[FFmpeg-devel] [PATCH] lavfi: port ow from libmpcodecs.
Clément Bœsch
ubitux at gmail.com
Thu May 16 00:49:39 CEST 2013
On Tue, May 14, 2013 at 10:59:44PM +0200, Stefano Sabatini wrote:
> On date Friday 2013-05-10 22:10:15 +0200, Clément Bœsch encoded:
> > TODO: Changelog & minor bump
> >
> > ---
> > Feel free to bikeshed for a new name.
> >
> > mp=ow will be dropped in the same time.
> > ---
> > doc/filters.texi | 26 ++++
> > libavfilter/Makefile | 1 +
> > libavfilter/allfilters.c | 1 +
> > libavfilter/vf_ow.c | 344 +++++++++++++++++++++++++++++++++++++++++++++++
> > 4 files changed, 372 insertions(+)
> > create mode 100644 libavfilter/vf_ow.c
> >
> > diff --git a/doc/filters.texi b/doc/filters.texi
> > index fe38b7f..8b9fb4a 100644
> > --- a/doc/filters.texi
> > +++ b/doc/filters.texi
> > @@ -5103,6 +5103,32 @@ testsrc=s=100x100, split=4 [in0][in1][in2][in3];
> >
> > @end itemize
> >
> > + at section ow
> > +
> > +Overcomplete Wavelet denoiser.
>
> Maybe tell when/if this filter is useful compared to other similar
> tools.
>
I have... absolutely no idea :)
> > +
> > +The filter accepts the following options:
> > +
> > + at table @option
>
> > + at item depth
> > +Set depth.
>
> Depth of the sea?
>
It's an arbitrary depth metric AFAICT; the next sentence describes the
behaviour.
> > +
> > +Larger depth values will denoise lower frequency components more, but
> > +slow down filtering
>
> Missing final dot.
>
Added.
> > +
> > +Default is @code{8}.
> > +
> > + at item luma_strength
> > +Set luma strength.
> > +
> > +Default is @code{1.0}
> > +
> > + at item chroma_strength
> > +Set chroma strength.
> > +
> > +Default is @code{1.0}
>
> Also some notes about the strenght is about would be nice.
>
I wonder what we could add...
> > + at end table
> > +
> > @section pad
> >
> > Add paddings to the input image, and place the original input at the
> > diff --git a/libavfilter/Makefile b/libavfilter/Makefile
> > index 73a24dd..f6daed1 100644
> > --- a/libavfilter/Makefile
> > +++ b/libavfilter/Makefile
> > @@ -154,6 +154,7 @@ OBJS-$(CONFIG_NULL_FILTER) += vf_null.o
> > OBJS-$(CONFIG_OCV_FILTER) += vf_libopencv.o
> > OBJS-$(CONFIG_OPENCL) += deshake_opencl.o unsharp_opencl.o
> > OBJS-$(CONFIG_OVERLAY_FILTER) += vf_overlay.o
> > +OBJS-$(CONFIG_OW_FILTER) += vf_ow.o
> > OBJS-$(CONFIG_PAD_FILTER) += vf_pad.o
> > OBJS-$(CONFIG_PERMS_FILTER) += f_perms.o
> > OBJS-$(CONFIG_PIXDESCTEST_FILTER) += vf_pixdesctest.o
> > diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
> > index b8f273d..8986547 100644
> > --- a/libavfilter/allfilters.c
> > +++ b/libavfilter/allfilters.c
> > @@ -151,6 +151,7 @@ void avfilter_register_all(void)
> > REGISTER_FILTER(NULL, null, vf);
> > REGISTER_FILTER(OCV, ocv, vf);
> > REGISTER_FILTER(OVERLAY, overlay, vf);
> > + REGISTER_FILTER(OW, ow, vf);
> > REGISTER_FILTER(PAD, pad, vf);
> > REGISTER_FILTER(PERMS, perms, vf);
> > REGISTER_FILTER(PIXDESCTEST, pixdesctest, vf);
> > diff --git a/libavfilter/vf_ow.c b/libavfilter/vf_ow.c
> > new file mode 100644
> > index 0000000..97c386a
> > --- /dev/null
> > +++ b/libavfilter/vf_ow.c
> > @@ -0,0 +1,344 @@
>
> > +/*
> > + * Copyright (C) 2007 Michael Niedermayer <michaelni at gmx.at>
> > + * Copyright (c) 2013 Clément Bœsch
> > + *
> > + * This file is part of FFmpeg.
> > + *
> > + * FFmpeg is free software; you can redistribute it and/or
> > + * modify it under the terms of the GNU Lesser General Public
> > + * License as published by the Free Software Foundation; either
> > + * version 2.1 of the License, or (at your option) any later version.
> > + *
> > + * FFmpeg is distributed in the hope that it will be useful,
> > + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> > + * Lesser General Public License for more details.
> > + *
> > + * You should have received a copy of the GNU Lesser General Public
> > + * License along with FFmpeg; if not, write to the Free Software
> > + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
>
> Keep original licensing or ask the author (Michael seems the only one
> who did significant changes, things may be different on awesome
> lawyers planet) to relicense it.
>
I restored the GPL. For a filter where I barely have any relevant
contribution nor a personal particular interest in it, I don't feel at
place to ask for a relicensing.
> > + */
> > +
> > +/**
> > + * @todo try to change to int
> > + * @todo try lifting based implementation
> > + * @todo optimize optimize optimize
>
> > + * @todo hard tresholding
>
> thresholding
>
Fixed.
> > + * @todo use QP to decide filter strength
> > + * @todo wavelet normalization / least squares optimal signal vs. noise thresholds
> > + */
> > +
> > +#include "libavutil/imgutils.h"
> > +#include "libavutil/opt.h"
> > +#include "libavutil/pixdesc.h"
> > +#include "avfilter.h"
> > +#include "internal.h"
> > +
> > +typedef struct {
> > + const AVClass *class;
> > + double luma_strength;
> > + double chroma_strength;
> > + int depth;
> > + float *plane[16+1][4];
> > + int linesize;
> > + int hsub, vsub;
> > +} OWContext;
> > +
> > +#define OFFSET(x) offsetof(OWContext, x)
> > +#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
> > +static const AVOption ow_options[] = {
> > + { "depth", "set depth", OFFSET(depth), AV_OPT_TYPE_INT, {.i64 = 8}, 8, 16, FLAGS },
> > + { "luma_strength", "set luma strength", OFFSET(luma_strength), AV_OPT_TYPE_DOUBLE, {.dbl = 1.0 }, 0, 1000, FLAGS },
> > + { "ls", "set luma strength", OFFSET(luma_strength), AV_OPT_TYPE_DOUBLE, {.dbl = 1.0 }, 0, 1000, FLAGS },
> > + { "chroma_strength", "set chroma strength", OFFSET(chroma_strength), AV_OPT_TYPE_DOUBLE, {.dbl = 1.0 }, 0, 1000, FLAGS },
> > + { "cs", "set chroma strength", OFFSET(chroma_strength), AV_OPT_TYPE_DOUBLE, {.dbl = 1.0 }, 0, 1000, FLAGS },
> > + { NULL }
> > +};
> > +
> > +AVFILTER_DEFINE_CLASS(ow);
> > +
> > +DECLARE_ALIGNED(8, static const uint8_t, dither)[8][8] = {
> > + { 0, 48, 12, 60, 3, 51, 15, 63 },
> > + { 32, 16, 44, 28, 35, 19, 47, 31 },
> > + { 8, 56, 4, 52, 11, 59, 7, 55 },
> > + { 40, 24, 36, 20, 43, 27, 39, 23 },
> > + { 2, 50, 14, 62, 1, 49, 13, 61 },
> > + { 34, 18, 46, 30, 33, 17, 45, 29 },
> > + { 10, 58, 6, 54, 9, 57, 5, 53 },
> > + { 42, 26, 38, 22, 41, 25, 37, 21 },
> > +};
> > +
> > +#define SQRT2 1.41421356237
>
> M_SQRT2? Or does it affects bit-exactness?
>
Mmh, I kept it for comparison with mp=ow, changed.
> [...]
>
> No more comments from me without some serious study, but I believe it
> should be fine assuming it is bitexact with the original filter.
>
It results in the same framecrc locally here with my tests.
> "owdenoise" looks like a nice name to me, in case "ow" seems too
> cryptic.
Renamed and applied, thanks. Feel free to improve the documentation.
--
Clément B.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20130516/0b28f30e/attachment.asc>
More information about the ffmpeg-devel
mailing list