TV de-interlacing option (ATV4 4K)?

[satfeed1]

Dear Amet!

How do I report this issue or where?

[satfeed1]

3.3

https://github.com/MrMC/mrmc/blob/maste ... EG-VERSION

Do you plan to upgrade FFmpeg version in the player?

[amet]

There are no plans to upgrade ffmpeg for MrMC 3.* as it’s a big core change and will introduce stability issues

[satfeed1]

Many thanks for the latest update: 3.9-4.

New deinterlace methods(w3fdif-simple and complex) is working perfectly with MPEG-2 4:2:2 and H.264 4:2:0 8 bit and H.264 4:2:0 10bit furthermore H.264 4:2:2 8bit.

It seems to work perfectly with H.264 4:2:2 10bit, but I need more sample...

Thank you once again for this important development!!!

[noggin]

Many thanks for the latest update: 3.9-4.

New deinterlace methods(w3fdif-simple and complex) is working perfectly with MPEG-2 4:2:2 and H.264 4:2:0 8 bit and H.264 4:2:0 10bit furthermore H.264 4:2:2 8bit.

This is great - 4:2:2 8-bit h.264 1080i25 stuff with a W3FDIF on a consumer device! Amazing! I'm watching 38Mbs 4:2:2 1080i25 stuff and it's looking almost great!

One thing I have noticed is that it looks as if there may be the annoying ffmpeg 4:2:2 chroma issue that you get if you let ffmpeg do its default 4:2:2 interlace->4:2:0 interlace conversion (it treats the chroma as p25 not i25) and thus saturated chroma elements look juddery.

You see this particularly on entertainment shows. Performers and lighting effects in desaturated costumes (black, grey, white or pale colours) have no judder, but those very saturated reds and blues have the tell tale 'frame rate judder' of frame-based chroma.

This happened to me for a long time when using ffmpeg to convert 1080i25 4:2:2 stuff to 4:2:0 1080i25 - until someone explained the issue.

When you do a command line 4:2:2->4:2:0 ffmpeg conversion in the interlaced domain you need to add an interlace aware scale filter (even though it feels like there isn't a scale taking place - there is in chroma terms I guess - from 1080 to 540 vertically?)

I use something like -filter:v scale=interl=1 to force an interlace friendly chroma resample.

Is the deinterlace for 4:2:2 content happening after an ffmpeg 4:2:2->4:2:0 conversion ?

[satfeed1]

I have already seen this chroma issue, but didn't know what's causing it. Thank you for your information. I hope guys will solve this problem as soon as possible.

[noggin]

Yep - it may be something different - but having seen the same artefacts in the past I thought I'd mention it.

[satfeed1]

I have found a little issue with W3FDIF Complex deinterlacing method. This method can't handle properly 4:2:2 10bit interlaced videos. The W3FDIF simple mode seems to work good.
I have tested 4:2:0 10bit and 4:2:2 8bit videos and those videos perfect with both(simple and complex) deinterlacing method.

Here are two test video files(4:2:2 10bit): https://we.tl/t-Uc6hg3syid

[noggin]

I have found a little issue with W3FDIF Complex deinterlacing method. This method can't handle properly 4:2:2 10bit interlaced videos. The W3FDIF simple mode seems to work good.
I have tested 4:2:0 10bit and 4:2:2 8bit videos and those videos perfect with both(simple and complex) deinterlacing method.

Here are two test video files(4:2:2 10bit): https://we.tl/t-Uc6hg3syid

What artefacts are you seeing with w3fdif=complex:all on the 10-bit stuff? I just converted the TEST_2 .ts clip from h264 10-bit 4:2:2 i25 stuff to h.265 10-bit 4:2:0 p50 (so I retain hardware acceleration on my players) using complex and simple and didn't get major problems with either on a quick inspection.

There is some judder on both though (particularly at the beginning), presumably because there are lots of corrupt decoded frame errors in the video stream reported by ffmpeg?

I used

Code: Select all

-acodec copy -vcodec libx265 -crf 14 -pix_fmt yuv420p10le -vf "w3fdif=simple:all" 

and

Code: Select all

-acodec copy -vcodec libx265 -crf 14 -pix_fmt yuv420p10le -vf "w3fdif=complex:all"

(I know I should be using c:a, c:v and filter:v...)

Am I missing an artefact in the complex ?


*** EDIT - or am I being stupid and are you just talking about simple vs complex Weston 3-field in MrMC? ***

[satfeed1]

When I play 4:2:2 10bit inerlaced videos with MrMC I get many skipped frames with W3FDIF Complex deinterlacing method.
I have tested this complex method with many 4:2:2 10bit interlaced videos and always got errors(but playing 4:2:2 8bit always perfect)