What's wrong with this picture?

[TB Player]

I'm doing a video conversion from PAL to NTSC, and in the process I was noticing some artifacts and wondering what causes them. So take a look at the first photo, then the second. In the second there are some obvious "lines" that show up, for example at the edges of the scarfs that the men are wearing. Would someone tell me what the cause of this is? Thanks.

[smrpix]

It comes from improperly resizing interlaced material. Please search these forums for PAL to NTSC or vice versa, conversion is frequently unnecessary. What are you trying to accomplish?

[jagabo]

You have interlaced video (each frame contains two separate half pictures). Since some frames don't have comb artifacts and some do the video must be telecined film. I don't see any improper resizing but you also have a problem converting interlaced YV12 to RGB.

Whether you should do anything about the interlaced video depends on what you plan to do with it. Make a DVD? Upload to Youtube? Etc.

[TB Player]

Thanks for the responses.

The source is a PAL DVD... it is from India and is not available in NTSC format. My client is asking for multiple NTSC DVDs from the source DVD. I can transcode the video using TMPGEnc or MainConcept, and that is what you're looking at... the top image is from a MainConcept Reference v1.6 conversion, and the bottom image is from TMPGEnc VMW5. But I'm wondering why the TMPGEnc video looks so much worse than the MainConcept video (and how to fix it). Both were transcoded interlaced.

[manono]

Samples would be nice. Especially from the PAL source, but also from both reencoded DVD versions. 10 seconds of the same scene showing steady movement will be plenty.

[jagabo]

If it's a progressive PAL DVD it would be best to resize the frame to 720x480 and encode progressive MPEG 2. Then use DgPulldown to add pulldown flags to 29.97 fps interlaced. If the source is interlaced it will be a bit more difficult. We'd have to see samples.

[TB Player]

No, the source is progressive. I'll see about getting samples up.

If I use DgPullDown will that change the length of the video (thus throwing off the audio sync)?

[manono]

If I use DgPullDown will that change the length of the video (thus throwing off the audio sync)?

No, that's one of the main reasons to use it - for PAL<->NTSC conversions so you don't have to reencode the audio.

[jagabo]

If I use DgPullDown will that change the length of the video (thus throwing off the audio sync)?

As manono pointed out, it doesn't change the length of the video. It uses the same technique movie DVDs use to convert 23.976 fps progressive to 59.94 fields per second without changing the length of the video: it repeats fields to make up for the difference. The technique can be used to change the frame rate from anywhere between 19.98 fps and 29.97 fps progressive to 59.94 fields per second.

[theewizard]

OK i'm confused

you have a method that works, aka the top picture from mainconcept
but you want to use tempgenc and and make it work, why ??

once you have processed the video and created the master folder and ISO image, you can burn as many correct 'good looking' DVD's needed

is your question merely a learning exercise ??

Thanks for the responses.

The source is a PAL DVD... it is from India and is not available in NTSC format. My client is asking for multiple NTSC DVDs from the source DVD. I can transcode the video using TMPGEnc or MainConcept, and that is what you're looking at... the top image is from a MainConcept Reference v1.6 conversion, and the bottom image is from TMPGEnc VMW5. But I'm wondering why the TMPGEnc video looks so much worse than the MainConcept video (and how to fix it). Both were transcoded interlaced.

[TB Player]

Yes, I can get a good conversation, but I'd like to know why the new version of TMPGEnc isn't performing as well as the old version of MainConcept. TMPGEnc has some nice features - it will import the full DVD structure, and it has a really nice editor. As far as workflow I think it's easier than MC. It looks completely different than the old version 2x but still allows the same configuration options. So when I saw the difference in quality between MC and TMPGEnc I wondered if I was using it correctly.

[hech54]

I've done several "quick and dirty" PAL to NTSC conversions with MainConcept Reference and have had no complaints either from me or friends/family who received the converted material. If it works, use it. Maybe there is a reason it was so damn expensive back then.

[jagabo]

Yes, I can get a good conversation, but I'd like to know why the new version of TMPGEnc isn't performing as well as the old version of MainConcept.

I suspect it's the other way around: you aren't seeing comb artifacts with MC because they are duplicating frames to convert 25 fps to 30 fps. That makes the video jerk 5 times a second (especially noticeable during smooth panning shots). Whereas TMPGEnc is performing 3:2:3:2:2 pulldown which plays much more smoothly.

[sanlyn]

I've no problem with TMPGENC MW as a basic h264 encoder. But conversions such as PAL->NTSC, etc., are best pre-processed with software that can address such issues more specifically. For instance, I would have gone into Avisynth and lossless AVI for this kind of work, as I've found no automated push-button software anywhere that really handles these conversions that well, and that includes some big names with big prices. I wouldn't be satisfied with the results shown in either of the images posted. Yes, doing it the hard way takes longer. Depends on what you want in the end.

[manono]

as I've found no automated push-button software anywhere that really handles these conversions that well, and that includes some big names with big prices.

AVSToDVD does it properly. It uses AviSynth and DGPulldown with HCEnc as the encoder (or does use HCEnc if you tell it to) . But, as you imply, if doing these kinds of conversions you should know enough to be able do it all manually.

We still haven't been offered any samples yet, though, so who knows how TB Player set up the encodes? One thing he did wrong, though, was to encode as interlaced where he should have been able to set things up to encode progressive. If the TMPGEnc one is hard telecined (which would require encoding interlaced), then that isn't optimal. If the MainConcept one has dupe frames, then (as jagabo mentioned) that's no good either.

[TB Player]

jagabo: Interesting perspective - I'm just getting back to the project and I'll re-check my results looking at it from your POV.

sanlyn: Thanks for the input. In my research for this project I did see Avisynth as an alternative solution (and I already have it installed in my PC) but I wanted to see first if MC/TMPGEnc would give me quicker yet acceptable results. After seeing these results I may give Avisynth a try, but I'm not sure what you mean regarding "lossless AVI" since I'm already in the MPEG domain.

manono: Sorry, just haven't had enough time to post samples. And why was creating interlaced MPEGs wrong - doesn't the video need to be interlaced to be in standard NTSC DVD format? Earlier jagabo wrote:

If it's a progressive PAL DVD it would be best to resize the frame to 720x480 and encode progressive MPEG 2. Then use DgPulldown to add pulldown flags to 29.97 fps interlaced.

Is this what you mean, and why is that an advantage over going directly to interlaced (does it produce better quality)?


And what would be the best way to create samples for you guys?

[manono]

Is this what you mean...

Yes, that's exactly what I mean. However, you'll have to turn off any DVD templates those encoders you used have in order to allow you to encode for 25fps and 720x480. This is more often done using AviSynth scripts and either HCEnc or CCE as the encoder.

...and why is that an advantage over going directly to interlaced (does it produce better quality)?

If, as I suspect, TMPGEnc encoded the telecine into the video, you have 29.97 frames per second with one-third being interlaced. Encoding interlacing takes way more bits for the same quality as does encoding all progressive frames. Compare that with encoding 25 all-progressive frames per second with 3:2:3:2:2 pulldown applied afterwards. You need something like a 30-35% higher bitrate to produce the same quality when encoding interlaced as compared to encoding progressive with pulldown.

And what would be the best way to create samples for you guys?

One way is to open a VOB in DGIndex and use the [ and ] buttons to isolate a small section. Then File->Save Project And Demux Video. Upload the resulting M2V. Choose a section with steady movement and 10 seconds will be plenty.

[TB Player]

OK, DGIndex... that was dead easy.

The PAL DVDs that are my sources are videos of a speaker (a Sikh guru) so in the main movies there isn't a lot of movement. But there are previews before each movie that have more movement, and that is what these samples are, although since even the previews don't have 10 sec of uninterrupted movement, there are two built-in cuts to get to over 10 seconds.

The first sample is from the progressive PAL DVD.
The second sample is MainConcept's interlaced NTSC video.
The third sample TMPGEnc interlaced NTSC video.

[jagabo]

Yes, as I suspected, MC duplicated every 5th frame (the camerawork in these shots is very jerky so it masks the jerkiness frame duplication creates), TMPG performed hard 3:2:3:2:2 pulldown. As manono suggested, encoding at 25 fps progressive and adding 3:2:3:2:2 pulldown flags will give better results.

HcEnc 25 fps progressive + pulldown, TMPG interlaced hard pulldown (4x point resize)



Click on the images a few times to see them full size. HcEnc file attached.

[TB Player]

Analyzing the samples, one thing I notice right away is that the MainConcept sample is BFF and the TMPGEnc sample is TFF, although I don't know how that effects the videos.

[aedipuss]

given the option dvd spec mpeg-2 is usually tff but bff is acceptable. since it started out progressive it doesn't make any difference to the video quality which is used.