I want to edit a video (change audio, add credits, apply some minor adjustments and stabilization), however, I would like to use as 'basis' the best possible quality. The source is a 720x576 mpeg2 video from a camcorder (PAL 25fps?, interlaced, pixel aspect of 1.4568* or 1.422̅2* so, the output would be 1049x576 or 1024x576). *-> different depending on the software
I would like to upscale that to 720p.
The problem is, for some reason beyond my understanding, no other software can achieve the quality of MPC-HC playback. I've downloaded trials of pretty much any editing software out there (vegas, the 2 from corel, avidemux, virtualdub, cyberlink, premiere, and others) and played extensively with the settings, as well as tried other video players (vlc, potplayer, windows media, mplayer (smplayer), mpv, quicktime) and none of them could achieve the quality of MPC-HC playback. I've tried neatvideo also.
I'm almost sure MPC-HC applies some kind of filter to the video due the interlacing or perhaps the deinterlacer is just superior. Changing the setting didn't seem to have any effect and I couldn't find what's so special about MPC-HC.
I'm attaching images showing the difference between MPC-HC and the others (including the video editors).
I just want a way to reproduce the MPC-HC quality in a video editor (except premiere and expensive ones) so the output file will have decent quality. Perhaps there's a way to 'dump' the MPC-HC video in a 'raw' file or something, so I could just use virtualdub.
Also, I don't understand why MPC-HC deinterlaces the video AND plays it at double framerate, because afaik it only should be played at twice the original framerate if you're discarding the other 'field' at every frame, so, you'll end up with two images for each original frame. So you either have both fields deinterlaced playing @ 25fps or you can just 'intersperse' them, so you have half the vertical resolution but twice the framerate, but it seems MPC-HC is actually managing to have both, and by analyzing frame-by-frame it's possible to see that it actually has higher resolution AND twice the framerate. It's quite crazy if you ask me. Quicktime for example displays a pixelated video @ 25fps, so, 3/4 of the 'data' is just being lost (one interpolation field and one frame (2 fields))... unless MCP-HC is applying some very smart twixtor-like frame interpolation what I really doubt.
Perhaps the best solution would be an 'output' for MPC-HC that writes a lossless file to disk (1024x576 @ 50fps in this case), afaik VLC allows you to do that.
Seriously, this is the first AND the last video I'm editing (that's why I want a 'free/cheapo' solution) this thing certainly is not for me. It's going to drive me crazy.
Any help is appreciated.
Thank you in advance.
+ Reply to Thread
Results 1 to 10 of 10
You're going to have to dip your toes into avisynth and QTGMC if you want better than MPCHC quality (yes, better)
Interlaced PAL 25fps means 50 fields per second. That's 50 moments in time per second represented. Double rate deinterlacing essentially makes each field into a frame (+/- other post processing and interpolation) . That's why motion should be smoother. Single rate deinterlacing would mean discarding every 2nd field (throwing away 1/2 the temporal information) to yield 25 frame/s instead of 50 frames/s.
Each field is like 1/2 a frame. "Smart" deinterlacers attempt to use temporal information from adjacent fields to fill in the gaps in information. What you're seeing in quicktime is a simple resize. That's why you get the aliasing or "stair stepping" deinterlacing artifacts . MPCHC depends on how you have it configured, but the results usually arent very good compared to QTGMC
Software players (including MPCHC) typically use low quality deinterlacers, because the emphasis is on speed. High quality deinterlacing requires more intensive CPU usage, is less amenable for realtime playback . What people typically do, is use avisynth and apply that to a lossless intermediate for editing in their NLE
If you combine two fields for a single frame you get 25fps but each interlaced field is kind of like a frame, in that it's a unique moment in time, so if you combine each field with the preceding field and then again with the following field you get 50fps. The latter method is generally how it's done on playback. Traditionally video is/was de-interlaced to 25fps when encoding because hardware players didn't support 50fps progressive video. These days support for 50fps progressive up to 720p would be fairly normal (Bluray players etc). If Quicktime is displaying "pixelated" video maybe it's not de-interlacing at all.
If you don't have a lossless video encoder installed, install MagicYUV
The Yadif de-interlacer deinterlacing to "double frame rate" (50fps output) should give you pretty much the same quality as when playing the original video using MPC-HC. I couldn't find a link to the VirtualDub Yadif plugin which does 50fps de-interlacing so I've attached the one I have to this post. I've never used it myself, but it seems to work. Stick it in the VirtualDub plugins folder.
Once you've opened the video in VirtualDub, add the Yadif de-interlacer from the list of filters (use the Video/Filters menu). Select the "double frame rate" option for 50fps output. The frame count in VirtualDub's navigation bar should double if you do. Under compression choose the MagicYUV encoder. Under the Audio menu you can leave it on "copy" or enable full processing and under compression select PCM to output lossless audio. Save the video/audio as an AVI.
With any luck that'll give you a lossless output nicely de-interlaced to 50fps. You should be able to open it with most editing programs (possibly not Quicktime). The output AVI will be 720x576 but you can resize it while re-encoding or while editing etc.
Last edited by hello_hello; 11th Mar 2014 at 08:03.
I thought 'single rate deinterlacing' were 'blending' two fields together to get a higher resolution image @ 25fps... in a second the camera produces 50 images, the best use for them imho would be to somehow get 25 images at twice the vertical resolution by 'blending' two of them together, since 1/50 of a second is not very representative, so, hardly any ghosting artifacts would be generated, instead of displaying the video @ 50fps, because imho a higher resolution image is more important and 25fps is enough for most cases. It seems though the only way to use the extra 'images' produced by the camera is to double the framerate, what doesn't make much sense imho.
In my case MPC-HC correctly played the video with deinterlacing without manual configuration. On VLC it's necessary to manually configure the deinterlacing. I've tried the Yadif (and Yadif2x) deinterlacing in VLC and Mplayer and the Yadif plugin for Vegas without success, the quality is always the same as the image I attached, almost pixel-by-pixel. The difference between the two (yadif and yadif 2x) is the framerate, the image quality is the same. In case of vegas, the yadif plugin actually is worse than the internal deinterlacer, which has the exactly same quality as the other editors and players. The only exception is MPC-HC, everything else, editor or player, produces the same (almost exactly when analyzing pixel-by-pixel) image quality. I would like to know what is the difference of MPC-HC.
On the screenshots it's not possible to see, but when the video is playing, you can notice the horizontal lines that produces discontinuation in shapes, of course you can't see the actual 'lines' but it's surely possible to notice a kind of blurred pixelation in the horizontal direction, as you can see on the images, the cracks are smooth and continuous in MPC-HC. How it achieves that extra definition is a mystery... at first I thought it were actually twice the vertical resolution, but it seems it's just a 'smart deinterlacer' according to poisondeathray.
Thank you guys.
"Single rate" only refers to whether or not 1/2 the information is dropped. It only refers to the output framerate, or "temporal resolution", not "spatial resolution" . Interlaced video doesn't have full spatial resolution, except potentially in sections that are static. If you want full spatial resolution, shoot progressive
Blending is one method of deinterlacing. It produces blurry images in motion, because the fields represent different moments in time . You can do that in vegas , set the deinterlacing to blend instead of "interpolate" .
"Weaving" 2 fields in sections that are static will give you full resolution
Smart deinterlacers use temporal information and interpolation techniques to fill in the gaps in motion weave 2 fields together - so it attempts to give you full frame resolution sections that have motion. In sections that are static, they weave the fields together (or at least they attempt to), so you get full frame resolution
It depends what "most cases" are. Have you watched sports? (football/soccer) at 25fps isn't nearly smooth enough . When you watch interlaced video on TV, that's what's happening (you're effectively seeing 50 different images per second)
You can search for previous threads on deinterlacing comparisions. The difference is like night/day. Or if you upload a sample and someone will walk you through some options
Thank you for your replies guys.
It seems QTGMC is the way to go. I saw a comparison and I think it is possibly even better than MPC-HC (it's ridiculous that a real-time deinterlacer is better than the ones in video editing software in the first place imho). The problem is, I don't have a single clue on how to get started quickly using AviSynth and QTGMC... it seems I have to program the thing myself with commands to apply the deinterlacing and stuff, hopefully it should be 'doable' to just apply the QTGMC to the videos and get them in a lossless format so I can just load them in VDub. I can't wait to finish this editing so I can simply uninstall all the things I installed and forget this real nightmare.
Video editing imho, specially because of these crazy formats, is far beyond insanity imho.
Not sure why you think this is supposed to be easy. Like playing an instrument -- it's drudgery until you learn enough to bang out a couple of tunes. --And then you're just getting started.
So relax and enjoy.
I don't know which images are which in your screenshots.
I still think you're wrong about MPC-HC v Yadif. I compared a video played using VLC and Yadifx2 with MPC-HC and they look virtually the same to me. Be careful you're comparing apples with apples. Screenshots may be deceptive because each de-interlaced frame mightn't be identical. I can stop MPC-HC and VLC on the same frames and the MPC-HC version will look better, but advance a frame or two and it's the reverse. It might depend which field is being de-interlaced as the first filed, but when actually playing the video the two methods seem to do a basically identical job to me.
If you resize the video to compare them (ie run them fullscreen) the method used to resize can have a big effect on the straightness of straight lines too.
Here's some samples. The original video is there along with de-interlaced versions using Yadif and QTGMC at both 25fps and 50fps. I'd be surprised if the original and the Yadif 50fps version look all that much different in MPC-HC (although the de-interlaced versions were also resized).
I'd still do something similar to what I suggested in my previous post if you want to edit. Create a script, open it with VirtualDub and encode a lossless, de-interlaced version. Only you'd use QTGMC instead of Yadif. That way you'd still have a lossless de-interlaced copy to edit etc which you can re-encode when you're done.
If you don't have Avisynth installed do so (version 2.6.0 Alpha 5). If you don't have MeGUI installed, give it a spin.
Open the mpeg2 video using MeGUI's File/Open menu and it'll offer to index it and extract the audio. Add the indexing job to the queue and run it. When it's done the script creator window will open with a preview. If there's any cropping, resizing or anamorphic encoding enabled, disable them. Switch to the "Script" tab. Simply copy the following below what's already there.
That'll give you 50fps. If you only want 25fps, add the following instead.
Save the script to your hard drive and close the script creator (a second preview might open, if so, just close it).
Open the saved script with VirtualDub. Use the method I described previously to save the video as a lossless AVI (add the audio MeGUI extracted under VirtualDub's audio menu). With QTGMC in the script everything will be slow. QTGMC is slow.
Attached are two files. Put the ones inside "QTGMC.zip" in the Avisynth plugins folder. You might want to create a shortcut to the html file for future reference. The "QTGMC 32-bit Plugins.zip" file contains mostly files which also go into the Avisynth plugins folder, but there's instructions for the others. If you get an error when trying to run QTGMC in a script, you've probably put the plugins in the wrong place.