Bicubicresize() and Bob() should default to Catrom, not Mitchell-Netravali

[Katie Boundary]

Catrom is the One True Cubic, a claim that I base on the following two facts:

1) It's the only cubic that preserves linear gradients (it follows the b+2c=1 rule) and also always passes through the original control points/samples (is a "no-op")

2) Spline16, Lagrange-2, Lanczos-2, etc. are all approximations of Catrom and not any other cubic

EDIT: in PSNR tests, it also kicks Mitchell-Netravali's ass all up and down the sidewalk every day of the week and twice on Sundays.

[hello_hello]

Create a couple of wrapper functions with whatever defaults you prefer and use those.

function KatieBicubic(clip Source, int target_width, int target_height, \
float "src_left", float "src_top", float "src_width", float "src_height", float "B", float "C")  {

B = default(B, 0)
C = default(C, 0.5)

BicubicResize(Source, target_width, target_height, \
src_left=src_left, src_top=src_top, src_width=src_width, src_height=src_height, b=B, c=C)  }

[wonkey_monkey]

Originally Posted by Katie Boundary

Bicubicresize() and Bob() should default to Catrom, not Mitchell-Netravali

It's already been suggested but was rejected for the sake of backwards compatability: http://forum.doom9.net/showthread.php?p=1849921

[StainlessS]

And here Katie CatromResize() as discussed in Katie PM.

Function CatromResize(clip a, int w, int h, float "src_left", float "src_top", float "src_width", float "src_height") {
    src_left   = Default(src_left  ,0.0) # Should provide defaults for optionals, otherwise are UNDEFINED
    src_top    = Default(src_top   ,0.0) #   Ditto : (If client specifies src_top then use it, otherwise default to 0.0)
    src_width  = Default(src_width ,0.0) #   Ditto
    src_height = Default(src_height,0.0) #   Ditto
    BicubicResize(a, w, h, 0, 0.5, src_left, src_top, src_width, src_height)
}

And here slightly more Advanced implementation

Function CatromResize(clip a, int w, int h, float "src_left", float "src_top", float "src_width", float "src_height") {
    BicubicResize(a, w, h, 0, 0.5, src_left, src_top, src_width, src_height)
}

As we know that BicubicResize() uses defaults of 0.0 for its undefined optional coords (src_left/top/width/height), we can just pass on args (client supplied or Undefined) to
BicubicResize and have it supply its defaults instead [EDIT: when are undefined].
Either we supply 0.0 default, or BicubicResize does, dont really matter which does it,
only need to supply our own defaults if we require different defaults than used by BicubicResize.

Note, HH moved args b,c, to end of arg KatieBicubic() list.

EDIT: Removed "c=" and "b=" from "c=0" and "b=0.5" [leaving the above in blue]
Otherwise need supply all of the src left/src_top etc together with names.
Once you supply a named argument in a function call, all following supplied args (if any) MUST also be supplied named.

UNTESTED,

EDIT: Katie, use one of above. PM script functions bugged due to red line above.

[hello_hello]

Resize8 has a Catmull-Rom kernel, converted to BicubicResize(b=0, c=0.5)

Resize8(640,480, kernel="Catmull-Rom")

[DTL]

Lanczos are sinc weighted by sinc and may be very 'long' in taps. It is from DSP-theory of signals processing. Bicubic is fixed number of samples kernel and not based on sinc. Lanczos2 uses same number of taps also may looks close to some bicubic b,c but the curve still not the equal I think because bicubic is based on different basis function (x^n, not sin(x)).

[Katie Boundary]

Originally Posted by StainlessS

And here Katie CatromResize() as discussed in Katie PM.

This is a discussion of what AVIsynth default behavior should be. It has nothing to do with the creation of custom AVSI files. In our PM conversations, I was merely using Catrom as an example to better understand how AVSI scripting works in general.

Also, your inbox is full.

[wonkey_monkey]

Yes StainlessS, how dare you try and be helpful. Shame on you!

[Sunspark]

I really like Mitchell-Netravali with the values b = 1/3, c = 1/3.

[StainlessS]

how dare you

Yeah, you'd think I'd have learnt by now.

[Katie Boundary]

Originally Posted by StainlessS

Yeah, you'd think I'd have learnt by now.

I'm not criticizing or complaining. I'm clarifying the purpose of the thread so that discussions can be properly contained.