Deinterlace Picture Quality is not good

[vdrfan]

hi, how can we see which Deinterlace Method is used. Properties Screen show only all available: DeinterlaceTech.: Bob, Adaptive, Motion Compensation, Inverse Telecine is it posible to force a method? thanks, onur

[v0lt]

how can we see which Deinterlace Method is used.

There is no such possibility.

Properties Screen show only all available: DeinterlaceTech.: Bob, Adaptive, Motion Compensation, Inverse Telecine is it posible to force a method?

There is no such possibility. This is just a list of what the selected video processor can do.

[vdrfan]

i get much better result with software deint. (LAV + Yadif) - but i would prefer MPC Decoder + MPCVR + HW Deinterl. is there a option in AMD Driver Settings to Force a better Deinterlacing Method.

[Klaus1189]

What GPU do you use?

[v0lt]

https://github.com/Aleksoid1978/VideoRenderer/wiki/Known-Issues

Deinterlacing does not work correctly on some AMD graphics adapters.

Reason: Some AMD graphics adapters require Future Frames for the D3D11 video processor. Currently MPC Video Renderer does not support Future Frames.

[vdrfan]

i use IGPU AMD Ryzen 5 8600G - Radeon 760M, old Vega IGPU(3400G) is working better. is there a plan to supprt Future Frames.

[Aleksoid1978]

Can you check in other D3D11 video renderers, like in system media player Movies & TV or Kodi, mpv ?

[vdrfan]

@Aleksoid1978 - mvp is terrible even with intel. vlc is ok, MPC-BR with MAD is ok.

i have found following problem: Video Decoder(LAV and MPC tested) cannot use Hardware Accelaration(DXVA2 or D3D11) on AMD(Radeon 760M tested) with certain SD MPEG2 Videos. MPC Video Decoder shows: output format: NV12 - should show D3D11 on Intel System everything is working: output format: D3D11 whit MADVR Picture Quality is OK(AMD and Intel) (?? maybe MAD uses Internal SW Deint. when DXVA2 is not active ??)

testfile: .ts with 50MB https://drive.google.com/file/d/1F_16sZUQodwP8IZAY_nPonYt37J0TJt9/view?usp=drive_link

[vdrfan]

comparison : intel - AMD INTEL:

AMD:

[clsid2]

Madvr does not use D3D11 video processor.

[vdrfan]

@clsid2, yes i know - that was a reason for me to switch to MPC-VR, but Madvr and LAV can play the video optical perfect.

[clsid2]

Does MPCVR deinterlace work if you use D3D9?

[vdrfan]

@clsid2 i think deinterlace is working but it looks bad (regardless d3d9 or d3d11). i think - main problem is that video decoder is not useing hwaccel on some SD mpeg2 videos. can somebody can check testvideo with AMD rdna graphics. i use Driver Adrenalin 24.4.1

[v0lt]

https://disk.yandex.ru/d/LKPmtavOINaE6w These are test build for testing. In this form, changes are not included in official releases, because video frames in these build are output 1-2 frames late.

No one has been able to implement a correct solution for MPC VR. Therefore, there are no plans for any specific time.

[vdrfan]

thanks V0lt, this build is working. Picture Quaity is much better. is there source code available of this build. master branch from Aleksoid1978 does not have that changes - for Future Frames. 1-2 frames lateness is no problem for me.

[v0lt]

master branch from Aleksoid1978 does not have that changes - for Future Frames.

4.5 years it's been there.

[clsid2]

Since it is currently disabled, is there a known problem with it? Or was it just considered unneeded? A possible optimization could be to use past/future frames only for interlaced content. Or are past frames needed for normal progressive content?

[v0lt]

Since it is currently disabled, is there a known problem with it?

It was always disabled. And in this state there were no problems. The problem was recently discovered on some modern AMD graphics adapters. If we enable future frame support incorrectly, we will end up with a 1-2 frame delay (and some problems at the beginning of playback) on all graphics adapters.

A possible optimization could be to use past/future frames only for interlaced content. Or are past frames needed for normal progressive content?

Most videos are marked as "possibly interlaced". Frame type marking for each frame occurs in the video decoder. Therefore, in most cases, we create a video processor that supports deinterlacing technologies, even for progressive video.

[vdrfan]

my suggestion make a interface where the decoder or player can activate Future Frames.

[Aleksoid1978]

No.

[clsid2]

You know when video is interlaced based on mediatype flags. You can only activate it in that case.

[clsid2]

Also, future frames can probably be maximized to 1 frame. That should be enough for deinterlacing. An adviced value that is higher than 1 is likely for inverse telecine purposes.

[v0lt]

You know when video is interlaced based on mediatype flags. You can only activate it in that case.

As I said, MPC Video Decoder and Microsoft DTV-DVD Video Decoder create an interlaced media type for most video (including progressive).

Also, future frames can probably be maximized to 1 frame.

This value varies from 0 to 2 for different graphics adapters. I don't have an AMD graphics adapter with this problem to check the different values. On my Nvidia, AMD and Intel graphics cards, deinterlacing works well without future frames.

[clsid2]

For most video the decoder knows that it is 100% progressive. So using interlaced flags then is just wrong.

The value is to get support for all functionality of RateConvCaps. But in cases where value is >1 it might be enough to use 1 if only deinterlacing capability is used. At least, that is my assumption. That avoids unnecessary long frame delays.

[vdrfan]

whith active Future Frames - do we need to delay Audio to be sync?

[Aleksoid1978]

If you want improvements in this direction, provide the developers with the appropriate equipment. After this there may be some changes. For now, stop writing here, topic is closed.