High Quality HDV (MPEG-2 1440x1080@50i) to WVC1 encoding?

[fico99]

Hi everyone,

I would like to convert my HDV (MPEG-2 1440x1080@50i) recordings from my Sony HDR-HC7E camcorder into the WMV-HD format using the VC-1 codec with the highest quality possible.

Can someone please tell me how I can do that? I tried Windows Media Encoder using the WMV9 Advanced Profile, but the picture quality and smoothness of the original HDV (MPEG-2) file is still far superior than the encoded WMV-HD :(

Thanks in advance,
Fico99

[Atak_Snajpera]

If you want high quality you should use x264 codec instead that crappy VC-1.

[fico99]

If you want high quality you should use x264 codec instead that crappy VC-1.

Why is VC-1 crappy compared to x264?

VC-1 in 1080p or 1080i plays smoother on a XP or Vista Media Center PC because you don't have to install any extra CODEC's and it does not consume as much CPU power as x264.

Anyway, I prefer VC-1 for Vista Media Center and x264 for my PS3 ;)

Untill now, I only got good results with the conversion of HDV (MPEG-2, 1440x1080@50i) to AVCHD using Nero Vision 4 or 5. It seems that I can't find the correct parameters for Windows Media Encoder using the WMV9 Advanced Profile and x264.exe to convert my HDV footage properly without any hickups or interlacing artefacts :(

Please help.

[Atak_Snajpera]

VC-1 in 1080p or 1080i plays smoother on a XP or Vista Media Center PC because you don't have to install any extra CODEC's and it does not consume as much CPU power as x264.

1080p with FFDShow installed will play smooth on Dual Core.

Untill now, I only got good results with the conversion of HDV (MPEG-2, 1440x1080@50i) to AVCHD using Nero Vision 4 or 5. It seems that I can't find the correct parameters for Windows Media Encoder using the WMV9 Advanced Profile and x264.exe to convert my HDV footage properly without any hickups or interlacing artefacts

Use RipBot264. Select CQ mode@22 (1pass - constant quality) and don't forget to enable deinterlace Top Field First -> 25 fps .

[fico99]

Thanks Atak_Snajpera :)

I will give it a try ;)

[fico99]

I just tried to convert an HDV sample with RipBot264 and I still have the following problems:

- 16:9 Aspect ratio is lost (should be 1440x1080 with SAR = 1,778:1)
- Deinterlacing is not OK (video playback is not smooth during panning shots)

Why do I have to deinterlace my 1080i50 HDV recordings anyway? The AVCHD standard (see http://en.wikipedia.org/wiki/AVCHD) is also based on 1080i. See also this URL about the different AVCHD formats: http://www.avchd-info.org/format/index.html

When I use Nero Vision 4, the converted HDV -> AVCHD plays fine, both on my XP Professional PC as well as on my PS3. If I export to an AVCHD folder structure to a Storage Card from within Nero Vision, I can even play the 1080i50 converted footage @50Hz directly from my PS3.

Disadvantage of Nero Vision is that you don't have as much quality settings as with x264.exe.

[Atak_Snajpera]

16:9 Aspect ratio is lost (should be 1440x1080 with SAR = 1,778:1)

Resize to Full HD 1920x1080 !

- Deinterlacing is not OK (video playback is not smooth during panning shots)

Use 50 fps then! Top Field First -> 50 fps .

Why do I have to deinterlace my 1080i50 HDV recordings anyway?

Because compression ( better quality) is more efficient in progressive mode.

[fico99]

OK, thanks for the update.

Currently, I am trying with:

- no deinterlacing
- resize to Full HD 1920 x 1080

It is still converting, but very slowly...

EDIT: the encoded MP4 with "no deinterlacing" does not play as good as the AVCHD encoded video. File size is almost as big as my original MPEG-2 file.

I am now trying again with the settings:

- Top Field First -> 50fps
- Resize to Full HD 1920 x 1080

and try the playback of both encoded files on my PS3 :rolleyes:

[Atak_Snajpera]

As you can see, it is still converting, very slowly...
High Quality = always slow encoding :) Buy quad core :)

[fico99]

High Quality = always slow encoding :) Buy quad core :)

Yeah right :D

[CruNcher]

Hi everyone,

I would like to convert my HDV (MPEG-2 1440x1080@50i) recordings from my Sony HDR-HC7E camcorder into the WMV-HD format using the VC-1 codec with the highest quality possible.

Can someone please tell me how I can do that? I tried Windows Media Encoder using the WMV9 Advanced Profile, but the picture quality and smoothness of the original HDV (MPEG-2) file is still far superior than the encoded WMV-HD :(

Thanks in advance,
Fico99

I think you mean the Subjective Sharpness is better and yeah that's a tricky thing to keep (especialy with H.264) when compressing (compressed source) again, what you wan't are transparent visual results (without the use of loosless encoding modes) and for that you need to visualy know the encoder inside out (all it's tools,internal features their visual characterstics and it's current Look and Feel). Neros Encoder is allready optimized for that kind of results that's why you feel it looks more true to your source :).


Resize to Full HD 1920x1080 !



Use 50 fps then! Top Field First -> 50 fps .



Because compression ( better quality) is more efficient in progressive mode.


Huh i allways belived Interlacing saves bandwidth ;)

[Atak_Snajpera]

Huh i allways belived Interlacing saves bandwidth

x264 + interlace switch is less effective than x264 + progressive in terms of compressibility. Read careffuly.

[CruNcher]

Yep that's right tough other Encoder are more efficient in this case deinterlacing and transcoding will most probably only result in a heavy speed lose (in the encoding chain) todays realtime per pixel adaptive deinterlacers are pretty efficient (Nvidia,Ati), but ok as X264 workarround he should use a good deinterlacer then @ least yadif or even a more advanced (slower) deinterlacer.
Still it wouldn't change anything about what i said the end result would not look as true to the source as Neros Encoder result, tough im also not sure what he really aims for a good lossy encode, a lossy visual transparent, or visual lossless encode as all of these are different things again :)

[fico99]

Resize to Full HD 1920x1080 !

Use 50 fps then! Top Field First -> 50 fps .


Atak_Snajpera, I tried to play both encoded files (with or without deinterlacing) on both my laptop (with Intel Core2 Duo, T7100@1.80GHz processor and 2GB RAM) and my PS3.

The result is very bad on both. The same HDV file encoded to AVCHD (MPEG-4/AVC 1440x1080@50i) with Nero Vision 4 plays perfect. Even better, Nero Vision 4 is able to create an an "AVCHD DVD" or an AVCHD file structure on a Storage Card or an External Hard Disk Drive which is also recognized by the PS3 as an "AVCHD DVD". The advantage of this is that the PS3 can now play this with a refresh rate of 50Hz on my HDTV (just like with DVD's or BD's). Video's that are started directly from the PS3's XMB interface always play @ 60Hz only which is very bad for 25fps or 50Hz based video. The quality of the Nero AVCHD encoder for this kind of footage is really superb. I think I will stick with Nero Vision for a while now untill there is another alternative.

Thank you very much for your help.

[Atak_Snajpera]

New version will allow you to export directly to BluRay structure. Should be ready by midnight.

[benwaggoner]

I would like to convert my HDV (MPEG-2 1440x1080@50i) recordings from my Sony HDR-HC7E camcorder into the WMV-HD format using the VC-1 codec with the highest quality possible.

Can someone please tell me how I can do that? I tried Windows Media Encoder using the WMV9 Advanced Profile, but the picture quality and smoothness of the original HDV (MPEG-2) file is still far superior than the encoded WMV-HD
What's youre workflow? .avs source? Windows Media encoder?

Also, what file size are you shooting for? You only mentioned quality as your target.

If you're using AVISynth, you'll definitely want to turn on deblocking/deringing in the decoder no matter your output codec. It'll reducing propogation of artifacts.

Assuming WME or other Format SDK like app, you'll want to use settings like:

Compression > General
Video: WMV 9 Advanced profile
Allow interlaced processing=on
Allow nonsquare pixel output=on
Video size: Same as video input
Frame rate: 25 fps
Key frame interval: 4?
Video bitrate: 20 Mbps? (you can go lower with VBR, certainly)
Buffer Size: 4
Video Smoothness: 90

Video Size:
Method: No cropping
Pixel Aspect Ratio:Anamorphic 16:8 (4:3)

Processing:
Maintain Interlacing: Top first (Top Field First in PAL HDV, right?)

If you're worried about speed, you can probably turn down complexity from the default of 4 to 3 without much quality hit.

And most importantly, install the WMV9 PowerToy:
http://www.citizeninsomniac.com/WMV/

And set these:
Perceptual Option: Conservative
In-Loop Filter: On
Overlap Filter: On
Motion Search Range: Adaptive
B-Frame Nubmer: 1
Lookahead: 16
Video Type: Auto Frame/Field/Progressive

And if you care about quality more than time, you can:
Complexity: 4
Motion Search Level:Fixed True Chroma
Motion Match Method: Adaptive

[benwaggoner]

x264 + interlace switch is less effective than x264 + progressive in terms of compressibility. Read careffuly.
Does it really double compression efficiency (so you need less than 1/2 the bits per pixel in progressive than interlaced)? Any of VC-1, MPEG-2, and H.264 should be able to do substantially better than that!

Assuming a decent bob on playback, 1080i25 should provide better quality on playback than 1080p50 until you get to really high bitrates (where you're mainly comparing the quality of the deinterlacing prefilter to a deinterlacing postfilter).

[fico99]

What's youre workflow? .avs source? Windows Media encoder?

Also, what file size are you shooting for? You only mentioned quality as your target.

If you're using AVISynth, you'll definitely want to turn on deblocking/deringing in the decoder no matter your output codec. It'll reducing propogation of artifacts.


Hi Ben,

Thank you very much for your feedback :)
This is EXACTLY the information that I was looking for :)

My current workflow is Windows Media Encoder, but it has problems decoding the original HDV file using FFDShow. Strange, because my Windows Media Player 11 plays my HDV source file just fine using FFDShow but when I try to convert it with Windows Media Encoder, FFDShow is also launched, but when WME wants to start with the 2nd pass the conversion suddenly stops and the output WMV-file is only 6KB :confused:

HDV PAL is indeed Top Field First and 1440x1080@50i.

Rgds,
Fico

EDIT: I think I have found the root cause for the Media Encoder problem. It seems that Haali Media Splitter caused WME to crash.
EDIT2: I do not manage to convert MPEG-2 to WMV with WME. Windows Media Encoder always stops when it wants to start with the 2nd pass???? Is FFDShow not compatible with WME??? What MPEG-2 decoder should I use with WME???? Please help!

[benwaggoner]

Sounds like some kind of FFDshow bug, then.

Do you need to do a VBR encode? You mentioned caring mainly about quality, so you could do a 1-pass CBR, and see if that works better for you.

Not sure if your goal is to make the file smaller, or just playable on Xbox.

[fico99]

Sounds like some kind of FFDshow bug, then.

Could be, when I disable FFDShow and use my Nero Video decoder instead, I got exactly the same problem :(

Which MPEG-2 decoder should I use best in combination with WME9?


Do you need to do a VBR encode? You mentioned caring mainly about quality, so you could do a 1-pass CBR, and see if that works better for you.

Not sure if your goal is to make the file smaller, or just playable on Xbox.

My goal is to compress my large HDV MPEG-2 files (10GB/hour) without loosing quality and to play from within both XP Media Center as well as Vista Media Center.

[benwaggoner]

Could be, when I disable FFDShow and use my Nero Video decoder instead, I got exactly the same problem :(

Which MPEG-2 decoder should I use best in combination with WME9?
Sheesh, not sure. I use Rhozet Carbon for this kind of stuff, which has a built-in MPEG-2 decoder.

AVISynth would be another approach that has worked for me in the past.

My goal is to compress my large HDV footage without loosing quality and to play from within both XP Media Center as well as Vista Media Center.
Gotcha. CBR is fine to make it playable, but VBR is what you want to reduce file size.

[fico99]

OK, I just figured out that I can use DVBPortal WMV Muxer/Encoder instead of that crappy Windows Media Encoder 9 :)

I am now encoding my HDV sample of 200MB to a WMV-HD with WMV9 Advanced Profile CBR 1-pass 12Mbps, buffer size 1sec, Video smoothness 100, video size same as video input, key frame interval 1 sec, allow interlaced processing and allow nonsquare pixel output.

[fico99]

OK, I managed to convert my HDV 1080i50 footage to interlaced WMV-HD using the WMV9 Advanced Profile but when I play this WMV in Windows Media Player 11 I got a lot of deinterlacing artefacts?

Can Windows Media Player 11 do deinterlacing on interlaced WMV-HD videos or do I need a special display adapter or video driver to do that? Currently I am testing the playback on my HP notebook which is using the "Mobile Intel(R) 965 Express Chipset" display adapter.

I still got the best results with Nero Vision 4 or greater by converting my HDV files to the AVCHD format. The only disadvantage of the AVCHD format is that it is using a special BDMV file structure that can only be read by my PS3 and Nero Showtime 3 or greater.

[benwaggoner]

OK, I managed to convert my HDV 1080i50 footage to interlaced WMV-HD using the WMV9 Advanced Profile but when I play this WMV in Windows Media Player 11 I got a lot of deinterlacing artefacts?

Can Windows Media Player 11 do deinterlacing on interlaced WMV-HD videos or do I need a special display adapter or video driver to do that? Currently I am testing the playback on my HP notebook which is using the "Mobile Intel(R) 965 Express Chipset" display adapter.
You may not have deinterlacing turned on in the decoder. Did you install WMV9 PowerToy? If so, go to the Decoder pane there and turn the Internal Deinterlacer to On. You'll also want to have DXVA turned on in WMP 11, with current video drivers installed that support VC-1 AP DXVA.

[fico99]

You may not have deinterlacing turned on in the decoder. Did you install WMV9 PowerToy? If so, go to the Decoder pane there and turn the Internal Deinterlacer to On. You'll also want to have DXVA turned on in WMP 11, with current video drivers installed that support VC-1 AP DXVA.

Ben, I have enabled the option "deinterlacing" using the WMV9 PowerToy and enabled "DirectX Video Acceleration" in WMP11, but the interlacing artefacts are still there :(

Maybe my display adapter is not powerfull enough? I will ask my father to try my encodings on his XP MCE PC with an nVidia GeForce 7600GT and nVidia's PureVideo HD codec.

Please note that I am able to play my original HDV footage just fine on the same PC, using WMP11 + FFDShow ('Cubic Interpolation' deinterlace filter enabled) without interlacing artefacts.

[benwaggoner]

Ben, I have enabled the option "deinterlacing" using the WMV9 PowerToy and enabled "DirectX Video Acceleration" in WMP11, but the interlacing artefacts are still there :(

Maybe my display adapter is not powerfull enough? I will ask my father to try my encodings on his XP MCE PC with an nVidia GeForce 7600GT and nVidia's PureVideo HD codec.

Please note that I am able to play my original HDV footage just fine on the same PC, using WMP11 + FFDShow ('Cubic Interpolation' deinterlace filter enabled) without interlacing artefacts.
Your setup sounds fine to me. My laptop has a 7600 Mobility and works fine with this. Although it is running Vista, and IIRC you're on XP...

Zambelli in a former life was guru of DXVA; perhaps he'll have a suggestion.

Any chance you could post a short sample of your video?

[fico99]

Any chance you could post a short sample of your video?

I have a very good HDV sample of 212MB with a very difficult scene to test. Unfortunately, I see no way how I can send you this sample because my website only allows 50MB of storage capacity.

If you want I can send you the HDV sample using yousendit.com in chunks of 50MB each?

I tried to cut the sample with HDVSplit (using GOP accuracy) to cut the MPEG-2 stream without altering the original stream, but then the scene that I want to send to you is still 138MB big.

[benwaggoner]

I have a very good HDV sample of 212MB with a very difficult scene to test. Unfortunately, I see no way how I can send you this sample because my website only allows 50MB of storage capacity.

If you want I can send you the HDV sample using yousendit.com in chunks of 50MB each?

I tried to cut the sample with HDVSplit (using GOP accuracy) to cut the MPEG-2 stream without altering the original stream, but then the scene that I want to send to you is still 138MB big.
Sure. Or just pick a <50 MB chunk with that hard scene in it.

[fico99]

Sure. Or just pick a <50 MB chunk with that hard scene in it.

Ben, I am currently sending 3 RAR parts of approx. 49MB each using yousendit.com containing an HDV sample of 138MB with ideal material to test.

I will post the high speed download links here ASAP.

[fico99]

Ben, here are the direct download links of my HDV sample:

HDV Sample - Part 1 of 3
http://download.yousendit.com/2756D4D3018DCBAB

HDV Sample - Part 2 of 3
http://download.yousendit.com/E1DB20844DAA553A

HDV Sample - Part 3 of 3
http://download.yousendit.com/8144C9A705DEA35E

Thanks in advance,
Fico

[benwaggoner]

Ben, here are the direct download links of my HDV sample:

HDV Sample - Part 1 of 3
http://download.yousendit.com/2756D4D3018DCBAB

HDV Sample - Part 2 of 3
http://download.yousendit.com/E1DB20844DAA553A

HDV Sample - Part 3 of 3
http://download.yousendit.com/8144C9A705DEA35E

Thanks in advance,
Fico
Downloading now. I'll try to get to it later this week (good excuse to get all the tools working on my new 8-core workstation, for which I am suffused with Nerd Love).

[fico99]

Thanks Ben :)

I am looking forward to read the results.

WMV-HD is really one of my favourite video formats because it plays on all Windows PC's without the need of installing exotic CODEC's. I also have the experience that you don't need as powerfull hardware to play WMV-HD compared with MPEG-4/AVC.

[Atak_Snajpera]

WMV-HD is really one of my favourite video formats because it plays on all Windows PC's without the need of installing exotic CODEC's. I also have the experience that you don't need as powerfull hardware to play WMV-HD compared with MPEG-4/AVC.

I would rather use Xvid instead of VC-1 crap. In my opinion WMV is exotic. Most people/players these days use H.264 standard. (iPOD,iPhone,PSP,PS3,BluRay players, Flash...)

[fico99]

I would rather use Xvid instead of VC-1 crap. In my opinion WMV is exotic. Most people/players these days use H.264 standard. (iPOD,iPhone,PSP,PS3,BluRay players, Flash...)

Atak_Snajpera,

With all respect, I challenge you to download my HDV sample and try to convert the video to Xvid or H.264 and try to play that video as smooth as the original HDV sample.

I have tried everything and I can't get it right.

The only acceptable solution so far was an AVCHD DVD converted/created with Nero Vision 4 or 5 and playback on my PS3. Unfortunately, this AVCHD DVD does not play smooth on any PC I have tried :(

Since ALL of my 1080p WMV-HD's with WVC1 video and WMA10 Professional 5.1 audio do playback very smoothly on ANY of the PC's I have tried, I have set my hope on the WMV-HD format.

If I could playback my AVCHD DVD's smoothly on the PC, that would a good solution too. But how? I don't have to buy a Quad Processor with a Geforce 9600 GT just for AVCHD playback, right :confused:

[Sagittaire]

I would rather use Xvid instead of VC-1 crap. In my opinion WMV is exotic. Most people/players these days use H.264 standard. (iPOD,iPhone,PSP,PS3,BluRay players, Flash...)

Well I think that vc1 produce really better result than XviD today.

[Atak_Snajpera]

But how? I don't have to buy a Quad Processor with a Geforce 9600 GT just for AVCHD playback, right
All you need is Dual Core @~2.5Ghz.

Xvid ~5Mbps 1440x1080 25fps plays without problems on my ancient laptop Turion 1.8 Ghz (single core)
http://www.mediafire.com/?x0dey9tx0ix

[obieobieobie]

I tried encoding his sample with megui and x264 with this script:


movie = mpeg2source("F:\test-2007_10_21-10_57_18.d2v")
LeakKernelDeInt(movie,order=1,sharp=true)
LanczosResize(1920,1088)
RemoveGrain(mode=2)
ColorMatrix("Rec.709->Rec.601",opt=0,hints=false,threads=0)

videoprofile in megui: 1p-goodquality, everything at default
did not encode audio.

here is the file (http://www.keepmyfile.com/download/90c6fb2205225) (~5 MB, ~1 Mbps)

[Atak_Snajpera]

1080p @ 1Mbps??? Are you insane?!?!?!?!

[obieobieobie]

1080p @ 1Mbps??? Are you insane?!?!?!?!

It was a deinterlacing test, since that seemed to be fico's biggest problem.

You are coming off as belligerent, imho.

[Ro-Co]

Hi everybody, I am Fico's father and following this thread with great interest.
I will download obieobieobie 's file and try it tomorrow on my dedicated HTPC powered by AMD X2 4200+ and NVidia 7600GT. This stuff connected to a Toshiba 52z3030d.
I hope you coding specialist can find the ultimate solution for us, I myself spent already days trying every codec and profile I could get my hands on.
To Atak_Snajpera: we don't rule out any satisfying codec, so convince us and try to code the files Fico posted.

[obieobieobie]

Redid the test, same script except for the resizing which is now 1440,1088 (to keep it mod16 and anamorphic)

megui settings are based on 1p-goodquality except that it is now constant quant=22 and with mixed reference frames enabled, everything else is default

resulting file is about 60 MB at about 10 Mbps, which is probably overkill, audio would add 1-2 MB

download here (http://www.keepmyfile.com/download/3c13142205232)

I think leak's kernel deinterlacer works well with this clip

[Dethis]

........
..
I think leak's kernel deinterlacer works well with this clip

While combing artifacts are perfectly absend there is too much studering (all vertical lines & leaves while panning).

Encoding interlaced and bob-deinterlacing during playback to 50fps gives perfect fluid motion.

[Atak_Snajpera]

To Atak_Snajpera: we don't rule out any satisfying codec, so convince us and try to code the files Fico posted.

Xvid ~5Mbps 1440x1080 25fps plays without problems on my ancient laptop Turion 1.8 Ghz (single core)
http://www.mediafire.com/?x0dey9tx0ix

Did you see it?

[obieobieobie]

Mediafire is just too slow. Half the time it won't start the download, and when it does start, it goes at 2 KB/s.

Could you re-upload it to keepmyfile or yousendit or something?

I would like to compare the clip with my x264 encode. I have to agree with Dethis that my clip stutters too much, but I thought it was because I watched it on a 60Hz display

[fico99]

Hi guys, thank you so much for all your efforts :)

Since I am currently on holiday in Italy and using my laptop on a limited internet connection, I only tried the first post (the 5MB video file).

The first thing I noticed about the de-interlaced video is that it is stuttering during the horizontal pan, which is exactly the problem that I was talking about all the time. For those who don't see this, pay attention to the leaves of the trees during the panning shot. They start to flicker when you de-interlace. It doesn't look natural anymore.

As I have said before, my only successfull recode so far was the conversion to the AVCHD DVD format using Nero Vision with playback on the PS3. On the PC, I didn't find a solution yet (unless keeping the original MPEG-2 format).

I will try the other test files as soon as I am back from holiday (1 april). In the mean time, my father will be more than happy to test them out on his hardware :)

[fico99]

I have to agree with Dethis that my clip stutters too much, but I thought it was because I watched it on a 60Hz display

Does the original HDV sample also stutter on your 60Hz display?

[obieobieobie]

Does the original HDV sample also stutter on your 60Hz display?

Yes, in VLC but not in PowerDVD 7.

[Ro-Co]

I downloaded the 3 files and none of them plays fluently on my system. They all suffer from the flickering effect while panning. I retried the original clip in HDV format and this plays perfectly without any stutter. The only artifact I can see is a slight motion blur on the individual leaves.
Anyway, I thank you guys for your time and expertise and I am anxious to try a sample from Microsofts Guru Ben Wagoner.

@obieobieobie: I play the original HDV clip in WMP, but you have to disable Haali splitter for MPEG TS streams else it starts up and causes stutter. The HDV container must use some kind of internal splitter I suppose.

@Atak_Snajpera: sorry I didn't see the download link at first but after your message I was able to dowload your test file at a reasonable speed.

[pandy]

i prefer Avisynth + Yadif() as a result i receive 1440x1080p50 then resize this to the 1280x720p50 and encode as You prefer (VC1 or h.264) also You may try to encode 1440x1080i25 as a interlace with SAR 4:3

[Ro-Co]

i prefer Avisynth + Yadif() as a result i receive 1440x1080p50 then resize this to the 1280x720p50 and encode as You prefer (VC1 or h.264) also You may try to encode 1440x1080i25 as a interlace with SAR 4:3

This seems not logical at all to me, you deinterlace and downscale whereas afterwards my screen has to upscale it again to 1080p. You're second approach seems better, maintaining the original resolution and let the TV do the deinterlacing, I tried this in vain with different codecs, I get always these flickering in horizontal pans.

[nm]

This seems not logical at all to me, you deinterlace and downscale whereas afterwards my screen has to upscale it again to 1080p.
720p50 is a commonly used HDTV video mode in Europe along with 1080i25 and 1080p24 (on Blu-ray/HD DVD), 1080p50 is still rare and it's not supported by most hardware decoders. Also decoding 1080p50 with software is about twice as hard on the CPU as 720p50. So, if you want to deinterlace properly while maintaining compatibility with current hardware and software solutions, you should use 720p50.

You're second approach seems better, maintaining the original resolution and let the TV do the deinterlacing, I tried this in vain with different codecs, I get always these flickering in horizontal pans.
Well, your TV won't deinterlace unless you use an interlaced display mode (1080i25), and even then it will probably only work through HDMI input with a standalone player since PC video players and graphics cards have a hard time keeping the fields of interlaced video in order.

When playing interlaced video on a PC through VGA/DVI/HDMI output of a normal graphics card, you'll need to deinterlace. If that doesn't work, the problem is in your decoder or video player (or display driver / graphics card, when using hardware deinterlacing). Remember that many (perhaps most?) people will not have these things set up correctly, so giving interlaced video files to PC users is usually a bad idea. AVCHD may work in some cases, but people that don't have PowerDVD Ultra or other Blu-ray player software will have problems with interlacing.

The flickering that you describe is probably caused by half-rate deinterlacing. If you use hardware deinterlacing, make sure that it is set up properly to a full-rate "vector-adaptive" or "pixel-adaptive" mode by using the setup utility of your display driver. Also make sure that the encoder is actually set to encode as interlaced.

[Ro-Co]

I know it's confusing, but broadcast standard in Europe is 1080i50 (fields/second not frames/second) I switch the resolution in my geforce driver to 1080p24 for BD or HDDVD rips, 1080i50 for PAL and 1080i60 for NTSC sources. In the first case my screen does a 5:5 pulldown and displays at 120Hz which gives a perfectly smooth image.
Our camera's have a 1080i substandard resolution of 1440X1080 (also often used by satellite) @50 fields/second.
Deinterlacing is normally done by PureVideo.
I saw some movies in the 1440X1080i format on Usenet servers and could try these, but unfortunately these are all porn and I don't want these on my PC.
And ofcourse my HTPC is connected to the screen via HDMI.

[nm]

I know it's confusing, but broadcast standard in Europe is 1080i50 (fields/second not frames/second)
We are talking about the same thing here. It can be written either 1080i50 or 1080i25 (http://en.wikipedia.org/wiki/1080i), with 25 meaning frames/second and you can infer the 50 fields/s from that.

Our camera's have a 1080i substandard resolution of 1440X1080 (also often used by satellite) @50 fields/second.
Yes, no doubt about this.

Deinterlacing is normally done by PureVideo.
If the original source video can be played back without flickering, but your interlaced H.264 or VC-1 encodings show the problem when played back with the same player, there is either a problem in the player or decoder software, or the video wasn't encoded properly as interlaced (you need to manually tell the encoder to do this).

[Ro-Co]

If the original source video can be played back without flickering, but your interlaced H.264 or VC-1 encodings show the problem when played back with the same player, there is either a problem in the player or decoder software, or the video wasn't encoded properly as interlaced (you need to manually tell the encoder to do this).

Now we are talking, but where's the clue, what are we doing wrong???

[Atak_Snajpera]

i prefer Avisynth + Yadif() as a result i receive 1440x1080p50 then resize this to the 1280x720p50 and encode as You prefer (VC1 or h.264) also You may try to encode 1440x1080i25 as a interlace with SAR 4:3

This seems not logical at all to me, you deinterlace and downscale whereas afterwards my screen has to upscale it again to 1080p. You're second approach seems better, maintaining the original resolution and let the TV do the deinterlacing, I tried this in vain with different codecs, I get always these flickering in horizontal pans.

If you encode material from HDV you should know that 1440x1080i25 is in fact 1440x540p50. So decoder has to resize 1440x540 picture to 1920x1080. (1 296 000 pixels have to be interpolated.)

[scharfis_brain]

that 1440x1080i25 is in fact 1440x540p50
This is simply not true.
Cause even and odd fields of the video represent different spatial positions.

If the whole image is static, nothing will have to interpolated from interlaced to progessive.

This is the trick with interlacing: static areas (and areas that move 2pixels/frame vertically) retain the FULL frame resolution.
Only areas with other movement need interpolation.

[Atak_Snajpera]

This is the trick with interlacing: static areas (and areas that move 2pixels/frame vertically) retain the FULL frame resolution.

All right I agree but if your camcorder is shacking like in example you get only 1440x540.

[pandy]

If you encode material from HDV you should know that 1440x1080i25 is in fact 1440x540p50. So decoder has to resize 1440x540 picture to 1920x1080. (1 296 000 pixels have to be interpolated.)

Partially yes - we need to interpolate 720, but IMHO temporal resolution is much more important than spatial (most of cheap HDV cam have rather poor spatial resolution so forcing 1920x1080i25 or p25 over p50 is not a good idea - we loose to much).

by using good interpolation algorithm subjective quality such video (1440x1080i25 >>1280x720p50) can be better than original source (especially when we think that most of TV displays dont have 1920x1080 but 1368x768 or even worse spatial resolution)

poorly done deinterlacing can be visualy worse than bilinear rescaling of 720p to the 768p or 1080p

[pandy]

This seems not logical at all to me, you deinterlace and downscale whereas afterwards my screen has to upscale it again to 1080p.

If You deinterlace Your source from 1440x1080i25 to the 1440x1080p25 then You loose temporal resolution - in worst case You loose almost half of information stored inside video...

You're second approach seems better, maintaining the original resolution and let the TV do the deinterlacing, I tried this in vain with different codecs, I get always these flickering in horizontal pans.

But most of TV deinterlace source in one way - by using BOB - they do almost the same as in first method but with rather worse quality (probably use bilinear interpolation, sometimes spline or bicubic). Also You must remeber that TV crop borders (crop overscan) and resize video after croping to the native resolution of display.

[scharfis_brain]

I tried to convert it to 1280x720p50:
http://rapidshare.com/files/102327504/test-2007_10_21-10_57_18-deshaked-720p.mkv.html

I used
- nnedi(field=-2) for bob-deinterlacing
- tweak(bright=-16, coring=false) for highlight recovery
and
- deshaker for stabilizing and resizing the video to 1280x720

[scharfis_brain]

I've done another encoding using deshakers "rolling shutter" option:
http://rapidshare.com/files/102347441/des-720p.mp4.html

[benwaggoner]

Okay, my head is swimming from all this. What's the ask?

Anyway, likely culprit when you get motion artifacts in interlaced content after compression is getting fields in the wrong order, like top field first getting flagged as bottom field first. PAL HDV is top field, right?

[Atak_Snajpera]

Partially yes - we need to interpolate 720, but IMHO temporal resolution is much more important than spatial (most of cheap HDV cam have rather poor spatial resolution so forcing 1920x1080i25 or p25 over p50 is not a good idea - we loose to much).

by using good interpolation algorithm subjective quality such video (1440x1080i25 >>1280x720p50) can be better than original source (especially when we think that most of TV displays dont have 1920x1080 but 1368x768 or even worse spatial resolution)

poorly done deinterlacing can be visualy worse than bilinear rescaling of 720p to the 768p or 1080p

I also Deinterlace HDV source with Yadif and then resize to 1280x720 + Sharpen(0.2)

[fico99]

PAL HDV is top field, right?

That's right.

Here are my results so far:

- Original HDV PAL (MPEG-2 1440x1080 @ 50Hz interlaced) plays smooth on both PC and PS3.
- Converted HDV PAL -> AVCHD PAL DVD (MPEG-4/AVC 1440x1080 @ 50 Hz interlaced) plays smooth on my PS3 but NOT on the PC.

So, my father and I are hoping that the WMV-HD format can offer smooth playback of our HDV PAL footage on the PC without loosing too much quality.

We could store everything permanently in the native HDV PAL format but the problem with this format is that it is storage-consuming (25Mbit/s) and it does not allow you to rewind/fast forward smoothly.
More info: http://en.wikipedia.org/wiki/HDV

[Dethis]

......
- Converted HDV PAL -> AVCHD PAL DVD (MPEG-4/AVC 1440x1080 @ 50 Hz interlaced) plays smooth on my PS3 but NOT on the PC.

..........
More info: http://en.wikipedia.org/wiki/HDV

You should really check CoreAVC 1.65 decoder configured @ bob-deinterlacing

[benwaggoner]

I did a quick WMV encode as tff @ 5 Mbps (2-pass VBR) and it played back great for me.

Make sure you encode as top field first. That's likely your problem.

[fico99]

I did a quick WMV encode as tff @ 5 Mbps (2-pass VBR) and it played back great for me.

Make sure you encode as top field first. That's likely your problem.

Ben, we always did our WMV encodes with top field first. Did you use the WMV9 Advanced Profile (aka. WVC1) with the option maintain interlacing? Is the horizontal panning shot in your WMV encode as smooth as in the original HDV sample?

Can you please send us a sample (with yousendit) of your WMV encode, so we can try it out on our hardware?

[Ro-Co]

I've done another encoding using deshakers "rolling shutter" option:
http://rapidshare.com/files/102347441/des-720p.mp4.html

Congratulations, you made the first encodings that are acceptable. I will analyze the files a little better today, what's you're secret? Is it this deshaker?

[nm]

Congratulations, you made the first encodings that are acceptable. I will analyze the files a little better today, what's you're secret? Is it this deshaker?
The secret is deinterlacing properly.

[fico99]

Congratulations scharfis_brain, I will try your encodes as soon as I am back home :)

Thanks man :)

[Atak_Snajpera]

There is no magic ... The secret is 50 fps instead of 25 fps :) But 50 fps means higher bitrate = higher CPU utilization.

[nm]

I made a couple more (http://www.cs.helsinki.fi/u/mikkila/video/hdvtest). 720p50 and 1080p50 deinterlaced with yadif+mcdeint in MEncoder, 1080i25 encoded as interlaced (x264's MBAFF). These samples may also play with hardware acceleration (DXVA), but probably not on other standalones than PS3 (and you'll need to remux to MP4 or MPEG-TS).

If the 1080i25 sample plays with slight judder compared to 720p50, it's probably caused by poor playback deinterlacing (half-rate instead of pixel-adaptive full-rate). If it judders or flickers vigorously, as in being completely unwatchable, the problem is caused by field order. Anyway, 720p50 is probably the most compatible and trouble-free alternative in this case.

[Ro-Co]

You should really check CoreAVC 1.65 decoder configured @ bob-deinterlacing

I have my reserves about installing codecs on my HTPC, as they can ruin everything and are sometimes hard to remove. My main decoder is NVidia's Purevideo which is considered the best around and it supports HW acceleration of my GPU. When I play HDV it uses this codec.
To decode X-264, H-264, MKV or DiffX I use FFDshow and Haali splitter.
I have no PowerDVD or WinDVD installed as I play everything straight into the MCE interface and controlled by Microsoft remote control.

[Atak_Snajpera]

CoreAVC is very efficient so you should really try it!

[Ro-Co]

I made a couple more (http://www.cs.helsinki.fi/u/mikkila/video/hdvtest[/url). 720p50 and 1080p50 deinterlaced with yadif+mcdeint in MEncoder, 1080i25 encoded as interlaced (x264's MBAFF). These samples may also play with hardware acceleration (DXVA), but probably not on other standalones than PS3 (and you'll need to remux to MP4 or MPEG-TS).


Sorry I get this message on your server:
Forbidden
You don't have permission to access /u/mikkila/video/hdvtest[/url on this server.

[Atak_Snajpera]

1280x720p50 @6Mbps
http://www.mediafire.com/?2840vwyp0ym

[nm]

Sorry I get this message on your server:
Forbidden
You don't have permission to access /u/mikkila/video/hdvtest[/url on this server.
Sorry, there was a typo in the link. I fixed it in the original post.

[nm]

My main decoder is NVidia's Purevideo which is considered the best around and it supports HW acceleration of my GPU.Purevideo Decoder is good since it automatically deinterlaces correctly, but unfortunately it only decodes MPEG-2.

To decode X-264, H-264, MKV or DiffX I use FFDshow and Haali splitter.Do you have hardware deinterlacing set in ffdshow? Otherwise you won't get good quality from interlaced MPEG-4 sources. CoreAVC is also worth trying, especially for HD videos.

[Ro-Co]

I did notice the typo error in the link and tested the 3 files, also Atak_Snajpera's latest.
All the progressive files play quite well (no more flickering). The interlaced file I cannot play properly and sorry I cannot find hardware acceleration in ffdshow. The deshaked file plays astonishly smooth but suffers most from resolution loss. I think this resolution loss (crispness) will be inevitable because we reduce file size so much.
Tomorrow I will try to code myself with your settings straight from Adobe Premiere's timeline.

[scharfis_brain]

My process whas this:
1st pass deshaker:
AVS-Script for 1st pass
setmtmode(2)
loadplugin("c:\x\DGDecode.dll")
loadplugin("c:\x\nnedi.dll")

mpeg2source("F:\forum\fico99\test-2007_10_21-10_57_18.d2v") #,cpu=6)
bob()
tweak(bright=-16,coring=false)
converttorgb(matrix="PC.601")
last+last.reverse().trim(0,50)

I used the folling settings in deshakers 1st pass options:
SAR: HDV 1440x1080 (1.333)
Video type: progressive
[x] Camera has a rolling shutter

for the 2nd pass of deshaker alter the script to use nnedi() interpolation:
setmtmode(2)
loadplugin("c:\x\DGDecode.dll")
loadplugin("c:\x\nnedi.dll")

mpeg2source("F:\forum\fico99\test-2007_10_21-10_57_18.d2v") #,cpu=6)
nnedi(field=-2)
tweak(bright=-16,coring=false) #recover blown out highlights
converttorgb(matrix="PC.601") #convert to RGB without colour distortions (ie Banding)
last+last.reverse().trim(0,50)

The settings for the 2nd pass of deshaker were these:
[ ]Same dest. props as source
Dest pixel aspect: Square pixel (1.0)
Dest. video size 1280x720
Presampling: Bicubic
Edge compensation: None (large borders)
[x] Use prev. & fut. frames to fill in borders
Prev. frames: 50
Future frames: 50
[x] soft borders. Edge transition width: 5 pixels
[x] Extrapolate colors into border.

Motion smoothness
Hor panning: 4000
Vertical Panning 4000
Rotation 400
Zoom 150

Max. corr. limits
Horizontal Panning 15
Vertical Panning 15
Rotation 5
Zoom 1


Save the deshaked video as LOSSLESS RGB (this is important!) Video file.

Use this script to hand over the deshaked RGB-Video to an encoder of your choice:
avisource("des-720p-rolling.avi")
trim(50,0) #trim away deshakers future-frame compensation
converttoyv12(matrix="PC.601") #convert back to YV12 without colour distortions (ie Banding)

[pandy]

PAL HDV is top field, right?

There is no fixed TFF/BFF configuration, many sources have various field order, some even worse - two field organization is mixed beetwen two joined streams (internal edition in camvid case)

[pandy]

I also Deinterlace HDV source with Yadif and then resize to 1280x720 + Sharpen(0.2)

IMHO sharpen before (not after) resizing is quite good solution.

[SeeMoreDigital]

Hi Nm,

Just to let you know, when I played your "test-x264-1080i25-crf25.mkv" sample in my hardware player, the image judders about very badly!

EDIT: The other two samples play fine.

[pandy]

There is no magic ... The secret is 50 fps instead of 25 fps :) But 50 fps means higher bitrate = higher CPU utilization.

Hmmmm, i can't agree - 25 vs 50fps means that more information can be encoded as MV - also due nature of the h.264 such stream should encode much more efficiently than done by h.262 or VC-1.
Due of nature of interlace progressive is encoded much better.

[nm]

Hi Nm,

Just to let you know, when I played your "test-x264-1080i25-crf25.mkv" sample in my hardware player, the image judders about very badly!
That's probably somewhat expected since x264 doesn't signal the field order separately. There is a patch that does this in SEI data, but I'm not sure if that helps either with your hardware (do other MBAFF streams work correctly?). See: http://forum.doom9.org/showthread.php?t=120410

[SeeMoreDigital]

...do other MBAFF streams work correctly?I have various HDTV captured samples that do.

[Atak_Snajpera]

Hmmmm, i can't agree - 25 vs 50fps means that more information can be encoded as MV - also due nature of the h.264 such stream should encode much more efficiently than done by h.262 or VC-1.
Due of nature of interlace progressive is encoded much better.

1280x720
CRF22 25fps = 34 MB ; 50fps = 58 MB

Surprised ? I'm not ...

IMHO sharpen before (not after) resizing is quite good solution.
Would you enlighten me why? Also I use very small value just to enhance fine details.

[scharfis_brain]

if the footage would exhibit no motion blur (eg. short sutter like < 1/1000 sec) then the bitrate won't rise much when goinjg from 25fps to 50fps.

But this video is shakey and has been shot with 1/50 sec shutter.
THis means that there is a lot of random motion blur present.
This motion blur changes from frame to frame making a good and near lossless prediction impossible.

[Atak_Snajpera]

if the footage would exhibit no motion blur (eg. short sutter like < 1/1000 sec) then the bitrate won't rise much when goinjg from 25fps to 50fps.

Probably 90%+ of footage taken from camcorders has motion blur so higher fps = higher bitrate. End of story.

[scharfis_brain]

Nah, nearly all consumer camcorders switch to very short shutter times in bright sunlight when set to Full Automatic mode...

[Atak_Snajpera]

Nah, nearly all consumer camcorders switch to very short shutter times in bright sunlight when set to Full Automatic mode...
and obviously you have always nice weather :)

[scharfis_brain]

Heh, You know what I meant ;)

Of course, in low light situations the cam will choose a slower shutter.
Further a 1/50 (PAL) or 1/60 (NTSC) shutter is preferrable over a short shutter IMO in terms of motion fluidity.

[Atak_Snajpera]

Further a 1/50 (PAL) or 1/60 (NTSC) shutter is preferrable over a short shutter IMO in terms of motion fluidity.
I agree. I've seen tests with tap water (very fast shutter vs slow). Slow shutter version looked natural while fast was like watching photos.

[pandy]

1280x720
CRF22 25fps = 34 MB ; 50fps = 58 MB

Surprised ? I'm not ...


Are You compare 25i vs 50p?


Would you enlighten me why? Also I use very small value just to enhance fine details.

Gibbs effect, Nyquist sampling etc - Sharpen before downsampling enhance "real" edge of sampled object, then resizing downsample "real" samples down - Sharpen after resizing simply differentiate false (resampled) edge of the object. Sharpening before downsampling can be compared to the preemphasis (downsampling act like deemphasis).

ps
pozdrowienia z Polski

[Atak_Snajpera]

Are You compare 25i vs 50p?
p vs p obviously :)

Gibbs effect, Nyquist sampling etc - Sharpen before downsampling enhance "real" edge of sampled object, then resizing downsample "real" samples down - Sharpen after resizing simply differentiate false (resampled) edge of the object. Sharpening before downsampling can be compared to the preemphasis (downsampling act like deemphasis).

eeee I see no difference

http://img291.imageshack.us/img291/1016/720pplussharpen02beforeoe0.th.png (http://img291.imageshack.us/my.php?image=720pplussharpen02beforeoe0.png)

http://img262.imageshack.us/img262/9264/720pplussharpen02qc2.th.png (http://img262.imageshack.us/my.php?image=720pplussharpen02qc2.png)

[Ro-Co]

But this video is shakey and has been shot with 1/50 sec shutter.
THis means that there is a lot of random motion blur present.
This motion blur changes from frame to frame making a good and near lossless prediction impossible.

Aha that's why we chose this clip, we know this is very hard to compress, but if we can get this clip right then the rest will be easy. Also this clip is quite easy to evaluate, much more easy than a waterfall f.i.
Anyway I was impressed by deshaker and will use this algorithm in my future projects.

[SeeMoreDigital]

Are You compare 25i vs 50p?Surely this should be 50i and not 25i?

[scharfis_brain]

both refer to the same thing.
it is just a nomenclature uncertainty.
25i refers to the framerate while 50i refers to the fieldrate, but both are the same...

[nm]

Hi Nm,

Just to let you know, when I played your "test-x264-1080i25-crf25.mkv" sample in my hardware player, the image judders about very badly!

Another try, this time with dattrax's x264_hrd_pulldown.04_interlace.diff.7z (http://forum.doom9.org/showthread.php?p=1047919#post1047919):

http://www.cs.helsinki.fi/u/mikkila/video/hdvtest/test-x264-1080i25-nalhrd.mkv (x264 command-line is here (http://www.cs.helsinki.fi/u/mikkila/video/hdvtest/encode-nalhrd.sh))

Does this work better with your hardware player?

[Ro-Co]

I made a breakthrough playing the interlaced samples. I wrote before that I could not find "HW deinterlace" in ffdshow, well this setting has been replaced by "set interlace flag in output media type" then you can choose between BOB and Weave. Choosing BOB does a good deinterlace, so no more flickering. But at the end of the clip I get picture deformation. I did some more research and found one of my processor cores was completely saturated (total CPU load was 65%). So maybe I wil have to find a way to make them work together 50/50. Anyway I wil try the other codec that was recommended here as it would be less CPU intensive.

[Ro-Co]

Just tested CoreAVC 1.6.5.0 professional and it's a complete disaster. This decoder uses the same CPU load though well balanced between the 2 cores. Deinterlacing simply is a mess, even the colors are wrong, tried every possible setting.

[Dethis]

Another try, this time with dattrax's x264_hrd_pulldown.04_interlace.diff.7z (http://forum.doom9.org/showthread.php?p=1047919#post1047919):

http://www.cs.helsinki.fi/u/mikkila/video/hdvtest/test-x264-1080i25-nalhrd.mkv (x264 command-line is here (http://www.cs.helsinki.fi/u/mikkila/video/hdvtest/encode-nalhrd.sh))

Does this work better with your hardware player?

Fine.
With this last 25i encoding i can properly play it back with coreavc & bob-deinterlace.
The former one, i only could play it properly using coreavc none(weave) & ffdshow kernel bob.
Encodings at 25i with Nero plays perfectly even with coreavc & hardware deinterlacing.

[nm]

What kind of problems do you have with this x264 encode when played back with CoreAVC & hardware deinterlacing?

[Dethis]

What kind of problems do you have with this x264 encode when played back with CoreAVC & hardware deinterlacing?

It plays at 25fps with combing artifacts.

Sorry i was wrong about the difference between nero vs (the last) x264 encodings. They have the same behaviour.
Both play perfectly with hardware deinterlacing but only if i use haalli renderer. If i use VMR9 renderless both need bob-deinterlacing to play back properly.

[SeeMoreDigital]

Another try, this time with dattrax's x264_hrd_pulldown.04_interlace.diff.7z (http://forum.doom9.org/showthread.php?p=1047919#post1047919):

http://www.cs.helsinki.fi/u/mikkila/video/hdvtest/test-x264-1080i25-nalhrd.mkv (x264 command-line is here (http://www.cs.helsinki.fi/u/mikkila/video/hdvtest/encode-nalhrd.sh))

Does this work better with your hardware player?That sample looks much, much better :D

Could you do me a favour please? And use the same technique to encode the following MPEG-2 source: -

http://www.sendspace.com/file/yqthb4


Many thanks

[nm]

Could you do me a favour please? And use the same technique to encode the following MPEG-2 source: -
Sure, here's the result: http://www.cs.helsinki.fi/u/mikkila/video/hdvtest/test2-x264-1080i30.mkv

I used the same encoding parameters as before, except for increasing the GOP size. This source caused some weird skipping with keyint < 32. Perhaps a bug in x264.

[SeeMoreDigital]

Sure, here's the result: http://www.cs.helsinki.fi/u/mikkila/video/hdvtest/test2-x264-1080i30.mkv

I used the same encoding parameters as before, except for increasing the GOP size. This source caused some weird skipping with keyint < 32. Perhaps a bug in x264.That encode works perfectly in my hardware player...

You have definately cracked how to generate AVC encodes from pure interlaced sources :D

[nm]

You have definately cracked how to generate AVC encodes from pure interlaced sources :D
We should thank dattrax and Trahald for their patches, and of course akupenguin and other x264 developers for the interlaced encoding support :)

[benwaggoner]

Ben, we always did our WMV encodes with top field first. Did you use the WMV9 Advanced Profile (aka. WVC1) with the option maintain interlacing?
Yes.

Is the horizontal panning shot in your WMV encode as smooth as in the original HDV sample?
Yes, for me.

Can you please send us a sample (with yousendit) of your WMV encode, so we can try it out on our hardware?
I've got all my upstream bandwidth on another deadline right now. But you can do it yourself easily enough:

AVISynth script:
LoadPlugin("C:\Program Files\AviSynth 2.5\plugins\DGDecode.dll")
MPEG2Source("test-2007_10_21-10_57_18.d2v", cpu=6)

And this batch file for encoding (with Zambelli's WMCmd.vbs):
cscript "C:\Program Files\Windows Media Components\Encoder\WMCmd.vbs" -i "Source_postproc.avs" -o "HDV_test_25i_5M.wmv" -videoonly -v_codec WVC1 -v_mode 4 -v_keydist 4 -v_preproc 17 -v_bitrate 5000000 -v_peakbitrate 25000000 -v_performance 80 -v_pixelratio 4 3 -v_adz 1 -v_loopfilter 1 -v_overlap 1 -v_msrange 0 -v_mslevel 2 -v_mmatch 0 -v_bframedist 1 -v_type 4

The above is tuned for max quality without much concern for speed. If you want to hit a speed/quality balance, I'd suggest these changes. It's be 2-3x faster and probably nearly as good with this source.

-v_performance 60
-v_mslevel 0
-v_mmatch 1
-v_type 3

[CruNcher]

That encode works perfectly in my hardware player...

You have definately cracked how to generate AVC encodes from pure interlaced sources :D

Tough it says nothing about BD compatibility yet only your Hardware player plays it btw what for a SAP is that a Sigma SAP?
also this interlacing is the most unefficients mode available in H.264 it only flags every MB as being interlaced it's no real MBAFF (ofcourse it does it's job but not yet as efficient as it could). The real great thing will be if MBAFF with a efficient heuristic is fully in.

[pandy]

p vs p obviously :)


so You loose almost half data from stream - only i vs p can be reliable comparison.



eeee I see no difference

http://img291.imageshack.us/img291/1016/720pplussharpen02beforeoe0.th.png (http://img291.imageshack.us/my.php?image=720pplussharpen02beforeoe0.png)

http://img262.imageshack.us/img262/9264/720pplussharpen02qc2.th.png (http://img262.imageshack.us/my.php?image=720pplussharpen02qc2.png)

sharpen before downsampling should have slightly better compression

[pandy]

Surely this should be 50i and not 25i?

there is no 50i (or a should say - 50i mean 50 frames per second and 100 fields per second).
I believe that BBC in http://www.bbc.co.uk/commissioning/production/docs/hd_standards.pdf thinking in the same way.
If we write 1920x1080i50 then i50 means 1920x1080 50 frames per second which means 100 fields (btw i know that exists interlace formats where full frame has more than 2 fields ie 3 or even more fields create full frame).

PS
from technical point of view 1280x720i50 (ie 1280x360p100) should be valid format for decoding hardware...

[SeeMoreDigital]

there is no 50i (or a should say - 50i mean 50 frames per second and 100 fields per second).
I believe that BBC in http://www.bbc.co.uk/commissioning/production/docs/hd_standards.pdf thinking in the same way.
If we write 1920x1080i50 then i50 means 1920x1080 50 frames per second which means 100 fieldsI think it makes much more logical sense that "i50" relates to "50 interlaced fields"....

[scharfis_brain]

yeah. like I wrote earlier: 50i and 25i are exactly the same.
It's just that the naming convention isn't clear.
But since there aren't any actual video standards that offer a 12.5 fps framerate or a 100 fps fieldrate, the terms 25i and 50i can be used without much danger.

(the same applies to 30i and 60i)

[SeeMoreDigital]

Indeed....

For the record, I've just had a look at the manuals of some of my "up-scaling" and "full HD" players... When discussing the video output, they use the term "i50/i60".

[benwaggoner]

While I agree that 50i/60i are commonly used phrases, I recommend we as an industry/community standardize on 25i/30i, which clearly indicates the amount of visual data per second.

Also, it makes 720p60 look better on one axis comapred to 1080i30, while 720p60 versus 1080i60 looks like it's just smaller.

[scharfis_brain]

25i like 50i has the 'visual data' of 50fps...

[benwaggoner]

25i like 50i has the 'visual data' of 50fps...
...if you write it as 1920x540i60 :).

Really, I guess it gets down to if we're trying to indicate the number of temporal samples per second, or the number of pixels per second.

From a compression and visual quality perspective, I'd argue that the latter is more important. It's not like there are displays out there that can't display at 50/60 Hz, but plenty that can't display full resolution.

[pandy]

If we write 1920x1080i25 then we have 25 frames(pictures), each with resolution 1920x1080, "i" means that these frames(pictures) are constructed from few (for TV - 2) interlaced fields.

[benwaggoner]

Ooops, just realized I pasted in the wrong version of the WMCmd.vbs script for the VC-1 encode. Here's the corrected one:

cscript "C:\Program Files\Windows Media Components\Encoder\WMCmd.vbs" -input "Source_postproc.avs" -output "HDV_test_25i_5M.wmv" -videoonly -v_codec WVC1 -v_mode 4 -v_keydist 4 -v_preproc 17 -v_bitrate 5000000 -v_peakbitrate 25000000 -v_performance 80 -pixelratio 4 3 -v_adz 1 -v_loopfilter 1 -v_overlap 1 -v_msrange 0 -v_mslevel 2 -v_mmatch 0 -v_bframedist 1 -v_type 4

Anyone give this a try and check out the output yet? I'm pretty pleased myself @ 5 Mbps.

[Ro-Co]

Anyone give this a try and check out the output yet? I'm pretty pleased myself @ 5 Mbps.

I already spent a few HOURS trying to figure out your workflow. It requires a bunch of programs to be installed, which I don't want to do without a guaranteed result.
As far as I understand you use: Avisynth - DGMPGDec codec pack for DGDecode - Zambelli's application to control MCEncode - Net-Framework - VirtualDub???
I encode straight from Adobe Premiere's timeline which has an interface to control the WMV encoder, but the results ar crap.
MCEncode itself crashes each time I open it. This WMV encoding is a real headache.

[benwaggoner]

I already spent a few HOURS trying to figure out your workflow. It requires a bunch of programs to be installed, which I don't want to do without a guaranteed result.
As far as I understand you use: Avisynth - DGMPGDec codec pack for DGDecode - Zambelli's application to control MCEncode - Net-Framework - VirtualDub???
Oh, that's not about the output format. Avisynth+DGMPGDec is all about getting a good adaptive postprocessing MPEG-2 decoder into the chain to improve the source quality. It'd help regardless of your output format.

I encode straight from Adobe Premiere's timeline which has an interface to control the WMV encoder, but the results ar crap.
MCEncode itself crashes each time I open it. This WMV encoding is a real headache.
Can you share the settings you're using? You should be able to replicate much of what I set up in command line with that using the Adobe Media Encoder GUI and Zambelli's WMV 9 PowerToy GUI to set the registry keys.

The only thing you can't do is use more than a 10 Mbps peak bitrate; that's a Premiere Pro GUI limitation. But that shouldn't hurt your quality that much.

[fico99]

Ben,

Would it be possible to send us a small sample of your best VC-1 encode via yousendit.com?
I really would like to see how good your encode is before spending any more time on the WMV-HD solution.

Thanks in advance.

[Ro-Co]

Can you share the settings you're using? You should be able to replicate much of what I set up in command line with that using the Adobe Media Encoder GUI and Zambelli's WMV 9 PowerToy GUI to set the registry keys.

The only thing you can't do is use more than a 10 Mbps peak bitrate; that's a Premiere Pro GUI limitation. But that shouldn't hurt your quality that much.

These are my settings:
Frame Width: 1.440
Frame Height: 1.080
Frame Rate: 25
Codec: Windows Media Advanced Profile
Encoding Passes: 1
Bitrate Mode: constant
Max Bitrate: 5.000
Image Quality: 50
Keyframe Interval: 5
Buffer Size: default

You can download the file here:
http://download.yousendit.com/EFC8880A12F24AF9

Strangely, I always tested these files in MWP 11 where they deinterlace badly. I tested them in MCE today and this file plays perfect there. When I push bitrate and image quality to the maximum (10.000) however, I have this terrible judder again. Is there a flag somewhere that has to be set via Powertoys to tell the decoder that this is an interlaced file?

[benwaggoner]

These are my settings:
Strangely, I always tested these files in MWP 11 where they deinterlace badly. I tested them in MCE today and this file plays perfect there. When I push bitrate and image quality to the maximum (10.000) however, I have this terrible judder again. Is there a flag somewhere that has to be set via Powertoys to tell the decoder that this is an interlaced file?
Yeah, in Zambelli's WMV9 PowerToy, set the "Internal Deinterlacer" to "True". It's in the "Decode" tab.

Also, don't push Image Quality to maximum - that'll cause dropped frames (telling the codec to prioritize image quality over frame rate). That may be what you're seeing. An image quality of 90 should be okay.

In Premiere, because you have the 10 Mbps max, you probably want to use 2-pass VBR unconstrained, so you can let the peaks go higher.

[Ro-Co]

We're coming closer and closer to near perfect results. Ben Wagonner, you were right about the image quality, reducing this makes the file perfectly smooth. The files I code myself now in VC1 nearly match the best X264 codings Nm made.
One issue is now unresolved, if I play the files in WMP11, they look terrible and cannot be played anymore in MCE (judder). When I reboot and play them in MCE first, everything works fine. Playing them again in WMP11 ruins everything.
I am now coding a complete 1h project and evaluate the result tonight.
As soon as my son has the opportunity to visit me we will evaluate the results and decide which way to go, VC1 or H264. Or maybe Silverlight???

[benwaggoner]

We're coming closer and closer to near perfect results. Ben Wagonner, you were right about the image quality, reducing this makes the file perfectly smooth. The files I code myself now in VC1 nearly match the best X264 codings Nm made.
One issue is now unresolved, if I play the files in WMP11, they look terrible and cannot be played anymore in MCE (judder). When I reboot and play them in MCE first, everything works fine. Playing them again in WMP11 ruins everything.
Did you try the registry key setting via WMV9 PowerToy?

I am now coding a complete 1h project and evaluate the result tonight.
As soon as my son has the opportunity to visit me we will evaluate the results and decide which way to go, VC1 or H264. Or maybe Silverlight???
Silverlight 2 doesn't have native support for interlaced VC-1 streams; it's progressive only.

[fico99]

Did you try the registry key setting via WMV9 PowerToy?


I did, but I discovered that WMP11 only de-interlaces the WMV9 Advanced Profile video on my laptop when I set the "Display Size" in WMP11 to 50%. When I play the video @ 100% or in Full Screen, the picture is distorted by interlacing artefacts.

I am using an HP/Compaq 6510b notebook with an Intel(R) Core(TM)2 Duo CPU T7100 @ 1.80GHz, 2GB of RAM and a Mobile Intel(R) 965 Express Chipset @ 1280 x 800 True Color (60Hz).

None of the MPEG-4 or WMV-HD samples posted in this thread play with the same quality/smoothness as my original HDV sample except the "Deshaker"-samples. However, they are soo smooth that it doesn't look the same as the original anymore. It's hard to say, but they look too smooth for me.

Maybe I just have to stick with the HDV format for PC use untill I have more powerfull hardware to play MPEG-4/AVC or AVCHD. My PS3 plays MPEG-4/AVC or AVCHD without any problem. Playing H264 is the PS3's second nature.

Now I will try all the MP4 and WMV samples on the PS3 and let you know the results...

[pandy]

I did, but I discovered that WMP11 only de-interlaces the WMV9 Advanced Profile video on my laptop when I set the "Display Size" in WMP11 to 50%. When I play the video @ 100% or in Full Screen, the picture is distorted by interlacing artefacts.

This probably side effect of downscaling by 2 - WMP instead deinterlacing simply add two lines (result is the same as deinterlacing by blending two fields)

[scharfis_brain]

if WMP (or any other player) shows no interlacing at 50% size, it is the graphics adapter's scaling that either throws away every 2nd line or blends every two lines.

[fico99]

Hey guys, thank you so much for sharing all this information. I have to admit that you guys are maybe right and that encoding 1080i50 HDV (50 fields/second) to 720p50 H.264 (50 frames/second) with good software pre-deinterlacing is the way to go.

Winnydows, the writer of my favorite video conversion tool Xvid4PSP, has implemented some new software deinterlacers that I am testing with my HDV sample. These deinterlacers (MCBob, NNEDI and SmoothDeinterlace) also double the framerate to 50fps. I will upload the best result as soon as I am done with encoding and testing.

The only thing that bothers me is that I am downscaling from 1440x1080@50i to (only) 1280x1080@50p. I would prefer to encode 1920x1080@50p or 1440x1080@50p instead, but I am afraid that a 50fps framerate at such high resulotions will be too difficult to play on most PC's :rolleyes:

Would it be possible to pre-deinterlace and encode to 1080p @ 25fps with relatively smooth motion? Or should I stick with encoding @ 50fps? What do you think?

[pandy]

Would it be possible to pre-deinterlace and encode to 1080p @ 25fps with relatively smooth motion? Or should I stick with encoding @ 50fps? What do you think?

This is highly subjective matter - at this moment i prefer 720p over 1080i - but this is highly dependent from personal taste, type of source (eg film source are 24p) also display technology is very important (ie CRT vs LCD etc).

[Gannjunior]

@scharfis_brain

how would manage (to hit the best quality) an AVCHD video (1440x1080i PAL 25fps) by avisynth script ?
I'm particular interested about deinterlacing by yadifmod and nneni.
(i believe no need to deshacker it)

here a little sample:
http://www.savefile.com/files/1426101

TIA,
ciao!

[scharfis_brain]

which format do you want your video be?
1) anamorphic PAL 704x576
1a) progressive (25fps)
1b) interlaced (50fps)
2) smallHD 1280x720
2a) 25fps
2b) 50fps

leaving it 1080 is useless, because the camera has not near this image resolution.

[Gannjunior]

i need generally progressive video.
but sometimes i could need interlaced 50.
best resolution as possible, so, at least, smallHD.

(you don't suggest to keep 1080, but if i would like to keep it, what would you suggest?).

thanks!

[SeeMoreDigital]

i need generally progressive video.
but sometimes i could need interlaced 50.
best resolution as possible, so, at least, smallHD.

(you don't suggest to keep 1080, but if i would like to keep it, what would you suggest?).What type of player do you ultimately intend to view your encodes on?

Support for decoding "interlaced" sources using software players is still very poor. Whereas with hardware players it's very good.


Cheers

[fico99]

Hi everyone,

As promised, the download link of my H.264 encode using Winnydows Xvid4PSP with the "MP4 PS3" preset with the settings x264 HQ Turbo / SmoothDeinterlace (50 fps) / lanczos4Resize to 1280x720:
http://download.yousendit.com/091510713D10B635

Yesterday, I tested my encodes on my fathers hardware setup (Toshiba 52Z3030D 52" Full HD LCD TV connected with HDMI to an HTPC using an AMD Athlon Dual Core with a GeForce 7600GT running @ 1920x1080p50). We discovered that with 720p50 (50fps) the picture quality is not as sharp as the original HDV file anymore and that the system cannot handle 1080p50 (50fps) because it cannot keep up with a framrate of 50fps @ 1920x1080.

Then we tried to play some AVCHD encodes (PAL 50Hz interlaced 1440x1080 @ 12Mbps and PAL 50Hz interlaced 1920x1080 @ 13 Mbps) using the PowerDVD 7 Ultra CODEC (with PureVideo HD HW acceleration) in WMP11 and XP MCE and the results were stunning! It looked as good as the original!

So, my conclusion so far:

- If you have a powerfull PC with a video card wich supports H.264 HW acceleration: archive HDV in AVCHD 1920x1080 @ 50 Hz interlaced with >10Mbps
- If you have an ordinary PC (like me) with a standard CPU and video card without H.264 acceleration: keep the HDV format or encode to 720p50 (50fps) using a smooth deinterlacer.

[scharfis_brain]

@Gannjunior:
if you need progressive video, with THIS particular camera, stick with 1280x720.
1440x1080 or 1920x1080 is useless, cause the sensor of the camera barely offers enough pixels to fill 1024x576 with a crisp image.

I used this script:
directshowsource("00013.m2ts")
assumetff()
bicubicresize(1280,height)
yadifmod(mode=1, edeint=nnedi(field=-2))
s=bicubicresize(width,720)
bv3 = s.MVAnalyse(isb = true, truemotion=true, delta = 3, idx = 10,blksize=16)
bv2 = s.MVAnalyse(isb = true, truemotion=true, delta = 2, idx = 10,blksize=16)
bv1 = s.MVAnalyse(isb = true, truemotion=true, delta = 1, idx = 10,blksize=16)
fv1 = s.MVAnalyse(isb = false, truemotion=true, delta = 1, idx = 10,blksize=16)
fv2 = s.MVAnalyse(isb = false, truemotion=true, delta = 2, idx = 10,blksize=16)
fv3 = s.MVAnalyse(isb = false, truemotion=true, delta = 3, idx = 10,blksize=16)

fc3 = s.MVFlow(fv3, idx=10, thSCD1=500)
fc2 = s.MVFlow(fv2, idx=10, thSCD1=500)
fc1 = s.MVFlow(fv1, idx=10, thSCD1=500)
bc1 = s.MVFlow(bv1, idx=10, thSCD1=500)
bc2 = s.MVFlow(bv2, idx=10, thSCD1=500)
bc3 = s.MVFlow(bv3, idx=10, thSCD1=500)

interleave(fc3, fc2, fc1, s, bc1, bc2, bc3)
ttempsmooth(maxr=3, lthresh=20, cthresh=30)

selectevery(7,3)

Although it is slow the mocomped denoising helps compression cause there is no more sharpness flicker between the fields.

http://rapidshare.com/files/104765595/yv12-720p-xvid-q3.avi.html

[fico99]

@Gannjunior:
if you need progressive video, with THIS particular camera, stick with 1280x720.
1440x1080 or 1920x1080 is useless, cause the sensor of the camera barely offers enough pixels to fill 1024x576 with a crisp image.

Although your encode looks really nice, I see a big difference in resolution detail between the encoded Xvid and the original AVCHD file, especially on my 50" Pioneer Plasma TV :rolleyes:

But you are right, higher than 1280x720 using progressive video is overkill. I am testing PAL 1080i AVCHD encodes on my PS3 now and it looks very nice. Now, I hope that Sony will bring out a new PS3 firmware so that PAL AVCHD files can be played from the XMB directly @ 50Hz without the need of an AVCHD DVD file structure. With the current System Software all videos that are launched directly from the XMB interface are played @ 60Hz only and this causes judder with 25fps or 50fps based footage.

[Gannjunior]

SeeMoreDigital,

about the interlaced material i would look it on 42" plasma TV of my friend...I believe too it is better to avoid "on fly" deint by pc player...

-----

scharfis_brain,
first: thank you very much!!
the script works perfectly.

only one thing: if i would 25p i don't think it's a good idea to add in the above script only a 'ConvertFps()'...don't it?

ciao !!!:)

[scharfis_brain]

just use selecteven().
it halves the framerate by only selecting even frames.

[fico99]

just use selecteven().
it halves the framerate by only selecting even frames.

But then you will loose the smoothness in the video and it will look more to film than video, right?

Gannjunior, why would you prefer 25p over 50p? Do you have a playback problem with 720p50?

My experiments and findings about the conversion of HDV PAL 50Hz interlaced 1440x1080 to H.264/AVC are the following:
- AVCHD 1440x1080i 25fps encodes are sharper than 720p 50fps encodes on my father's Toshiba 52Z3030D 52" Full HD LCD TV (playback with XP MCE).
- I see no difference in picture quality between AVCHD 1440x1080i 25fps or 720p 50fps encodes on my Pioneer PDP-507XD 'HD Ready' Plasma TV (playback with PS3).
- 720p 25fps cannot be used because of judder.
- A good AVCHD 1440x1080i 25fps encode can look as good as the original HDV source!
- WMV-HD encodes cannot compete with AVCHD encodes.
- You need powerfull hardware to play AVCHD encodes.

My conclusion:

AVCHD 1440x1080i or 1920x1080i 25fps, using a bitrate of >10Mbps, has past ALL our tests and looks as good as our original HDV footage. Based on all the testing we have done in the last 2 weeks, we have decided to archive our HDV footage in the AVCHD format. Even better, if the prices of the BD-R(W) will ever come down, you can easily burn your encodes to Blu-ray disc. In the mean time, we will archive everything on dual (manually mirrored) external USB/FireWire/eSATA 1TB hard disk drives (one in my father's house and one in my house for site redundancy).

[SeeMoreDigital]

fico99,

Should you ever decide to buy yourself a dedicated hardware device that supports playback of high-def MPEG-2, MPEG-4 (Part-2 and Part-10) and VC-1 sources, your finding might change again!

[nm]

Should you ever decide to buy yourself a dedicated hardware device that supports playback of high-def MPEG-2, MPEG-4 (Part-2 and Part-10) and VC-1 sources, your finding might change again!
In what way? I think the points presented by fico are quite reasonable in that case too.

[fico99]

fico99,

Should you ever decide to buy yourself a dedicated hardware device that supports playback of high-def MPEG-2, MPEG-4 (Part-2 and Part-10) and VC-1 sources, your finding might change again!

I am convinced that the AVCHD 1440x1080i 25fps is one of the most futureproof formats and I cannot imagine that newer dedicated hardware would not support this format natively.

Also new hardware and software for the PC now starts to fully support this AVCHD format too.

[SeeMoreDigital]

As I've mentioned before in this thread....

"Support for decoding "interlaced" sources using software players is still very poor. Whereas with hardware players it's very good"

I am convinced that the AVCHD 1440x1080i 25fps is one of the most futureproof formats... Hmmm...

Personally, I don't understand why any "high-def" camcorder manufacturer is bothering to support "interlaced" capturing... Are we not supposed to be living in a progressive world ;)

[fico99]

@Gannjunior:
if you need progressive video, with THIS particular camera, stick with 1280x720.
1440x1080 or 1920x1080 is useless, cause the sensor of the camera barely offers enough pixels to fill 1024x576 with a crisp image.

scharfis_brain,

Today, I was on visit with my father and compared the original AVCHD with your encoded XviD again on his 52" Full LCD TV. I have to apologize what I said before and have to admit that the difference in sharpness is not that big. When we paused the video in the middle when the bike is passing at high speed, we see even less artefacts in the encoded XviD video.
Only the detail of static objects, like the sign of the tankstation, the license plates, the trees or the structure of the road are sharper on the original AVCD footage.

Thanks for sharing all this interesting information with us :)

[Gannjunior]

hi scharfis_brain,

this is result using your script, seesaw and time remapping in premire:
test timeremapping (http://www.savefiles.net/d/tdr4zx83ekyd3.html)

i'me very happy!!!

----
another thing that i've notice about a sony SR11: it has the field inverted compared to SD1, since i need to assumeBFF. here it is the "raw":

00073 MTS (http://www.magazine-video.com/tests/00073.MTS)

getting 25p (by adding selecteven as you suggest) there is no problem.
But getting 50p, something goes wrong...the video is not perfect, not perfect fluid (as i obtain in the other SD1 sample).

Do I forget anything?

ciao!

[Gannjunior]

maybe the problem in here:

assuming TFF this is ok:
yadifmod(mode=1, edeint=nnedi(field=-2))

but assuming BFF i would have to get:
yadifmod(mode=1, order=0, edeint=nnedi(field=+2))

i tried. i works fine for 25p (by adding selecteven() at the end ) but in the 50p i obtain something strange, there is something wrong in the video...

[scharfis_brain]

I don't see you problem here!

AVCSource("00073.dga")
AssumeTFF()
yadifmod(mode=1,edeint=nnedi(field=-2))

produces a smooth full framerate output.

[Gannjunior]

hi scharfis,
sorry if i answer now but i was in country during previous days...

if i use TFF i obtain a smooth full framerate output but with the typical horizontal lines of deinterlacing...
using BFF, no horizontal lines of deint but no full framerat output...

[scharfis_brain]

1) post your full script!

2) provide screenshots for your tff / bff cases!

but with the typical horizontal lines of deinterlacing...
what do you mean?
assumetff() / assumebff() in my posted script above only changes the order of the frames being put out.
It doesn't lead to a change of the look of the individual frames.

[Gannjunior]

this is the script i used.


AVCSource("m:\SR11e_00073.dga")

s=assumetff().yadifmod(mode=1, edeint=nnedi(field=-2))
#s=assumebff().yadifmod(mode=1, order=0, edeint=nnedi(field=-2))

#blackmanresize(1280,height)

#s=blackmanresize(width,720)
bv3 = s.MVAnalyse(isb = true, truemotion=true, delta = 3, idx = 10,blksize=16)
bv2 = s.MVAnalyse(isb = true, truemotion=true, delta = 2, idx = 10,blksize=16)
bv1 = s.MVAnalyse(isb = true, truemotion=true, delta = 1, idx = 10,blksize=16)
fv1 = s.MVAnalyse(isb = false, truemotion=true, delta = 1, idx = 10,blksize=16)
fv2 = s.MVAnalyse(isb = false, truemotion=true, delta = 2, idx = 10,blksize=16)
fv3 = s.MVAnalyse(isb = false, truemotion=true, delta = 3, idx = 10,blksize=16)

fc3 = s.MVFlow(fv3, idx=10, thSCD1=500)
fc2 = s.MVFlow(fv2, idx=10, thSCD1=500)
fc1 = s.MVFlow(fv1, idx=10, thSCD1=500)
bc1 = s.MVFlow(bv1, idx=10, thSCD1=500)
bc2 = s.MVFlow(bv2, idx=10, thSCD1=500)
bc3 = s.MVFlow(bv3, idx=10, thSCD1=500)

interleave(fc3, fc2, fc1, s, bc1, bc2, bc3)
ttempsmooth(maxr=3, lthresh=20, cthresh=30)

selectevery(7,3) #### 50p -> solo selectevery


#selectEven() ### 25p -> selectevery + selecteven





Compared to your script it is identical except i removed resize. Using this script the ouput is fluid but with classical interlacing lines:

http://img176.imageshack.us/img176/7448/50p000140100658cv5.th.jpg (http://img176.imageshack.us/my.php?image=50p000140100658cv5.jpg)

Instead using exactly your script (so with 720p resize):

AVCSource("m:\SR11e_00073.dga")

#assumebff()
assumeTff
blackmanresize(1280,height)
yadifmod(mode=1, edeint=nnedi(field=-2))
s=blackmanresize(width,720)
bv3 = s.MVAnalyse(isb = true, truemotion=true, delta = 3, idx = 10,blksize=16)
bv2 = s.MVAnalyse(isb = true, truemotion=true, delta = 2, idx = 10,blksize=16)
bv1 = s.MVAnalyse(isb = true, truemotion=true, delta = 1, idx = 10,blksize=16)
fv1 = s.MVAnalyse(isb = false, truemotion=true, delta = 1, idx = 10,blksize=16)
fv2 = s.MVAnalyse(isb = false, truemotion=true, delta = 2, idx = 10,blksize=16)
fv3 = s.MVAnalyse(isb = false, truemotion=true, delta = 3, idx = 10,blksize=16)

fc3 = s.MVFlow(fv3, idx=10, thSCD1=500)
fc2 = s.MVFlow(fv2, idx=10, thSCD1=500)
fc1 = s.MVFlow(fv1, idx=10, thSCD1=500)
bc1 = s.MVFlow(bv1, idx=10, thSCD1=500)
bc2 = s.MVFlow(bv2, idx=10, thSCD1=500)
bc3 = s.MVFlow(bv3, idx=10, thSCD1=500)

interleave(fc3, fc2, fc1, s, bc1, bc2, bc3)
ttempsmooth(maxr=3, lthresh=20, cthresh=30)

selectevery(7,3) #### 50p -> solo selectevery

#selectEven() ### 25p -> selectevery + selecteven

the result are perfect, fluid and with any interlacing problem.

At last, about the bff case:


#AVCSource("E:\A PARTE\Video_Cervarezza_2007\00013.dga")

AVCSource("m:\SR11e_00073.dga")

s=assumebff().yadifmod(mode=1, order=0, edeint=nnedi(field=2))

#blackmanresize(1280,height)

#s=blackmanresize(width,720)
bv3 = s.MVAnalyse(isb = true, truemotion=true, delta = 3, idx = 10,blksize=16)
bv2 = s.MVAnalyse(isb = true, truemotion=true, delta = 2, idx = 10,blksize=16)
bv1 = s.MVAnalyse(isb = true, truemotion=true, delta = 1, idx = 10,blksize=16)
fv1 = s.MVAnalyse(isb = false, truemotion=true, delta = 1, idx = 10,blksize=16)
fv2 = s.MVAnalyse(isb = false, truemotion=true, delta = 2, idx = 10,blksize=16)
fv3 = s.MVAnalyse(isb = false, truemotion=true, delta = 3, idx = 10,blksize=16)

fc3 = s.MVFlow(fv3, idx=10, thSCD1=500)
fc2 = s.MVFlow(fv2, idx=10, thSCD1=500)
fc1 = s.MVFlow(fv1, idx=10, thSCD1=500)
bc1 = s.MVFlow(bv1, idx=10, thSCD1=500)
bc2 = s.MVFlow(bv2, idx=10, thSCD1=500)
bc3 = s.MVFlow(bv3, idx=10, thSCD1=500)

interleave(fc3, fc2, fc1, s, bc1, bc2, bc3)
ttempsmooth(maxr=3, lthresh=20, cthresh=30)

selectevery(7,3) #### 50p -> solo selectevery


#selectEven() ### 25p -> selectevery + selecteven



I obtain no interlacing lines in the frames, but the playback results "slow", as a rallenty.


ciao!

[scharfis_brain]

AVCSource("m:\SR11e_00073.dga")
s=assumetff().yadifmod(mode=1, edeint=nnedi(field=-2))

of course, this cannot work!
yadifmod gets signalled tff, while nnedi gets signalled bff.
This causes yadifmod and nnedi being out of phase.

let me rewrite it without the variable 'last' being hidden:

last=AVCSource("m:\SR11e_00073.dga")
s=last.assumetff().yadifmod(mode=1, edeint=last.nnedi(field=-2))

do you see now, what was going on with your script?
you have to be very careful inheriting clip using the invisibla 'last'.

[Gannjunior]

oh...i'm a silly man!
i've just corrected my script, thanks !

however i tried it, but the problem of interlaced lines remains..

[scharfis_brain]

post your script again!

[Gannjunior]

here it is:

source=AVCSource("m:\SR11e_00073.dga")

s=source.assumetff().yadifmod(mode=1, edeint=source.nnedi(field=-2))
#s=source.assumebff().yadifmod(mode=1, order=0, edeint=source.nnedi(field=2))
#blackmanresize(1280,height)

#s=blackmanresize(width,720)
bv3 = s.MVAnalyse(isb = true, truemotion=true, delta = 3, idx = 10,blksize=16)
bv2 = s.MVAnalyse(isb = true, truemotion=true, delta = 2, idx = 10,blksize=16)
bv1 = s.MVAnalyse(isb = true, truemotion=true, delta = 1, idx = 10,blksize=16)
fv1 = s.MVAnalyse(isb = false, truemotion=true, delta = 1, idx = 10,blksize=16)
fv2 = s.MVAnalyse(isb = false, truemotion=true, delta = 2, idx = 10,blksize=16)
fv3 = s.MVAnalyse(isb = false, truemotion=true, delta = 3, idx = 10,blksize=16)

fc3 = s.MVFlow(fv3, idx=10, thSCD1=500)
fc2 = s.MVFlow(fv2, idx=10, thSCD1=500)
fc1 = s.MVFlow(fv1, idx=10, thSCD1=500)
bc1 = s.MVFlow(bv1, idx=10, thSCD1=500)
bc2 = s.MVFlow(bv2, idx=10, thSCD1=500)
bc3 = s.MVFlow(bv3, idx=10, thSCD1=500)

interleave(fc3, fc2, fc1, s, bc1, bc2, bc3)
ttempsmooth(maxr=3, lthresh=20, cthresh=30)

selectevery(7,3) #### 50p -> solo selectevery


#selectEven() ### 25p -> selectevery + selecteven

[scharfis_brain]

you didn't change anything!
you just exchanged the invisible variable 'last' by 'source'.
So the error remains the same:

source.assumetff().yadifmod(...) #signalling TFF to yadifmod
edeint=source.nnedi(...) #signalling BFF to nnedi

see? this cannot work!

you have to signal BOTH yadifmod AND nnedi the same fieldorder!

only this will work properly:

source=AVCSource("m:\SR11e_00073.dga").assumetff()
s=source.yadifmod(mode=1, edeint=source.nnedi(field=-2))

now both will be signalled tff.

[Gannjunior]

everything is perfect now.

You've been really really patient and kind. ;-)

THANK YOU.