AWarpsharp2/aWarp4
From Avisynth wiki
(Difference between revisions)
m (minor formatting) |
(small documentation update) |
||
| (One intermediate revision by one user not shown) | |||
| Line 1: | Line 1: | ||
{{FilterCat4|External_filters|Plugin_functions|Adjustment_filters|Sharpeners}} | {{FilterCat4|External_filters|Plugin_functions|Adjustment_filters|Sharpeners}} | ||
| − | + | <tt>aWarp4</tt> is designed for an even more custom warp-sharpening than <tt>[[AWarpsharp2/aWarp|aWarp]]</tt> allows. Unlike aWarp, this uses a 4x larger source clip than the edge mask. This is useful for improving the quality of the subpixel interpolation quality in the final step. [[Supersampling]] allows to produce a sharper result with lower {{Template:FuncDef3|depth}}, which equals to less deformations. | |
| − | + | ||
| − | + | ||
| − | < | + | |
| − | + | ||
| − | aWarp4 is designed for an even more custom warp-sharpening than [[AWarpsharp2/aWarp|aWarp]] allows. Unlike aWarp, this uses a 4x larger source clip than the edge mask. This is useful for improving the quality of the subpixel interpolation quality in the final step. Supersampling allows to produce a sharper result with lower {{Template: | + | |
<br> | <br> | ||
<br> | <br> | ||
| Line 12: | Line 7: | ||
<br> | <br> | ||
::{{Par2| |clip| |}} | ::{{Par2| |clip| |}} | ||
| − | :::Input clip; must be exactly 4 times the width and height of the edge mask clip. | + | :::Input clip; must be exactly 4 times the width and height of the edge mask clip. The 4x upsampled clip should always be top-left aligned, take a look at the examples below. |
<br> | <br> | ||
::{{Par2| |clip| |}} | ::{{Par2| |clip| |}} | ||
| − | :::Edge mask clip; usually the input clip processed by [[AWarpsharp2/aSobel|aSobel]] and followed by [[AWarpsharp2/aBlur|aBlur]]. | + | :::Edge mask clip; usually the input clip processed by <tt>[[AWarpsharp2/aSobel|aSobel]]</tt> and followed by <tt>[[AWarpsharp2/aBlur|aBlur]]</tt>. |
| − | :::If you want more stable results it's good idea to do anti-aliasing and temporal filtering on the edge mask before passing it to the warping stage. | + | :::If you want more stable results it's a good idea to do anti-aliasing and temporal filtering (if necessary) on the edge mask before passing it to the warping stage. |
<br> | <br> | ||
::{{Par2|depth|int|3}} | ::{{Par2|depth|int|3}} | ||
| − | :::Strength of the final warping. Negative values result in warping in opposite direction. | + | :::Strength of the final warping. Negative values result in warping in opposite direction, i.e. will blur the image instead of sharpening. |
:::Range: -128 to 127 | :::Range: -128 to 127 | ||
| + | :::<span style="color:red">'''Note:'''</span> Chroma channels are internally processed with <code>{{Template:FuncDef|depth}}/2</code>. | ||
<br> | <br> | ||
::{{Par2|chroma|int|4}} | ::{{Par2|chroma|int|4}} | ||
| − | :::Processing mode for chroma | + | :::Processing mode for the chroma channels (U and V): |
| − | + | :::* 0 : fill with 0x80(128), output is grayscale. | |
| − | + | :::* 1 : don't care - chroma will be trashed. | |
| − | + | :::* 2 : copy chroma channels from the input clip. | |
| − | + | :::* 3 : process chroma; create an edge mask from each chroma channel and use those to warp each chroma channel individually. | |
| − | + | :::* 4 : process chroma; use the edge mask from the luma to warp the chroma channels. | |
| − | + | :::* 5 : same as 3, but don't process luma. | |
| − | + | :::* 6 : same as 4, but don't process luma. | |
| + | :::Luma plane (Y) is always processed, except for mode 5 and 6 which simply copy the luma channel from the input clip. | ||
<br> | <br> | ||
== Examples == | == Examples == | ||
| − | Correct usage examples:<br> | + | Correct usage examples[http://forum.doom9.org/showpost.php?p=1567371&postcount=67]:<br> |
<br> | <br> | ||
*Using [[Resize|Spline36Resize]] to upscale: | *Using [[Resize|Spline36Resize]] to upscale: | ||
| Line 58: | Line 55: | ||
aWarp4(input4x, edge_mask, depth=2) | aWarp4(input4x, edge_mask, depth=2) | ||
<br> | <br> | ||
| − | *The 4x upsampled clip should always be top-left aligned, so Spline36Resize(width*4, height*4) or nnedi3_rpow2(rfactor=4) '''will not''' produce correct results. | + | *The 4x upsampled clip should always be top-left aligned, so <code>Spline36Resize(width*4, height*4)</code> or <code>nnedi3_rpow2(rfactor=4)</code> '''will not''' produce correct results. |
<br> | <br> | ||
-------------------------------------------------------- | -------------------------------------------------------- | ||
'''Back to [[aWarpSharp2#Filters|aWarpSharp2]] ←''' | '''Back to [[aWarpSharp2#Filters|aWarpSharp2]] ←''' | ||
Latest revision as of 00:55, 10 November 2015
aWarp4 is designed for an even more custom warp-sharpening than aWarp allows. Unlike aWarp, this uses a 4x larger source clip than the edge mask. This is useful for improving the quality of the subpixel interpolation quality in the final step. Supersampling allows to produce a sharper result with lower depth, which equals to less deformations.
[edit] Syntax and Parameters
- aWarp4 (clip, clip, int "depth", int "chroma")
- clip =
- Input clip; must be exactly 4 times the width and height of the edge mask clip. The 4x upsampled clip should always be top-left aligned, take a look at the examples below.
- clip =
- int depth = 3
- Strength of the final warping. Negative values result in warping in opposite direction, i.e. will blur the image instead of sharpening.
- Range: -128 to 127
- Note: Chroma channels are internally processed with
depth/2.
- int depth = 3
- int chroma = 4
- Processing mode for the chroma channels (U and V):
- 0 : fill with 0x80(128), output is grayscale.
- 1 : don't care - chroma will be trashed.
- 2 : copy chroma channels from the input clip.
- 3 : process chroma; create an edge mask from each chroma channel and use those to warp each chroma channel individually.
- 4 : process chroma; use the edge mask from the luma to warp the chroma channels.
- 5 : same as 3, but don't process luma.
- 6 : same as 4, but don't process luma.
- Luma plane (Y) is always processed, except for mode 5 and 6 which simply copy the luma channel from the input clip.
- Processing mode for the chroma channels (U and V):
- int chroma = 4
[edit] Examples
Correct usage examples[1]:
- Using Spline36Resize to upscale:
input = AviSource("Blah.avi") input4x = input.Spline36Resize(width*4, height*4, src_left=0.375, src_top=0.375) edge_mask = input.aSobel(thresh=128, chroma=1).aBlur(blur=3, type=1, chroma=1)
aWarp4(input4x, edge_mask, depth=3, chroma=4)
The following example is written differently but is identical to the one above.
AviSource("Blah.avi") aWarp4(Spline36Resize(width*4, height*4, 0.375, 0.375), aSobel().aBlur())
- Using nnedi3_rpow2 and Spline36Resize to upscale:
input = AviSource("Blah.avi") input4x = input.nnedi3_rpow2(rfactor=2).Spline36Resize(width*4, height*4, 0.25, 0.25) edge_mask = input.aSobel().aBlur()
aWarp4(input4x, edge_mask)
- Using nnedi3_rpow2 to upscale:
input = AviSource("Blah.avi") input4x = input.nnedi3_rpow2(rfactor=2).nnedi3_rpow2(rfactor=2) edge_mask = input.aSobel().aBlur()
aWarp4(input4x, edge_mask, depth=2)
- The 4x upsampled clip should always be top-left aligned, so
Spline36Resize(width*4, height*4)ornnedi3_rpow2(rfactor=4)will not produce correct results.
Back to aWarpSharp2 ←
