MVTools without idx

[Fizick]

Released MVTools v2.5.11.20.
•Fixed Mflow broken in 2.5.11.2, fixed several other small bugs.
•MBlockFps: added support of overlap, renamed parameter thres to ml.
•MCompensate, MMask: added partial time compensation.
•Masked functions: improved occlusion mask.
•MBlockFPS, MMask - added option to use SAD as occlusion mask.
•Documented (after Firesledge) negative delta values (v.2.2.1 quick hack for static scene compensation and motion tracking) and fixed some functions accordingly.

Avisynth 2.6 header, but no new color formats.

[stax76]

I know it's very difficult but a merge of all flavors and builds would be awesome. Out of the 34 x64 plugins included with StaxRip x64 only ffms2 and KNLMeansCL support both AviSynth+ x64 and VapourSynth x64 as single DLL.

[chainik_svp]

BTW svpflow plugin code contains an intermediate abstraction layer to separate the core logic from *synth interface, so the same sources're used to build both Avisynth and Vapoursynth plugins.
It's more like a demo version now and may not be suitable for a complicated plugins but it's totally enough for MSuper/MAnalayse/M***Fps.

[pinterf]

Fizick, download link to the new version is available only from the Russian page http://avisynth.org.ru/mvtools/mvtools2-rus.html

[Fizick]

Anyboby else have a problem with english page ?

[Boulder]

Nope, seems to work fine for me.

[pinterf]

Must be the company proxy. It works from home.

[Fizick]

v2.5.11.21 (22.04.2016 by Fizick)
•MflowXXX: remove limit of motion vectors length (was 127/pel).
•MMask: possible vector scaling for kind=3,4,5.

The vector length limit problem was discussed (without me ) here:
http://forum.doom9.org/showthread.php?t=165075

It was not bug but feature (documented).
Vectors always was stored as short , but previous versions used interpolation (resize) code with BYTE (char) as unity. Now it is short int.

[Groucho2004]

Originally Posted by Fizick

v2.5.11.21 (22.04.2016 by Fizick)
•MflowXXX: remove limit of motion vectors length (was 127/pel).
•MMask: possible vector scaling for kind=3,4,5.

Thanks!

[jackoneill]

1) avisynth.org.ru appears to be dead? It loads "Default Web Site Page".

2) In v2.5.11.3, in MVDepan.cpp, line 711:

	int nf = (backward) ? ndest: ndest; // set next frame number as data frame if backward

One of those should be either ndest+1 or ndest-1, right? Also, the comment contradicts the comment at line 549. Which one is correct? Due to 1) I was unable to check if this was fixed in the latest version.

[Reel.Deel]

Originally Posted by jackoneill

1) avisynth.org.ru appears to be dead? It loads "Default Web Site Page".

Fizick's plugins are also hosted here: http://www.avisynth.nl/users/fizick/

Edit: the DL link still points to avisynth.org.ru, here's the correct DL link for MVTools: http://www.avisynth.nl/users/fizick/...v2.5.11.22.zip

[jackoneill]

Originally Posted by Reel.Deel

Fizick's plugins are also hosted here: http://www.avisynth.nl/users/fizick/

Edit: the DL link still points to avisynth.org.ru, here's the correct DL link for MVTools: http://www.avisynth.nl/users/fizick/...v2.5.11.22.zip

Ah, thanks.

[Fizick]

jackoneill, sorry I can not give you answer about MVDepan right now. It is old code. And it is quite possible that the code is correct (by result)


And yes, I now see some problem with my main site. May be cheap hosting limit is exceeded.

[jackoneill]

Originally Posted by Fizick

jackoneill, sorry I can not give you answer about MVDepan right now. It is old code. And it is quite possible that the code is correct (by result)

Oh well. I noticed another thing though: the "zerow" parameter appears to be ignored. The variable "zeroWeight" at line 531 shadows the class member.

[jackoneill]

Another question about MVDepan: if the "range" parameter is greater than zero, it will calculate data (motionx, motiony, motionrot, motionzoom) for several frames in a single GetFrame call. However, it uses just one frame from the mask clip in the calculations for all those frames. Is this a bug?

[jackoneill]

While integrating the recent changes into the VapourSynth plugin, I found several bugs.

MVBlockFPS.cpp
--------------

- At lines 861 and 862 the pointers pDst[1], pDst[2], pSrc[1], pSrc[2], pRef[1], and pRef[2] should each have nWidth_B/2 added.

- At lines 865, 866, and 867 the pitches need to be multiplied by nHeight_B, not nHeight-nHeight_B.

- Short2Bytes is not called before returning from MVBlockFps::GetFrame. Frames are returned uninitialised when overlap is used.

- There is a line missing after line 730.

MaskFun.cpp
-----------

- In function MakeSADMaskTime MaskPitch should be used, not nBlkX, when writing to Mask.



Older things I hadn't noticed or forgot about:

MVFlowBlur.cpp
--------------

At line 74 it should be nVPadding instead of nHPadding.

The arrays MaskSmallB, MaskSmallF, MaskFullYB, MaskFullYF, MaskFullUVB, and MaskFullUVF are never used.



And also: shouldn't FlowFps reject vector clips with negative delta? And in Degrain2 and Degrain3 only the first two vector clips are checked for negative delta.