So is any DV codec better than MainConcept at decoding? Does it matter?

[Asterra]

I'll be exporting some DV clips as raw from Premiere Pro soon, to do things like deinterlace. I was noticing that the footage seems a little worse than I feel DV generally should be, and could not help but recall that there was a lot of debate over which DV codec to use.

The subject says it all, really. I want to make sure that the decoding of my DV footage is as good as the format allows. There may or may not be difference between the various codecs (MC, Sony, Canopus, Microsoft) in this regard. I'm really not certain. But for an example of the type of problem one might unexpectedly encounter when simply decoding compressed video, one needs only recall the DVD chroma bug (and its astonishing proliferation).

[ronnylov]

It was many years ago I used Mainconcept DV Codec so I can't say if it is good or not. But I really like the results from the free Cedocida DV codec. Since it is free you can try it yourself and see if there is any difference. There are settings optimized for different types if input and output which can be useful when converting between different formats.

[elmuz]

Even Panasonic should be free...

[titigrom]

De-interlacing affects sharpness of the footage. When outputting, quality is affected too by the choice of the output format (avi file with divx/xvid even avi-dv due to recompression, mpeg-2 or mp4 file with h264/x264). For DV export I choose MPEG-2 or MPEG-2 DVD format because it's close to original (keeps interlaced, high bitrate). My opinions, I may be wrong on some terms but I think it depends on your goal : archiving or watching on tv set, computer...

As you mentioned "there may or may not be difference between the various codecs". May be interested in this thread

https://forum.videohelp.com/threads/308716-general-questions-about-color-space-conversi...987&viewfull=1

[jagabo]

Panasonic is full of bugs. And it always decodes to RGB with the standard rec.601 matrix. So any super-black or super-bright areas are crushed before they get to your encoder. I don't remember for sure (it's been banished from my computers for quite some time) but I think it duplicates chroma samples rather than interpolating them leading to vertical color banding. Cedocida is much better.

https://forum.videohelp.com/threads/239994-DV-CODEC-for-VHS-Capture?p=1421497&viewfull=1#post1421497

Those bands are caused by the decoder duplicating chroma samples during the 4:1:1 to 4:4:4 upscale. Some editors have filters for fixing that after the fact. Like VirtualDub's 4:1:1 helper.

Attached is a sample if you want to test your decoder.

[Asterra]

There may be something wrong with that dv video you posted. The only thing I've got which even recognizes it as video is MediaInfo. Perhaps the absence of an audio track? Hard to say.

[jagabo]

It hasen't given me any problems. But I made a new one with audio and compressed with Cedocida. GSpot shows this as Type 2 DV AVI. I zipped it up for error detection, not to mention a much smaller file.

[Asterra]

Ah, thanks. That one gave my players / Premiere Pro etc. no difficulties.

I feel like I should be disappointed that Premiere Pro's decoder (presumably MainConcept) handles that difficult chroma challenge splendidly. Meanwhile, whatever DV codec is being used in the likes of MPC handles it badly, generating a blocky effect. Maybe I really am just expecting too much of DV video.

Perhaps I will try some deblocking in Avisynth when it comes time to deinterlace. The occasionally obvious macroblocking is what I'm really hung up on.

[jagabo]

Right click on the video window while playing in MPC and hover over Filters. That will show you the list of filters being used. Click on the video decoder and see if there are any settings you can change.

Normal AviSynth deblocking filters won't work for this DV artifact. You can use ReInterpolate411:
http://avisynth.org.ru/docs/english/externalfilters/reinterpolate411.htm

ffdshow's DV decoder is one of the most common with this problem.

[Asterra]

Oh, I'm not concerned with how DV looks in the likes of MPC. If Premiere Pro's decoder is doing the best job possible (as seems to be the case), that's all I need. Though I suppose I ought to see if it's doing nasty things to the levels like the Panasonic codec reportedly does.

Rather than "deblocking," I perhaps should have said mosquito noise, as that is the nature of the artifacting that seems most prevalent with the video I'm dealing with.

[Asterra]

So I've been fiddling with Premiere Pro's lack of codec transparency and the consequences, in addition to how Avisynth and other resources handle DV (on my system at least). It seems that:

1) Premiere Pro correctly handles chroma (with interpolation) internally and when exporting.
2) Virtualdub also correctly handles chroma internally.
3) Avisynth does not correctly handle chroma without the use of ReInterpolate411, which in turn requires a converttoyuy2(interlaced=true).
4) Telling Premiere Pro to export unmodified DV as DV (smart render) results in a DV video identical to the original, except it exports type 2 only.

5) The most important and unfortunate observation I have made is that Premiere Pro does not possess the ability to render DV to raw without performing a noticeably destructive process on the image it prior to export. I tried it four ways:

Microsoft AVI, codec: "None", maximum depth.
Microsoft AVI, codec: "None", not maximum depth (PPro says this defaults to 8-bit when unchecked, but a drop-down menu below makes things uncertain).
Uncompressed Microsoft AVI, UYVY, maximum depth.
Uncompressed Microsoft AVI, UYVY, 8-bit depth.
(v210 was not liked by Avisynth, even with directshowsource.)

I then took a good, close look at the same frame in each of those renders in addition to a smart render from Premiere Pro and the original DV file. The smart render and original DV file proved identical. But all four of the uncompressed renders were visibly different and generally noisier. Perhaps most unsettling of all is the fact that all four of those individual raw renders were different from each other.

Now, I'm not exactly panicking over this, as I do still have the option of smart rendering my DV edits from Premiere Pro and doing my extra processing on that render. But eventually I'm going to be telling Premiere Pro to render my entire video as raw so I can pass it through x264, and as far as I can tell, this carries an unavoidable, inexplicable degradation of the quality.

[jagabo]

You're wandering around a mine field of codecs and colorspace conversions!

1) Premiere Pro correctly handles chroma (with interpolation) internally and when exporting.

I'm not sure "correctly" is the right word. It's a matter of intentions. The reason for duplicating chroma samples is to make sure that if you convert from 4:1:1 to 4:2:2 or 4:4:4 and back to 4:1:1 the original chroma samples are unchanged. When values are interpolated the result looks smoother but conversion back to 4:1:1 may generate small errors. So duplication may be better if you plan on going back and forth between 4:1:1 and 4:4:4 several times.

2) Virtualdub also correctly handles chroma internally.

Were you using its internal DV decoder or a system installed VFW decoder? Actually, I know VirtualDub's internal decoder interpolates. Microsoft's current decoder interpolates (not sure about really old versions). ffdshow duplicates.

3) Avisynth does not correctly handle chroma without the use of ReInterpolate411, which in turn requires a converttoyuy2(interlaced=true).

AviSynth isn't decoding the video. If you use AviSource() then a system installed VFW decoder is doing the decoding, usually to YUY2 or RGB. If you use DirectShowSource() a system installed DirectShow DV decoder is being used. Again, usually to YUY2 or RGB.

4) Telling Premiere Pro to export unmodified DV as DV (smart render) results in a DV video identical to the original, except it exports type 2 only.

There shouldn't be any difference if you haven't done any filtering. Smart mode simply copies the compressed frames from the source to the destination when possible. So there is no decompression, colorspace conversion, or recompression if you're not filtering.

5) The most important and unfortunate observation I have made is that Premiere Pro does not possess the ability to render DV to raw without performing a noticeably destructive process on the image it prior to export. I tried it four ways:

Microsoft AVI, codec: "None", maximum depth.
Microsoft AVI, codec: "None", not maximum depth (PPro says this defaults to 8-bit when unchecked, but a drop-down menu below makes things uncertain).
Uncompressed Microsoft AVI, UYVY, maximum depth.
Uncompressed Microsoft AVI, UYVY, 8-bit depth.
(v210 was not liked by Avisynth, even with directshowsource.)

I then took a good, close look at the same frame in each of those renders in addition to a smart render from Premiere Pro and the original DV file...

What are you using to compare the frames? Two media players playing the same video side by side may look different. Because one may be using the graphics card's video overlay feature, the other not (only one can use video overlay at a time, video overlay has it's own proc amp settings in the graphics card's setup applet). Even if you are comparing with editors that aren't using Video Overlay similar videos might look different depending on whether or not the video decoder performs the usual rec.601 contrast expansion or not. For example a DV decoder may perform the contrast expansion whereas RGB export from Premiere might not. I don't know Premiere so I can't say whether it does or not but that's the type of issue you're dealing with.