't affect tests.

[sambat]

Unfortunately, Donald Graft's site with all his filters seems to be gone now.

Here?

http://rationalqm.us/index.html

[manono]

Unfortunately, Donald Graft's site with all his filters seems to be gone now.

Here?

http://rationalqm.us/index.html

Cool, thanks. The original neuron2 site seems to be gone, but replaced by this one.

[manono]

Here is the original. This clip is particularly bad for causing the horizontal interlace lines through objects in motion (when encoded to DVD format), would you be able to tell me if its the edited uncompressed file, or the original source footage that might be causing this problem?

I'm not going through 5664 frames, but the 500 or so I examined look just fine. There shouldn't be any problem converting this for PAL DVD. In AviSynth-speak:

LanczosResize(750,576)#or whatever resizer you like
SeparateFields()
SelectEvery(4,0,3)
Weave()

Then encode it as interlaced and 16:9. I have no idea how you did it yourself.

Your black levels seem okay, if a bit high. Your white levels are definitely high, blown out in places, even.

I thought this thread was about duplicate frames?

[kieranvyas]

I'm not going through 5664 frames, but the 500 or so I examined look just fine. There shouldn't be any problem converting this for PAL DVD.

Thank you for looking through those 500 frames on the source footage!!! Would you be able to have a look at some frames on the uncompressed edited version?? To see if I've messed it up somehow in PP.

I decided to send my project to a professional company for encoding as my results were just rubbish, even after troubleshooting. That was the exact avisynth I am using but in places there were interlace lines through objects in motion and it looked awful. So this company replied to me saying my uncompressed edited file had an 'odd field pattern' and 'duplicated frames' which would explain any stuttering or interlace lines showing. They said that lots of work needed to be done before the encoding stage. If my source footage is fine, then I must be messing it up in PP somehow. If you could confirm this by testing some frames on the uncompressed edited file I posted previously I would be eternally grateful. Thank you!

[jagabo]

Maybe you have the field order set wrong in your encoder? AviSynth assumes BFF by default so the output of the aforementioned script will be BFF.

ffVideoSource("TODD IN PEN FIRST TIME..mov") 
Spline36Resize(720,576)
AssumeTFF()
SeparateFields()
SelectEvery(4,0,3)
Weave()

Encoded wiht HCEnc, TFF, no duplicate frames or fields. It's overly sharp and will flicker a lot on playback. Some blurring before downscaling would be better.

[kieranvyas]

Maybe you have the field order set wrong in your encoder? AviSynth assumes BFF by default so the output of the aforementioned script will be BFF.

Thanks! But it is set to TFF, and I have been assured the problem lies in my uncompressed export from PP, before the encoding stage. And my script is:

AssumeTFF().SeparateFields().SelectEvery(4, 0, 3).Weave()

Hopefully manono can test the uncompressed file I posted to confirm this!

[jagabo]

Yes, your AVI file in post #7 has many duplicate frames. That could cause comb artifacts (after conversion to interlaced MPEG 2) upon playback as the player might get confused about whether the video is progressive or interlaced. It will definitely cause jerky playback. I added an HCEnc encoded video to my last post.

By the way, that file isn't uncompressed, it's losslessly compressed with Lagarith.

[kieranvyas]

Yes, your AVI file in post #7 has many duplicate frames. That could cause comb artifacts (after conversion to interlaced MPEG 2) upon playback as the player might get confused about whether the video is progressive or interlaced. It will definitely cause jerky playback. I added an HCEnc encoded video to my last post.

By the way, that file isn't uncompressed, it's losslessly compressed with Lagarith.

Thank you for checking that!!!! So as my source footage is fine, that must mean something is happening in PP to cause the problem! Great to know, now I just have to figure out what it is that is going wrong in PP.

[kieranvyas]

Yes, your AVI file in post #7 has many duplicate frames.

Would you be able to tell me how you checked for duplicate frames, (I tried the MultiDecimate filter but didn't fully understand it)? So I can run these tests myself when I fix the problem. Thank you very much!

[jagabo]

Yes, your AVI file in post #7 has many duplicate frames.

Would you be able to tell me how you checked for duplicate frames, was it the MultiDecimate filter? So I can run these tests myself when I fix the problem. Thank you very much!

Since it's an AVI file I just opened the video in VirtualDub and stepped through frame by frame. It had duplicate with a pattern like 1 1 2 3 3 4 5 5 6 7 8 8. So about 3 duplicates in every 12 frames. Since it has the same 50 fps frame rate it runs longer.

[kieranvyas]

Yes, your AVI file in post #7 has many duplicate frames.

Would you be able to tell me how you checked for duplicate frames, was it the MultiDecimate filter? So I can run these tests myself when I fix the problem. Thank you very much!

Since it's an AVI file I just opened the video in VirtualDub and stepped through frame by frame. It had duplicate with a pattern like 1 1 2 3 3 4 5 5 6 7 8 8. So about 3 duplicates in every 12 frames. Since it has the same 50 fps frame rate it runs longer.

Great! Thanks a lot to you and everyone else, you've really helped me.

[poisondeathray]

You can do the same thing in premiere. Advance frame by frame and check for duplicates or missing frames

The MOV looks to be re-wrapped AVCHD (not prores) . There is no problem in premiere here on that MOV on the PC version CC2015. I noticed that your AVI said PP CC 2014 in the metadata - so you might consider upgrading to the newest version to see if it helps. Also clear out the cache files before you do any tests. But there are differences between the Mac and PC version of PP

Another thing you can try is actually converting it to prores before editing. It's a lot more stable and consistent to edit on Macs.

[kieranvyas]

You can do the same thing in premiere. Advance frame by frame and check for duplicates or missing frames

I've just done that and checked for duplicates, but I realise i've done a really really stupid thing and sent you a clip where the first part is in slow motion (slowed down in PP). I believe the slow motion part has the duplicated frames. If you jump to the second part of the clip (after frame 116) there are no duplicated frames that I can see. So maybe that isn't the problem after all! Sending me back to square 1

[poisondeathray]

Yes, the second part of the AVI is fine



Upload a section of the final render, that exhibits the problem when viewed on a broadcast monitor. Describe exactly what the problem is - is it just horizontal lines or combing similar to seeing the video not deinterlaced on a flat panel? Or is it something else ?

Also if you know how it was setup, connections - e.g what was connected to what, HD-SDI to what, how was it played etc...

[kieranvyas]

Yes, the second part of the AVI is fine

Upload a section of the final render, that exhibits the problem when viewed on a broadcast monitor. Describe exactly what the problem is - is it just horizontal lines or combing similar to seeing the video not deinterlaced on a flat panel? Or is it something else ?

Also if you know how it was setup, connections - e.g what was connected to what, HD-SDI to what, how was it played etc...

Here is some footage (11secs) of the worst parts.

worst3.1 = lossless export from PP (1st half)
worst3.2 = lossless export from PP (2nd half)
worst3 = final encode

This footage is comprised of 3 clips. Apologies for probably using the incorrect terminology but the symptoms are:

#1 the video is stuttering causing a motion blur around the dog, like it is ghosting
#2 lines over the dog during motion, distortion (the source footage of this clip is the one attached in a previous post above)
#3 large wavy lines/distortion on the bottom of the screen

I don't have access to a broadcast monitor at the moment unfortunately, so I am experiencing these problems on a flat panel TV and DVD player.

Avisynth script for worst3 (final encode):

AviSource("worst3.avi")
ConvertToYV12()
Spline36Resize(720,576)  
Blur(0,0.5)
ColorMatrix(mode="Rec.709->Rec.601",clamp=0) 
AssumeTFF().SeparateFields().SelectEvery(4, 0, 3).Weave()

[jagabo]

I don't see anything wrong with the interlacing in worst3.m2v. It has TFF interlaced frames, encoded interlaced TFF, 50 different fields per second. It should play back normally on any device that handles interlaced video properly. If you see interlaced comb artifacts on playback it's a shortcoming of the player's or TV's deinterlacing. Thin horizontal lines of the fencing will cause aliasing artifacts.

[poisondeathray]

I think he's changed the problem description

I think this was discussed in your other thread - Those artifacts around the dog are compression artifacts mostly related to low bitrate. You can't encode high motion (difficult to compress scenes) with low bitrate MPEG2 and expect it to look ok. Those blocky artifacts around the dog may give the subjective impression of "ghosting" or "motion blur", because the separation between foreground object (dog) and background (grass) isn't as distinct. But I'm curious about the choice of words "stuttering". I don't see any evidence of that

[kieranvyas]

I think he's changed the problem description

I think this was discussed in your other thread - Those artifacts around the dog are compression artifacts mostly related to low bitrate. You can't encode high motion (difficult to compress scenes) with low bitrate MPEG2 and expect it to look ok. Those blocky artifacts around the dog may give the subjective impression of "ghosting" or "motion blur", because the separation between foreground object (dog) and background (grass) isn't as distinct. But I'm curious about the choice of words "stuttering". I don't see any evidence of that

Thank you for your patience! On two different DVD players that first clip stutters like it is struggling to play. I just need to figure out if it's a problem caused in the editing process, or the encoding.

I am planning on paying a professional company to encode the DVD, but they took my lossless sample and said 'it had an odd field pattern' but they may have been referring to the slow motion parts. If you could do me one last favour and check one of the lossless samples I attached in my previous post, and confirm there is nothing wrong with it, then I will know it is the encoding that is the problem and hopefully this company will fix that - as much as is possible.

THANK YOU!

[poisondeathray]

I am planning on paying a professional company to encode the DVD, but they took my lossless sample and said 'it had an odd field pattern' but they may have been referring to the slow motion parts. If you could do me one last favour and check one of the lossless samples I attached in my previous post, and confirm there is nothing wrong with it, then I will know it is the encoding that is the problem and hopefully this company will fix that - as much as is possible.

There is nothing wrong with any of the lossless lagarith AVI samples. There is no "field" pattern per se - because they are all progressive still at that point (50p)

Yes , there are duplicate frames in that one slow motion section you referred to. There are other options to generate slow motion "inbetween" frames besides frame duplicates - you can look into motion interpolation (aka "optical flow") options (e.g. twixtor, kronos, timewarp - or avisynth options like mvtools , mflowfps, svpflow) . But they have drawbacks such as edge morphing artifacts

Don't expect too much improvement from the encoding from a professional company. You simply can't get around that low bitrate MPEG2 limitation.

The only way to improve it (more than a little bit) is to use higher bitrate e.g. maybe use DVD9 . Other things you can do are structure the edit differently - intersperse more low motion scenes with high motion scenes. That way the buffer has more time to recover, you can distribute the bitrate more effectively. Also consider using DC10 and maybe a low bitrate matrix, softer resizing at least for the high motion scenes . You can also prefilter the high motion scenes with mild denoising which will reduce the compression requirements

[kieranvyas]

I am planning on paying a professional company to encode the DVD, but they took my lossless sample and said 'it had an odd field pattern' but they may have been referring to the slow motion parts. If you could do me one last favour and check one of the lossless samples I attached in my previous post, and confirm there is nothing wrong with it, then I will know it is the encoding that is the problem and hopefully this company will fix that - as much as is possible.

There is nothing wrong with any of the lossless lagarith AVI samples. There is no "field" pattern per se - because they are all progressive still at that point (50p)

Yes , there are duplicate frames in that one slow motion section you referred to. There are other options to generate slow motion "inbetween" frames besides frame duplicates - you can look into motion interpolation (aka "optical flow") options (e.g. twixtor, kronos, timewarp - or avisynth options like mvtools , mflowfps, svpflow) . But they have drawbacks such as edge morphing artifacts

Ok that is brilliant!!! Honestly wouldn't be able to figure this out without you guys on this forum. Thank you for taking the time to educate a newbie.

[poisondeathray]

It's actually not that difficult to check it yourself in the future - jagabo mentioned one way - just open the AVI into vdub and go frame by frame . You're looking specifically for duplicate frames or gaps in motion (jumps) .

I added an edit to my post above - make sure you read the part about expectations for professional compression. Don't get your hopes up high - the biggest obstacle you're facing is low bitrate MPEG2. A professional compressionst can do a few tricks , but nothing that will magically make it look drastically better at that bitrate . If you're willing to go DVD9 as suggested in your other thread - that would help