cleaning video, does this make any sense?

[Soopafresh]

The only noise reduction plugins for VirtualDub that even come close to Avisynth's can be found here:

http://acobw.narod.ru/

[lordsmurf]

NeatVideo acts as a VirtualDub filter. Hard to beat that one, although even on my quad core it runs at only 3fps (with a bunch of other filters also in the chain).

[deadrats]

why are you deinterlacing if you are planning for SD blu-ray? Or are you making these mkv's for another purpose?

did you not say that denoise filters don't work that great on interlaced content? or was that someone else? if denoise filters work on interlaced sources then i won''t bother deinterlacing, i'll just denoise, maybe brighten up the output a bit, encode to interlaced mpeg-2 within the mkv (vdub mod only allows for avi, ogm or mkv output and the only one that willwork properly with mpeg-2 is mkv) then remux to m2ts.

[poisondeathray]

why are you deinterlacing if you are planning for SD blu-ray? Or are you making these mkv's for another purpose?

did you not say that denoise filters don't work that great on interlaced content? or was that someone else? if denoise filters work on interlaced sources then i won''t bother deinterlacing, i'll just denoise, maybe brighten up the output a bit, encode to interlaced mpeg-2 within the mkv (vdub mod only allows for avi, ogm or mkv output and the only one that willwork properly with mpeg-2 is mkv) then remux to m2ts.

Yes that's what I said. Only some filters work on interlaced content properly, and the ones that do, work less effectively. Neat video is an example that has a setting for interlaced, but it's not nearly as good as when using it on progressive stuff.

[jagabo]

With a lot of noise filters in VirtualDub you can get acceptable results simply by separating the fields:

Deinterlace -> unfold fields side by side
filter
Deinterlace -> fold side-by-side fields together

Be aware that VirtualDub does not handle interlaced YV12 sources correctly. You should use AviSynth to ConvertToRGB(interlaced=true) first.

VirtualDubMod handles interlaced YV12 properly.

[poisondeathray]

jagabo, don't you want to expand to pc.601 0-255 as well for color correction import into vdub, then bring it back to 16-235 when encoding?

ie.
ConvertToRGB(matrix="PC.601", interlaced=true)

[jagabo]

ConvertToRGB(matrix="PC.601", interlaced=true)

If you do that you'll need to make sure your output codec understands the levels are PC.601, not the usual rec.601. Or adjust the levels to rec.601 within VirtualDub (black at 0,0,0, full bright at 255,255,255).

[poisondeathray]

Did I misinterpret what you said about vdub losing superwhites and superdarks?

I have been doing color correction in other programs, and they all have little quirks as well. e.g. premiere will clamp 0-16 , 235-255 on some sources if you're not careful on how you do it.

[jagabo]

Did I misinterpret what you said about vdub losing superwhites and superdarks?

Well, I said you should make sure you adjust levels in AviSynth before giving the video to VirtualDub. Ie, black at Y=16, white at Y=235. Then when you give the video to VirtualDub there will be no superblacks or superwhites to lose (or none worth worrying about, maybe a little overshoot here and there).

But yes, if you want to do all your filtering in VirtualDub, and you have superblacks and superwhites in the source, you should use ConvertToRGB(matrix="PC.601", interlaced=true|false). Then in VirtualDub either make sure your output codec understands that you are giving it PC.601 levels (for example, frame serving to TMPGEnc Plus allows this), or adjust the levels to rec.601 in VirtualDub before output.

This thread has some discussion of video and RGB levels:
https://forum.videohelp.com/topic359049.html

[deadrats]

just 2 more potentially silly question: many encoders, tmpg express, main concept, procoder, pretty much anyone i have ever seen, allow you to choose whether you want interlaced or progressive encoding, what happens if your source is interlaced, you don't de-interlace (either with a built in filter or via avisynth) and you set the encoding parameters to encode progressive output? does the encoder automatically de-interlace? does it ignore the "progressive" option and encode as interlaced?

the second question deals with frame rate: recently i have run across a couple of dvd rips where the guy that ripped them had created a 60fps h264 640x480 video and the results were incredible (unfortunately due to the nature of the content i can't provide a short clip). many have tried to ask him how exactly he does it but he has been kind of coy, giving only round about answers, he did say that not all dvd's were ideal candidates for frame rate doubling and he also indicated that the dvd's had to be interlaced. i'm assuming that what he did was some sort of "bob" de-interlacing and then set the output frame rate at 60fps. in addition i assume if you are "unfolding" each interlaced frame and building a true frame you are effectively doubling the number of frames which means you also need to double the bit rate. what this guy does is use 4mb/s h264, presumably because 4mb/s is kind of high for h264 at 640x480 and thus there's enough bit rate to be go around for all the added frames. if anyone has any idea what the proper procedure is for frame rate doubling please let me know, i would love to try it out on some of my dvd's (though i believe we already covered that h264 720x480 progressive at 60fps is not blu-ray complaint).

thanks.

[poisondeathray]

just 2 more potentially silly question: many encoders, tmpg express, main concept, procoder, pretty much anyone i have ever seen, allow you to choose whether you want interlaced or progressive encoding, what happens if your source is interlaced, you don't de-interlace (either with a built in filter or via avisynth) and you set the encoding parameters to encode progressive output? does the encoder automatically de-interlace? does it ignore the "progressive" option and encode as interlaced?

Some programs will automatically deinterlace , e.g. adobe premiere and adobe media encoder will do this. Others will encode the scanlines and you will have a big mess. It will take a minute to test this out on a sample clip with your specific application

the second question deals with frame rate: recently i have run across a couple of dvd rips where the guy that ripped them had created a 60fps h264 640x480 video and the results were incredible (unfortunately due to the nature of the content i can't provide a short clip). many have tried to ask him how exactly he does it but he has been kind of coy, giving only round about answers, he did say that not all dvd's were ideal candidates for frame rate doubling and he also indicated that the dvd's had to be interlaced. i'm assuming that what he did was some sort of "bob" de-interlacing and then set the output frame rate at 60fps. in addition i assume if you are "unfolding" each interlaced frame and building a true frame you are effectively doubling the number of frames which means you also need to double the bit rate. what this guy does is use 4mb/s h264, presumably because 4mb/s is kind of high for h264 at 640x480 and thus there's enough bit rate to be go around for all the added frames. if anyone has any idea what the proper procedure is for frame rate doubling please let me know, i would love to try it out on some of my dvd's (though i believe we already covered that h264 720x480 progressive at 60fps is not blu-ray complaint).

This is referred to as "bobbing' . TempGaussMC_Beta1() with default settings is a smart bobber. The clip jagabo encoded was at 1/2 rate so he probably used selecteven() or selectodd() , which throws half the frames out. Most programs will have a bob mode, even TMPGenc has a bob mode. This preserves the temporal resolution instead of throwing 1/2 the frames out. This gives the super smooth look you are referring to that looks like soap opera, real live action . This is in contrast the the choppy 24p Hollywood style film

720x480/60p is not complaint with blu -ray. When you leave it interlaced, the blu-ray or dvd player will deinterlace it anyway, and you should still get the full 60 fields/second temporal resolution. (i.e it would make no difference for the smoothness). It's only when you deinterlace it single rate to 30p, that it becomes relatively choppy

x264 (with proper settings) is very efficient at encoding similar progressive frames, so doubling the number of frames, might only need 1.3-1.5x more bitrate for a certain "quality level" (like psnr or ssim) instead of 2x (but it will still vary by content complexity). But it's interlaced encoding efficiency is poor currently.

[jagabo]

If you encode an interlaced video as if it is progressive (and assuming the editor doesn't deinterlace it for you) you will end up blurring the two chroma channels together. You will also have playback problems. Over interlaced analog cables you may get fast-jerky playback as the player doesn't know which field to send first. Or you will see comb artifacts whenever there is motion (the player and the TV don't know the video is interlaced and display the frames as if they are progressive).

I agree that a well bob'd video usually doesn't need twice the bitrate as the changes between frames are smaller. But a poorly bob'd video may require more than double the birate:

xvid at 30i (target quantizer = 3) 60 KB:
fps30i.avi

poorly bob'd to 60p (target quantizer = 3) 1.36 MB:
bob60.avi

[deadrats]

I agree that a well bob'd video usually doesn't need twice the bitrate as the changes between frames are smaller. But a poorly bob'd video may require more than double birate:

ok, so what is the proper way to "bob" video and is it true that only certain dvd's are good candidates for this process? is that is true, then how do you determine if a dvd would be a good candidate other than actually trying it and seeing if the results came out good?

[poisondeathray]

The only "good" candidates are interlaced ones. These have a temporal resolution of 60 fields per second. If you have a progressive film source, it will be 24p.

Noise & non predictive motion (like camera shake) require much more bitrate for equivalent "quality". Hence the use of denoising and stabilizing filters

Poorly bobbed (or poorly deinterlaced) footage - like if you use TMPGEnc - leaves jaggies which , to the encoder, acts as "noise" and requires more bitrate to encode. If you use a high quality deinterlacer, not only will it look better, you will require less bitrate for a similar level of "quality"

[jagabo]

Hard telecined film should be inverse telecined back to progressive film frames. A simple TFM().TDecimate() in AviSynth works pretty well most of the time.