From Avisynth wiki
Jump to: navigation, search

Back to MVTools2


Gets a source clip and prepares a special Super clip with multilevel (hierarchical scaled) frames data. The Super clip is used by both MAnalyse and motion compensation (client) functions.
The audio properties of the Super clip are used to store and transfer its parameters (specifically num_audio_samples), therefore any existing audio is killed in the Super clip. That is one of reasons why you can additionally use a source clip with client functions. You may have a look to the Super clip yourself (it has a normal format).
Supported formats: Y, YUV and planar RGB (8-32) bits, and YUY2. You can use a different Super clip to generate vectors with MAnalyze and a different Super clip format for the actual action. This enables support for planar RGB with MFlowFPS, and lets MDegrain work with 32 bit float input clips.

Syntax and Parameters

MSuper (clip, int "hpad", int "vpad", int "pel", int "levels", bool "chroma",
      int "sharp", int "rfilter", clip "pelclip", bool "isse", bool "planar", bool "mt")

int  hpad = 8
Horizontal padding added to source frame (both left and right). A small padding is added for more correct motion estimation near frame borders.
int  vpad = 8
Vertical padding added to source frame (both top and bottom), for the same reason as hpad.
int  pel = 2
Motion estimation accuracy. Value can only be 1, 2 or 4.
  • 1 : means precision to the pixel.
  • 2 : means precision to half a pixel.
  • 4 : means precision to quarter of a pixel, produced by spatial interpolation (more accurate but slower and not always better due to big level scale step).
int  levels = 0
Number of hierarchical levels for super clip frames. MAnalyse needs all levels, but for other client functions a single, finest level is enough (coarser levels are not used). 0 is auto (all possible levels are produced).
bool  chroma = true
When true, process chroma planes too. False for luma only.
int  sharp = 2
Sub-pixel interpolation method for when pel == 2 || 4.
  • 0 : soft interpolation (bilinear).
  • 1 : bicubic interpolation (4 tap Catmull-Rom)
  • 2 : sharper Wiener interpolation (6 tap, similar to Lanczos).
int  rfilter = 2
Hierarchical levels smoothing and reducing (halving) filter.
  • 0 : simple 4 pixels averaging like unfiltered SimpleResize (old method)
  • 1 : triangle (shifted) filter like ReduceBy2 for more smoothing (decreased aliasing)
  • 2 : triangle filter like BilinearResize for even more smoothing
  • 3 : quadratic filter for even more smoothing
  • 4 : cubic filter like BicubicResize(b=1,c=0) for even more smoothing
You may also try to apply some external filter to thesuperclip or its coarse bottom part (by appropriate crop and overlay).
clip  pelclip =
Optional upsampled source clip for using instead of internal sub-pixel interpolation (for pel > 1).
Pixels at rows and columns positions multiple to pel (0,2,4,... for pel=2) (without padding) must be original source pixels; other pixels must be interpolated.
Example for pel=2:
MSuper(..., pel=2, pelclip=LanczosResize(width*2,height*2,src_left=0.25, src_top=0.25))
Another useful filter for pelclip is the Eedi3 edge-directed resampler.
Recent note: it is true for luma, but is not exactly corresponded to chroma pixels positions of internal MVTools interpolation. Nevertheless, vectors and motion compensation are quite similar for usual clips, same chroma would be with src_left=0.5 for YUY2 and additionally src_top=0.5 for YV12.[clarification needed]



Back to MVTools2

Personal tools