Frfun7
From Avisynth wiki
Abstract | |
---|---|
Author | Prunedtree, pinterf |
Version | 0.9 |
Download | frfun7_v0.9.7z |
Category | Spatial Denoisers |
License | GPLv2 |
Discussion | Doom9 Thread (2006) - Doom9 Thread (2021) |
Contents |
Description
frfun7 is a spatial fractal denoising plugin.
Requirements
- [x86]: AviSynth+ or AviSynth 2.6
- [x64]: AviSynth+
- Progressive input only
- Supported color formats: Y8, YV12, YV16, YV24
Syntax and Parameters
- frfun7 (clip, float "lambda", float "T", float "Tuv", int "P", int "TP", int "R1")
- clip =
- Input clip.
- clip =
- float lambda = 1.1
- Adjust the power of the local denoising.
- float lambda = 1.1
- float T = 6.0
- Limits the max luma denoising power for edges; 0 disables processing.
- float T = 6.0
- float Tuv = 2.0
- Limits the max chroma denoising power for edges; 0 disables processing.
- float Tuv = 2.0
- int P = 0
- It's a "speed -vs- quality" trade off setting.
- int P = 0
- 0 : faster but slightly lower quality than frfun7_rev6 (may create minor artifacts around line edges).
- 1 : adaptive overlapping (see also TP1). slower than frfun7_rev6 but the quality is a little bit better.
- 2 : temporal
- 4 : adaptive radius
- Internally the parameter is treated as a bit mask but probably it has no point.
- Parameter is available since frfun7 2013.
- int TP1 = 0
- A threshold which affects P=1 (adaptive overlapping).
- Introduced as a separate parameter in v0.7. This value had to be encoded into P as TP1*1000 previously.
- 0 will always run into a final filtering part, the bigger it is, probably the more pixels it will skip. (?)
- int TP1 = 0
- int R1 = 3
- Radius for first pass of the internal algorithm.
- First pass in pre v0.7 was fixed to 3 (and was no separate parameter).
- Valid values are 2 or 3.
- int R1 = 3
Examples
frfun7 with default settings:
AviSource("Blah.avi") frfun7(lambda=1.1, T=6.0, Tuv=2.0, P=0, TP1=0, R1=3)
Changelog
Version Date Changes
v0.9 2022/06/29 - Performance increase in Intel sse2 - backport from VapourSynth (dubhater): * p=0 184 fps to 304 fps 65% faster *p=1 36 fps to 61 fps 69% faster *p=2 233 fps to 282 fps 21% faster *p=4 263 fps to 401 fps 52% faster *note: The output is slightly different at the left and right edges of the frame. - additional fix bug when p=1 and tp1>0
v0.8 2022/05/18 - backport some fixes and non-Intel C code from VapourSynth port by dubhater: - Fix P=2 (temporal) blocky rectangular artifacts at the most top and bottom area - Fix bug when p=1 and tp1>0 - Avoid loss of precision when p=1 or p=2, Fixes https://github.com/pinterf/Frfun7/issues/1 - Add C-only (other than Intel SIMD) code path, enable usage of ENABLE_INTEL_SIMD build option.
v0.7 2021/09/10 - Code refresh and additions by pinterf - Source to VS2019 solution - Add/guess missing source parts, rename files - Update AviSynth headers - Add version resource - Avisynth V2.6 style plugin - Implement all mmx inline assembler as SIMD intrinsics - x64 build - fix some rounding and other issue - re-enable T=0, Tuv=0 cases (unprocessed plane copy) - add experimental TP1 (default 0) a threshold for P=1 (temporal overlapping) mode - add experimental R1 (default 3, can be set to 2) first pass radius
2013 2013/09/04 - no longer buffers the input; yields a nice speed increase. - "P" parameter added
rev6 2006/05/10 - bug fixes - remove mod8 restriction - process first and last frame
rev1 2006/05/05 - initial release
Archived Downloads
Version | Download | Mirror | Mirror 2 |
---|---|---|---|
2013 | frfun7 2013.zip | frfun7 2013.zip | frfun7_2013.zip |
rev6 | frfun7_rev6.zip | frfun7_rev6.zip | frfun7_rev6.zip |
External Links
- GitHub - Source code repository.
- GitHub - Source code repository (VapourSynth port).
- Doom9 Forum - frfun7_rev6 discussion.
- Doom9 Forum - frfun7 2013 discussion.
- Doom9 Forum - frfun7 2021 discussion.
Back to External Filters ←