diff mbox

[FFmpeg-devel] avfilter: add gblur filter

Message ID CAPYw7P7XpoaBVYxrDJEezRonT-N9v2dtK3TWcm+2TSgEVU7KBA@mail.gmail.com
State Accepted
Headers show

Commit Message

Paul B Mahol Sept. 1, 2016, 1:03 p.m. UTC
Hi,

patch attached.

Comments

Michael Niedermayer Sept. 4, 2016, 11:15 a.m. UTC | #1
On Thu, Sep 01, 2016 at 03:03:18PM +0200, Paul B Mahol wrote:
> Hi,
> 
> patch attached.

>  doc/filters.texi         |   17 ++
>  libavfilter/Makefile     |    1 
>  libavfilter/allfilters.c |    1 
>  libavfilter/vf_gblur.c   |  276 +++++++++++++++++++++++++++++++++++++++++++++++
>  4 files changed, 295 insertions(+)
> e14454fae179168294a89bf21ea0d2b1fb0fa41a  0001-avfilter-add-gblur-filter.patch
> From dca28955cfa774c0efed615cb5a0434f663731ec Mon Sep 17 00:00:00 2001
> From: Paul B Mahol <onemda@gmail.com>
> Date: Thu, 1 Sep 2016 01:18:45 +0200
> Subject: [PATCH] avfilter: add gblur filter
> 
> Signed-off-by: Paul B Mahol <onemda@gmail.com>
> ---
>  doc/filters.texi         |  17 +++
>  libavfilter/Makefile     |   1 +
>  libavfilter/allfilters.c |   1 +
>  libavfilter/vf_gblur.c   | 276 +++++++++++++++++++++++++++++++++++++++++++++++
>  4 files changed, 295 insertions(+)
>  create mode 100644 libavfilter/vf_gblur.c
> 
> diff --git a/doc/filters.texi b/doc/filters.texi
> index 88c6f65..cf43edb 100644
> --- a/doc/filters.texi
> +++ b/doc/filters.texi
> @@ -8231,6 +8231,23 @@ option may cause flicker since the B-Frames have often larger QP. Default is
>  
>  @end table
>  
> +@section gblur
> +
> +Apply Gaussian blur filter.
> +
> +The filter accepts the following options:
> +
> +@table @option
> +@item sigma
> +Set sigma, standard deviation of Gaussian blur. Default is @code{0.5}.
> +
> +@item steps
> +Set number of steps for Gaussian approximation. Defauls is @code{4}.
> +
> +@item planes
> +Set which planes to filter. By default all planes are filtered.
> +@end table
> +
>  @section geq
>  
>  The filter accepts the following options:
> diff --git a/libavfilter/Makefile b/libavfilter/Makefile
> index 81b40ac..0074ca3 100644
> --- a/libavfilter/Makefile
> +++ b/libavfilter/Makefile
> @@ -183,6 +183,7 @@ OBJS-$(CONFIG_FRAMESTEP_FILTER)              += vf_framestep.o
>  OBJS-$(CONFIG_FREI0R_FILTER)                 += vf_frei0r.o
>  OBJS-$(CONFIG_FSPP_FILTER)                   += vf_fspp.o
>  OBJS-$(CONFIG_GEQ_FILTER)                    += vf_geq.o
> +OBJS-$(CONFIG_GBLUR_FILTER)                  += vf_gblur.o
>  OBJS-$(CONFIG_GRADFUN_FILTER)                += vf_gradfun.o
>  OBJS-$(CONFIG_HALDCLUT_FILTER)               += vf_lut3d.o dualinput.o framesync.o
>  OBJS-$(CONFIG_HFLIP_FILTER)                  += vf_hflip.o
> diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
> index ee6f9cb..0234833 100644
> --- a/libavfilter/allfilters.c
> +++ b/libavfilter/allfilters.c
> @@ -200,6 +200,7 @@ void avfilter_register_all(void)
>      REGISTER_FILTER(FREI0R,         frei0r,         vf);
>      REGISTER_FILTER(FSPP,           fspp,           vf);
>      REGISTER_FILTER(GEQ,            geq,            vf);
> +    REGISTER_FILTER(GBLUR,          gblur,          vf);
>      REGISTER_FILTER(GRADFUN,        gradfun,        vf);
>      REGISTER_FILTER(HALDCLUT,       haldclut,       vf);
>      REGISTER_FILTER(HFLIP,          hflip,          vf);
> diff --git a/libavfilter/vf_gblur.c b/libavfilter/vf_gblur.c
> new file mode 100644
> index 0000000..fc316af
> --- /dev/null
> +++ b/libavfilter/vf_gblur.c
> @@ -0,0 +1,276 @@
> +/*
> + * Copyright (c) 2011 Pascal Getreuer
> + *
> + * 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
> + */
> +
> +#include "libavutil/imgutils.h"
> +#include "libavutil/opt.h"
> +#include "libavutil/pixdesc.h"
> +#include "avfilter.h"
> +#include "formats.h"
> +#include "internal.h"
> +#include "video.h"
> +
> +typedef struct GBlurContext {
> +    const AVClass *class;
> +
> +    float sigma;
> +    int steps;
> +    int planes;
> +
> +    int depth;
> +    int planewidth[4];
> +    int planeheight[4];
> +    float *buffer;
> +    int nb_planes;
> +} GBlurContext;
> +
> +#define OFFSET(x) offsetof(GBlurContext, x)
> +#define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
> +
> +static const AVOption gblur_options[] = {
> +    { "sigma",  "set sigma",            OFFSET(sigma),  AV_OPT_TYPE_FLOAT, {.dbl=0.5}, 0.0, 1024, FLAGS },
> +    { "steps",  "set number of steps",  OFFSET(steps),  AV_OPT_TYPE_INT,   {.i64=4},     1,   24, FLAGS },
> +    { "planes", "set planes to filter", OFFSET(planes), AV_OPT_TYPE_INT,   {.i64=0xF},   0,  0xF, FLAGS },
> +    { NULL }
> +};
> +
> +AVFILTER_DEFINE_CLASS(gblur);
> +
> +static void gaussianiir2d(float *buffer, int width, int height,
> +                          float sigma, int steps)
> +{
> +    const int numpixels = width * height;
> +    float nu, boundaryscale, postscale;
> +    double lambda, dnu;
> +    int i, x, y, step;
> +    float *ptr;
> +
> +    if (sigma <= 0 || steps < 0)
> +        return;
> +
> +    lambda = (sigma * sigma) / (2.0 * steps);
> +    dnu = (1.0 + 2.0*lambda - sqrt(1.0 + 4.0 * lambda)) / (2.0 * lambda);
> +    postscale = pow(dnu / lambda, 2*steps);
> +    boundaryscale = 1.0 / (1.0 - dnu);
> +    nu = (float)dnu;
> +
> +    /* Filter horizontally along each row */
> +    for (y = 0; y < height; y++) {
> +        for (step = 0; step < steps; step++) {
> +            ptr = buffer + width*y;
> +            ptr[0] *= boundaryscale;
> +
> +            /* Filter rightwards */
> +            for (x = 1; x < width; x++)
> +                ptr[x] += nu*ptr[x - 1];
> +
> +            ptr[x = width - 1] *= boundaryscale;
> +
> +            /* Filter leftwards */
> +            for (; x > 0; x--)
> +                ptr[x - 1] += nu*ptr[x];
> +        }
> +    }
> +

> +    /* Filter vertically along each column */
> +    for (x = 0; x < width; x++) {
> +        for (step = 0; step < steps; step++) {
> +            ptr = buffer + x;
> +            ptr[0] *= boundaryscale;
> +
> +            /* Filter downwards */
> +            for (i = width; i < numpixels; i += width)
> +                ptr[i] += nu*ptr[i - width];
> +
> +            ptr[i = numpixels - width] *= boundaryscale;
> +
> +            /* Filter upwards */
> +            for (; i > 0; i -= width)
> +                ptr[i - width] += nu*ptr[i];
> +        }
> +    }

This is a bit hard on the CPU cache, doing multiple colums at once
likely is faster
also this can be done much faster with SIMD

[...]
diff mbox

Patch

From dca28955cfa774c0efed615cb5a0434f663731ec Mon Sep 17 00:00:00 2001
From: Paul B Mahol <onemda@gmail.com>
Date: Thu, 1 Sep 2016 01:18:45 +0200
Subject: [PATCH] avfilter: add gblur filter

Signed-off-by: Paul B Mahol <onemda@gmail.com>
---
 doc/filters.texi         |  17 +++
 libavfilter/Makefile     |   1 +
 libavfilter/allfilters.c |   1 +
 libavfilter/vf_gblur.c   | 276 +++++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 295 insertions(+)
 create mode 100644 libavfilter/vf_gblur.c

diff --git a/doc/filters.texi b/doc/filters.texi
index 88c6f65..cf43edb 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -8231,6 +8231,23 @@  option may cause flicker since the B-Frames have often larger QP. Default is
 
 @end table
 
+@section gblur
+
+Apply Gaussian blur filter.
+
+The filter accepts the following options:
+
+@table @option
+@item sigma
+Set sigma, standard deviation of Gaussian blur. Default is @code{0.5}.
+
+@item steps
+Set number of steps for Gaussian approximation. Defauls is @code{4}.
+
+@item planes
+Set which planes to filter. By default all planes are filtered.
+@end table
+
 @section geq
 
 The filter accepts the following options:
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index 81b40ac..0074ca3 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -183,6 +183,7 @@  OBJS-$(CONFIG_FRAMESTEP_FILTER)              += vf_framestep.o
 OBJS-$(CONFIG_FREI0R_FILTER)                 += vf_frei0r.o
 OBJS-$(CONFIG_FSPP_FILTER)                   += vf_fspp.o
 OBJS-$(CONFIG_GEQ_FILTER)                    += vf_geq.o
+OBJS-$(CONFIG_GBLUR_FILTER)                  += vf_gblur.o
 OBJS-$(CONFIG_GRADFUN_FILTER)                += vf_gradfun.o
 OBJS-$(CONFIG_HALDCLUT_FILTER)               += vf_lut3d.o dualinput.o framesync.o
 OBJS-$(CONFIG_HFLIP_FILTER)                  += vf_hflip.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index ee6f9cb..0234833 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -200,6 +200,7 @@  void avfilter_register_all(void)
     REGISTER_FILTER(FREI0R,         frei0r,         vf);
     REGISTER_FILTER(FSPP,           fspp,           vf);
     REGISTER_FILTER(GEQ,            geq,            vf);
+    REGISTER_FILTER(GBLUR,          gblur,          vf);
     REGISTER_FILTER(GRADFUN,        gradfun,        vf);
     REGISTER_FILTER(HALDCLUT,       haldclut,       vf);
     REGISTER_FILTER(HFLIP,          hflip,          vf);
diff --git a/libavfilter/vf_gblur.c b/libavfilter/vf_gblur.c
new file mode 100644
index 0000000..fc316af
--- /dev/null
+++ b/libavfilter/vf_gblur.c
@@ -0,0 +1,276 @@ 
+/*
+ * Copyright (c) 2011 Pascal Getreuer
+ *
+ * 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
+ */
+
+#include "libavutil/imgutils.h"
+#include "libavutil/opt.h"
+#include "libavutil/pixdesc.h"
+#include "avfilter.h"
+#include "formats.h"
+#include "internal.h"
+#include "video.h"
+
+typedef struct GBlurContext {
+    const AVClass *class;
+
+    float sigma;
+    int steps;
+    int planes;
+
+    int depth;
+    int planewidth[4];
+    int planeheight[4];
+    float *buffer;
+    int nb_planes;
+} GBlurContext;
+
+#define OFFSET(x) offsetof(GBlurContext, x)
+#define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
+
+static const AVOption gblur_options[] = {
+    { "sigma",  "set sigma",            OFFSET(sigma),  AV_OPT_TYPE_FLOAT, {.dbl=0.5}, 0.0, 1024, FLAGS },
+    { "steps",  "set number of steps",  OFFSET(steps),  AV_OPT_TYPE_INT,   {.i64=4},     1,   24, FLAGS },
+    { "planes", "set planes to filter", OFFSET(planes), AV_OPT_TYPE_INT,   {.i64=0xF},   0,  0xF, FLAGS },
+    { NULL }
+};
+
+AVFILTER_DEFINE_CLASS(gblur);
+
+static void gaussianiir2d(float *buffer, int width, int height,
+                          float sigma, int steps)
+{
+    const int numpixels = width * height;
+    float nu, boundaryscale, postscale;
+    double lambda, dnu;
+    int i, x, y, step;
+    float *ptr;
+
+    if (sigma <= 0 || steps < 0)
+        return;
+
+    lambda = (sigma * sigma) / (2.0 * steps);
+    dnu = (1.0 + 2.0*lambda - sqrt(1.0 + 4.0 * lambda)) / (2.0 * lambda);
+    postscale = pow(dnu / lambda, 2*steps);
+    boundaryscale = 1.0 / (1.0 - dnu);
+    nu = (float)dnu;
+
+    /* Filter horizontally along each row */
+    for (y = 0; y < height; y++) {
+        for (step = 0; step < steps; step++) {
+            ptr = buffer + width*y;
+            ptr[0] *= boundaryscale;
+
+            /* Filter rightwards */
+            for (x = 1; x < width; x++)
+                ptr[x] += nu*ptr[x - 1];
+
+            ptr[x = width - 1] *= boundaryscale;
+
+            /* Filter leftwards */
+            for (; x > 0; x--)
+                ptr[x - 1] += nu*ptr[x];
+        }
+    }
+
+    /* Filter vertically along each column */
+    for (x = 0; x < width; x++) {
+        for (step = 0; step < steps; step++) {
+            ptr = buffer + x;
+            ptr[0] *= boundaryscale;
+
+            /* Filter downwards */
+            for (i = width; i < numpixels; i += width)
+                ptr[i] += nu*ptr[i - width];
+
+            ptr[i = numpixels - width] *= boundaryscale;
+
+            /* Filter upwards */
+            for (; i > 0; i -= width)
+                ptr[i - width] += nu*ptr[i];
+        }
+    }
+
+    for (i = 0; i < numpixels; i++)
+        buffer[i] *= postscale;
+
+    return;
+}
+
+static int query_formats(AVFilterContext *ctx)
+{
+    static const enum AVPixelFormat pix_fmts[] = {
+        AV_PIX_FMT_YUVA444P, AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV440P,
+        AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVJ440P,
+        AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUV420P,
+        AV_PIX_FMT_YUVJ422P, AV_PIX_FMT_YUVJ420P,
+        AV_PIX_FMT_YUVJ411P, AV_PIX_FMT_YUV411P, AV_PIX_FMT_YUV410P,
+        AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV444P9,
+        AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV444P10,
+        AV_PIX_FMT_YUV420P12, AV_PIX_FMT_YUV422P12, AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV440P12,
+        AV_PIX_FMT_YUV420P14, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV444P14,
+        AV_PIX_FMT_YUV420P16, AV_PIX_FMT_YUV422P16, AV_PIX_FMT_YUV444P16,
+        AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA444P9,
+        AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA444P10,
+        AV_PIX_FMT_YUVA420P16, AV_PIX_FMT_YUVA422P16, AV_PIX_FMT_YUVA444P16,
+        AV_PIX_FMT_GBRP, AV_PIX_FMT_GBRP9, AV_PIX_FMT_GBRP10,
+        AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRP14, AV_PIX_FMT_GBRP16,
+        AV_PIX_FMT_GBRAP, AV_PIX_FMT_GBRAP12, AV_PIX_FMT_GBRAP16,
+        AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY16,
+        AV_PIX_FMT_NONE
+    };
+
+    return ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
+}
+
+static int config_input(AVFilterLink *inlink)
+{
+    GBlurContext *s = inlink->dst->priv;
+    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
+
+    s->depth = desc->comp[0].depth;
+    s->planewidth[1] = s->planewidth[2] = AV_CEIL_RSHIFT(inlink->w, desc->log2_chroma_w);
+    s->planewidth[0] = s->planewidth[3] = inlink->w;
+    s->planeheight[1] = s->planeheight[2] = AV_CEIL_RSHIFT(inlink->h, desc->log2_chroma_h);
+    s->planeheight[0] = s->planeheight[3] = inlink->h;
+
+    s->nb_planes = av_pix_fmt_count_planes(inlink->format);
+
+    s->buffer = av_malloc_array(inlink->w, inlink->h * sizeof(*s->buffer));
+    if (!s->buffer)
+        return AVERROR(ENOMEM);
+    return 0;
+}
+
+static int filter_frame(AVFilterLink *inlink, AVFrame *in)
+{
+    GBlurContext *s = inlink->dst->priv;
+    AVFilterLink *outlink = inlink->dst->outputs[0];
+    AVFrame *out;
+    int plane;
+
+    if (av_frame_is_writable(in)) {
+        out = in;
+    } else {
+        out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
+        if (!out) {
+            av_frame_free(&in);
+            return AVERROR(ENOMEM);
+        }
+        av_frame_copy_props(out, in);
+    }
+
+    for (plane = 0; plane < s->nb_planes; plane++) {
+        const int height = s->planeheight[plane];
+        const int width = s->planewidth[plane];
+        float *bptr = s->buffer;
+        const uint8_t *src = in->data[plane];
+        const uint16_t *src16 = (const uint16_t *)in->data[plane];
+        uint8_t *dst = out->data[plane];
+        uint16_t *dst16 = (uint16_t *)out->data[plane];
+        int y, x;
+
+        if (!s->sigma || !(s->planes & (1 << plane))) {
+            if (out != in)
+                av_image_copy_plane(out->data[plane], out->linesize[plane],
+                                    in->data[plane], in->linesize[plane],
+                                    width * ((s->depth + 7) / 8), height);
+            continue;
+        }
+
+        if (s->depth == 8) {
+            for (y = 0; y < height; y++) {
+                for (x = 0; x < width; x++) {
+                    bptr[x] = src[x];
+                }
+                bptr += width;
+                src += in->linesize[plane];
+            }
+        } else {
+            for (y = 0; y < height; y++) {
+                for (x = 0; x < width; x++) {
+                    bptr[x] = src16[x];
+                }
+                bptr += width;
+                src16 += in->linesize[plane] / 2;
+            }
+        }
+
+        gaussianiir2d(s->buffer, s->planewidth[plane], s->planeheight[plane], s->sigma, s->steps);
+
+        bptr = s->buffer;
+        if (s->depth == 8) {
+            for (y = 0; y < height; y++) {
+                for (x = 0; x < width; x++) {
+                    dst[x] = bptr[x];
+                }
+                bptr += width;
+                dst += out->linesize[plane];
+            }
+        } else {
+            for (y = 0; y < height; y++) {
+                for (x = 0; x < width; x++) {
+                    dst16[x] = bptr[x];
+                }
+                bptr += width;
+                dst16 += out->linesize[plane] / 2;
+            }
+        }
+    }
+
+    if (out != in)
+        av_frame_free(&in);
+    return ff_filter_frame(outlink, out);
+}
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+    GBlurContext *s = ctx->priv;
+
+    av_freep(&s->buffer);
+}
+
+static const AVFilterPad gblur_inputs[] = {
+    {
+        .name         = "default",
+        .type         = AVMEDIA_TYPE_VIDEO,
+        .config_props = config_input,
+        .filter_frame = filter_frame,
+    },
+    { NULL }
+};
+
+static const AVFilterPad gblur_outputs[] = {
+    {
+        .name = "default",
+        .type = AVMEDIA_TYPE_VIDEO,
+    },
+    { NULL }
+};
+
+AVFilter ff_vf_gblur = {
+    .name          = "gblur",
+    .description   = NULL_IF_CONFIG_SMALL("Apply Gaussian Blur filter."),
+    .priv_size     = sizeof(GBlurContext),
+    .priv_class    = &gblur_class,
+    .uninit        = uninit,
+    .query_formats = query_formats,
+    .inputs        = gblur_inputs,
+    .outputs       = gblur_outputs,
+    .flags         = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
+};
-- 
2.5.0