[FFmpeg-devel] [PATCH] avfilter: add mergeplanes

Stefano Sabatini stefasab at gmail.com
Thu Oct 3 18:09:41 CEST 2013


On date Wednesday 2013-10-02 14:56:54 +0000, Paul B Mahol encoded:
> Signed-off-by: Paul B Mahol <onemda at gmail.com>
> ---
>  doc/filters.texi             |  22 +++
>  libavfilter/Makefile         |   1 +
>  libavfilter/allfilters.c     |   1 +
>  libavfilter/vf_mergeplanes.c | 314 +++++++++++++++++++++++++++++++++++++++++++
>  4 files changed, 338 insertions(+)
>  create mode 100644 libavfilter/vf_mergeplanes.c
> 
> diff --git a/doc/filters.texi b/doc/filters.texi
> index efadaf1..571afa4 100644
> --- a/doc/filters.texi
> +++ b/doc/filters.texi
> @@ -5248,6 +5248,28 @@ lutyuv=y='bitand(val, 128+64+32)'
>  @end example
>  @end itemize
>  
> + at section mergeplanes
> +
> +Merge color channel components from several video streams.
> +
> +This filter accepts the following options:
> + at table @option
> + at item inputs
> +Set number of inputs. Default is @code{4}

missing ending points

> + at item format
> +Set output pixel format. Default is @code{yuva444p}.
> + at end table

>From this description it is not clear what the inputs must be, nor how
the input planes are merged.

> +
> + at subsection Examples
> +
> + at itemize
> + at item
> +Merge three gray video streams of same width and height into single video stream:
> + at example
> +mergeplanes=3:yuv444p
> + at end example
> + at end itemize
> +
>  @section mcdeint
>  
>  Apply motion-compensation deinterlacing.
> diff --git a/libavfilter/Makefile b/libavfilter/Makefile
> index b2d3587..0200e29 100644
> --- a/libavfilter/Makefile
> +++ b/libavfilter/Makefile
> @@ -157,6 +157,7 @@ OBJS-$(CONFIG_LUT_FILTER)                    += vf_lut.o
>  OBJS-$(CONFIG_LUTRGB_FILTER)                 += vf_lut.o
>  OBJS-$(CONFIG_LUTYUV_FILTER)                 += vf_lut.o
>  OBJS-$(CONFIG_MCDEINT_FILTER)                += vf_mcdeint.o
> +OBJS-$(CONFIG_MERGEPLANES_FILTER)            += vf_mergeplanes.o
>  OBJS-$(CONFIG_MP_FILTER)                     += vf_mp.o
>  OBJS-$(CONFIG_MPDECIMATE_FILTER)             += vf_mpdecimate.o
>  OBJS-$(CONFIG_NEGATE_FILTER)                 += vf_lut.o
> diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
> index ed11d67..8f33281 100644
> --- a/libavfilter/allfilters.c
> +++ b/libavfilter/allfilters.c
> @@ -153,6 +153,7 @@ void avfilter_register_all(void)
>      REGISTER_FILTER(LUTRGB,         lutrgb,         vf);
>      REGISTER_FILTER(LUTYUV,         lutyuv,         vf);
>      REGISTER_FILTER(MCDEINT,        mcdeint,        vf);
> +    REGISTER_FILTER(MERGEPLANES,    mergeplanes,    vf);
>      REGISTER_FILTER(MP,             mp,             vf);
>      REGISTER_FILTER(MPDECIMATE,     mpdecimate,     vf);
>      REGISTER_FILTER(NEGATE,         negate,         vf);
> diff --git a/libavfilter/vf_mergeplanes.c b/libavfilter/vf_mergeplanes.c
> new file mode 100644
> index 0000000..3d200e4
> --- /dev/null
> +++ b/libavfilter/vf_mergeplanes.c
> @@ -0,0 +1,314 @@
> +/*
> + * Copyright (c) 2013 Paul B Mahol
> + *
> + * 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/avassert.h"
> +#include "libavutil/avstring.h"
> +#include "libavutil/imgutils.h"
> +#include "libavutil/opt.h"
> +#include "libavutil/pixdesc.h"
> +#include "avfilter.h"
> +#include "internal.h"
> +#include "framesync.h"
> +
> +typedef struct InputParam {
> +    int nb_planes;
> +    int planewidth[4];
> +    int planeheight[4];
> +    enum AVPixelFormat format;
> +} InputParam;
> +
> +typedef struct MergePlanesContext {
> +    const AVClass *class;
> +    int nb_inputs;
> +    const enum AVPixelFormat out_fmt;
> +    int nb_planes;
> +    int planewidth[4];
> +    int planeheight[4];
> +    const AVPixFmtDescriptor *outdesc;
> +    InputParam inputp[4];
> +
> +    FFFrameSync fs;
> +    FFFrameSyncIn fsi[3]; /* must be immediately after fs */
> +} MergePlanesContext;
> +
> +#define OFFSET(x) offsetof(MergePlanesContext, x)
> +#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
> +static const AVOption mergeplanes_options[] = {
> +    { "inputs", "set number of inputs", OFFSET(nb_inputs), AV_OPT_TYPE_INT, {.i64=4}, 2, 4, FLAGS },
> +    { "format", "set output pixel format", OFFSET(out_fmt), AV_OPT_TYPE_PIXEL_FMT, {.i64=AV_PIX_FMT_YUVA444P}, .flags=FLAGS},
> +    { NULL }
> +};
> +
> +AVFILTER_DEFINE_CLASS(mergeplanes);
> +
> +static int query_formats(AVFilterContext *ctx)
> +{
> +    MergePlanesContext *s = ctx->priv;
> +    AVFilterFormats *formats;
> +    int i;
> +
> +    for (i = 0; i < s->nb_inputs; i++) {
> +        formats = NULL;
> +        ff_add_format(&formats, s->inputp[i].format);
> +        ff_formats_ref(formats, &ctx->inputs[i]->out_formats);
> +    }
> +
> +    formats = NULL;
> +    ff_add_format(&formats, s->out_fmt);
> +    ff_formats_ref(formats, &ctx->outputs[0]->in_formats);
> +
> +    return 0;
> +}
> +

> +static int process_frame(FFFrameSync *fs)
> +{
> +    AVFilterContext *ctx = fs->parent;
> +    AVFilterLink *outlink = ctx->outputs[0];
> +    MergePlanesContext *s = fs->opaque;
> +    AVFrame *in[4] = { NULL };
> +    AVFrame *out;
> +    int i, p, j = 0, ret;
> +
> +    for (i = 0; i < s->nb_inputs; i++) {
> +        if ((ret = ff_framesync_get_frame(&s->fs, i, &in[i], 0)) < 0)
> +            return ret;
> +    }
> +
> +    out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
> +    if (!out)
> +        return AVERROR(ENOMEM);
> +    out->pts = av_rescale_q(s->fs.pts, s->fs.time_base, outlink->time_base);
> +

> +    for (i = 0; i < s->nb_inputs; i++) {

nit: initialize j = 0 here

> +        InputParam *inputp = &s->inputp[i];
> +
> +        for (p = 0; p < inputp->nb_planes; p++, j++) {
> +            av_image_copy_plane(out->data[j], out->linesize[j],
> +                                in[i]->data[p], in[i]->linesize[p],
> +                                inputp->planewidth[p], inputp->planeheight[p]);
> +        }
> +    }
> +
> +    return ff_filter_frame(outlink, out);
> +}
> +
> +static int config_output(AVFilterLink *outlink)
> +{
> +    AVFilterContext *ctx = outlink->src;
> +    MergePlanesContext *s = ctx->priv;
> +    FFFrameSyncIn *in = s->fs.in;
> +    int depth[4][4];
> +    int plane, i, j;
> +
> +    ff_framesync_init(&s->fs, ctx, s->nb_inputs);
> +    s->fs.opaque = s;
> +    s->fs.on_event = process_frame;
> +
> +    outlink->w = ctx->inputs[0]->w;
> +    outlink->h = ctx->inputs[0]->h;
> +    outlink->time_base = ctx->inputs[0]->time_base;
> +    outlink->frame_rate = ctx->inputs[0]->frame_rate;
> +    outlink->sample_aspect_ratio = ctx->inputs[0]->sample_aspect_ratio;
> +
> +    s->planewidth[1]  =
> +    s->planewidth[2]  = FF_CEIL_RSHIFT(outlink->w, s->outdesc->log2_chroma_w);
> +    s->planewidth[0]  =
> +    s->planewidth[3]  = outlink->w;
> +    s->planeheight[1] =
> +    s->planeheight[2] = FF_CEIL_RSHIFT(outlink->h, s->outdesc->log2_chroma_h);
> +    s->planeheight[0] =
> +    s->planeheight[3] = outlink->h;
> +    s->nb_planes = av_pix_fmt_count_planes(s->out_fmt);
> +
> +    for (i = 0; i < s->nb_inputs; i++) {
> +        InputParam *inputp = &s->inputp[i];
> +        AVFilterLink *inlink = ctx->inputs[i];
> +        const AVPixFmtDescriptor *indesc = av_pix_fmt_desc_get(inlink->format);
> +
> +        if (outlink->sample_aspect_ratio.num != inlink->sample_aspect_ratio.num ||
> +            outlink->sample_aspect_ratio.den != inlink->sample_aspect_ratio.den) {
> +            av_log(ctx, AV_LOG_ERROR, "input #%d link %s SAR %d:%d "
> +                                      "do not match output link %s SAR %d:%d\n",
> +                                      i, ctx->input_pads[i].name,
> +                                      inlink->sample_aspect_ratio.num,
> +                                      inlink->sample_aspect_ratio.den,
> +                                      ctx->output_pads[0].name,
> +                                      outlink->sample_aspect_ratio.num,
> +                                      outlink->sample_aspect_ratio.den);
> +            return AVERROR(EINVAL);
> +        }
> +
> +        inputp->planewidth[1]  =
> +        inputp->planewidth[2]  = FF_CEIL_RSHIFT(inlink->w, indesc->log2_chroma_w);
> +        inputp->planewidth[0]  =
> +        inputp->planewidth[3]  = inlink->w;
> +        inputp->planeheight[1] =
> +        inputp->planeheight[2] = FF_CEIL_RSHIFT(inlink->h, indesc->log2_chroma_h);
> +        inputp->planeheight[0] =
> +        inputp->planeheight[3] = inlink->h;
> +        inputp->nb_planes = av_pix_fmt_count_planes(inlink->format);
> +

> +        for (j = 0; j < inputp->nb_planes; j++)
> +            depth[i][j] = indesc->comp[j].depth_minus1;

maybe depth_minus1+1 for readability's sake

> +
> +        in[i].time_base = inlink->time_base;
> +        in[i].sync   = 1;
> +        in[i].before = EXT_STOP;
> +        in[i].after  = EXT_STOP;
> +    }
> +

> +    for (i = 0, plane = 0; i < s->nb_inputs; i++) {
> +        InputParam *inputp = &s->inputp[i];
> +
> +        for (j = 0; j < inputp->nb_planes; j++, plane++) {
> +            if (plane >= s->nb_planes)
> +                goto fail;
> +            if (s->outdesc->comp[plane].depth_minus1 != depth[i][j]) {
> +                av_log(ctx, AV_LOG_ERROR, "output plane %d depth %d does not "
> +                                          "match input %d plane %d depth %d\n",
> +                                          plane, s->outdesc->comp[plane].depth_minus1 + 1,
> +                                          i, j, depth[i][j] + 1);
> +                goto fail;
> +            }
> +            if (s->planewidth[plane] != inputp->planewidth[j]) {
> +                av_log(ctx, AV_LOG_ERROR, "output plane %d width %d does not "
> +                                          "match input %d plane %d width %d\n",
> +                                          plane, s->planewidth[plane],
> +                                          i, j, inputp->planewidth[j]);
> +                goto fail;
> +            }
> +            if (s->planeheight[plane] != inputp->planeheight[j]) {
> +                av_log(ctx, AV_LOG_ERROR, "output plane %d height %d does not "
> +                                          "match input %d plane %d height %d\n",
> +                                          plane, s->planeheight[plane],
> +                                          i, j, inputp->planeheight[j]);
> +                goto fail;
> +            }
> +        }
> +    }
> +

> +    if (plane != s->nb_planes) {
> +        av_log(ctx, AV_LOG_ERROR, "sum of input planes do not match "
> +                                  "number of planes in output format\n");
> +        goto fail;
> +    }

Nit: providing some numbers help users/debugging.

The sum of input planes (%d) does not match the number of planes in
output format (%d).

> +
> +    return ff_framesync_configure(&s->fs);
> +fail:
> +    return AVERROR(EINVAL);
> +}
> +
> +static int filter_frame(AVFilterLink *inlink, AVFrame *in)
> +{
> +    MergePlanesContext *s = inlink->dst->priv;
> +    return ff_framesync_filter_frame(&s->fs, inlink, in);
> +}
> +
> +static av_cold int init(AVFilterContext *ctx)
> +{
> +    MergePlanesContext *s = ctx->priv;
> +    int i, ret;
> +
> +    s->outdesc = av_pix_fmt_desc_get(s->out_fmt);
> +    if (!(s->outdesc->flags & AV_PIX_FMT_FLAG_PLANAR)) {
> +        av_log(ctx, AV_LOG_ERROR, "Only planar formats are supported\n");
> +        return AVERROR(EINVAL);
> +    }
> +
> +    switch (s->nb_inputs) {
> +    case 4:
> +    case 3:
> +        s->inputp[0].format =
> +        s->inputp[1].format =
> +        s->inputp[2].format =
> +        s->inputp[3].format = AV_PIX_FMT_GRAY8;
> +        break;
> +    case 2:
> +        switch (s->out_fmt) {

> +        case AV_PIX_FMT_YUVA420P: s->inputp[0].format = AV_PIX_FMT_YUV420P; break;
> +        case AV_PIX_FMT_YUVA422P: s->inputp[0].format = AV_PIX_FMT_YUV422P; break;
> +        case AV_PIX_FMT_YUVA444P: s->inputp[0].format = AV_PIX_FMT_YUV444P; break;

can be simplified: s->inputp[0].format = s->out_fmt; break;

Also, what happens if you copy a gray8 full-range image to YUV420P
luma plane?

> +        case AV_PIX_FMT_GBRAP:    s->inputp[0].format = AV_PIX_FMT_GBRP;    break;
> +        default:
> +            return AVERROR(EINVAL);

meaningful feedback missing here

> +        }
> +        s->inputp[1].format = AV_PIX_FMT_GRAY8;
> +        break;
> +    default:
> +        av_assert0(0);
> +    }
> +
[...]
-- 
FFmpeg = Foolish and Fantastic Muttering Peaceless Ecletic Gorilla


More information about the ffmpeg-devel mailing list