f_select.c

Go to the documentation of this file.

/*
 * Copyright (c) 2011 Stefano Sabatini
 *
 * 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
 */
 
/**
 * @file
 * filter for selecting which frame passes in the filterchain
 */
 
#include "config_components.h"
 
#include "libavutil/avstring.h"
#include "libavutil/eval.h"
#include "libavutil/fifo.h"
#include "libavutil/imgutils.h"
#include "libavutil/internal.h"
#include "libavutil/opt.h"
#include "libavutil/pixdesc.h"
#include "avfilter.h"
#include "audio.h"
#include "filters.h"
#include "formats.h"
#include "video.h"
#include "scene_sad.h"
 
static const char *const var_names[] = {
    "TB",                ///< timebase
 
    "pts",               ///< original pts in the file of the frame
    "start_pts",         ///< first PTS in the stream, expressed in TB units
    "prev_pts",          ///< previous frame PTS
    "prev_selected_pts", ///< previous selected frame PTS
 
    "t",                 ///< timestamp expressed in seconds
    "start_t",           ///< first PTS in the stream, expressed in seconds
    "prev_t",            ///< previous frame time
    "prev_selected_t",   ///< previously selected time
 
    "pict_type",         ///< the type of picture in the movie
    "I",
    "P",
    "B",
    "S",
    "SI",
    "SP",
    "BI",
    "PICT_TYPE_I",
    "PICT_TYPE_P",
    "PICT_TYPE_B",
    "PICT_TYPE_S",
    "PICT_TYPE_SI",
    "PICT_TYPE_SP",
    "PICT_TYPE_BI",
 
    "interlace_type",    ///< the frame interlace type
    "PROGRESSIVE",
    "TOPFIRST",
    "BOTTOMFIRST",
 
    "consumed_samples_n",///< number of samples consumed by the filter (only audio)
    "samples_n",         ///< number of samples in the current frame (only audio)
    "sample_rate",       ///< sample rate (only audio)
 
    "n",                 ///< frame number (starting from zero)
    "selected_n",        ///< selected frame number (starting from zero)
    "prev_selected_n",   ///< number of the last selected frame
 
    "key",               ///< tell if the frame is a key frame
 
    "scene",
 
    "concatdec_select",  ///< frame is within the interval set by the concat demuxer
 
    "ih",                ///< ih: Represents the height of the input video frame.
    "iw",                ///< iw: Represents the width of the input video frame.
 
    "view",
 
    NULL
};
 
enum var_name {
    VAR_TB,
 
    VAR_PTS,
    VAR_START_PTS,
    VAR_PREV_PTS,
    VAR_PREV_SELECTED_PTS,
 
    VAR_T,
    VAR_START_T,
    VAR_PREV_T,
    VAR_PREV_SELECTED_T,
 
    VAR_PICT_TYPE,
    VAR_I,
    VAR_P,
    VAR_B,
    VAR_S,
    VAR_SI,
    VAR_SP,
    VAR_BI,
    VAR_PICT_TYPE_I,
    VAR_PICT_TYPE_P,
    VAR_PICT_TYPE_B,
    VAR_PICT_TYPE_S,
    VAR_PICT_TYPE_SI,
    VAR_PICT_TYPE_SP,
    VAR_PICT_TYPE_BI,
 
    VAR_INTERLACE_TYPE,
    VAR_INTERLACE_TYPE_P,
    VAR_INTERLACE_TYPE_T,
    VAR_INTERLACE_TYPE_B,
 
    VAR_CONSUMED_SAMPLES_N,
    VAR_SAMPLES_N,
    VAR_SAMPLE_RATE,
 
    VAR_N,
    VAR_SELECTED_N,
    VAR_PREV_SELECTED_N,
 
    VAR_KEY,
 
    VAR_SCENE,
 
    VAR_CONCATDEC_SELECT,
 
    VAR_IH,
    VAR_IW,
 
    VAR_VIEW,
 
    VAR_VARS_NB
};
 
typedef struct SelectContext {
    const AVClass *class;
    char *expr_str;
    AVExpr *expr;
    double var_values[VAR_VARS_NB];
    int bitdepth;
    int nb_planes;
    ptrdiff_t width[4];
    ptrdiff_t height[4];
    int do_scene_detect;            ///< 1 if the expression requires scene detection variables, 0 otherwise
    ff_scene_sad_fn sad;            ///< Sum of the absolute difference function (scene detect only)
    double prev_mafd;               ///< previous MAFD                           (scene detect only)
    AVFrame *prev_picref;           ///< previous frame                          (scene detect only)
    double select;
    int select_out;                 ///< mark the selected output pad index
    int nb_outputs;
} SelectContext;
 
#define OFFSET(x) offsetof(SelectContext, x)
#define DEFINE_OPTIONS(filt_name, FLAGS)                            \
static const AVOption filt_name##_options[] = {                     \
    { "expr", "set an expression to use for selecting frames", OFFSET(expr_str), AV_OPT_TYPE_STRING, { .str = "1" }, .flags=FLAGS }, \
    { "e",    "set an expression to use for selecting frames", OFFSET(expr_str), AV_OPT_TYPE_STRING, { .str = "1" }, .flags=FLAGS }, \
    { "outputs", "set the number of outputs", OFFSET(nb_outputs), AV_OPT_TYPE_INT, {.i64 = 1}, 1, INT_MAX, .flags=FLAGS }, \
    { "n",       "set the number of outputs", OFFSET(nb_outputs), AV_OPT_TYPE_INT, {.i64 = 1}, 1, INT_MAX, .flags=FLAGS }, \
    { NULL }                                                            \
}
 
static av_cold int init(AVFilterContext *ctx)
{
    SelectContext *select = ctx->priv;
    int i, ret;
 
    if ((ret = av_expr_parse(&select->expr, select->expr_str,
                             var_names, NULL, NULL, NULL, NULL, 0, ctx)) < 0) {
        av_log(ctx, AV_LOG_ERROR, "Error while parsing expression '%s'\n",
               select->expr_str);
        return ret;
    }
    select->do_scene_detect = !!strstr(select->expr_str, "scene");
 
    for (i = 0; i < select->nb_outputs; i++) {
        AVFilterPad pad = { 0 };
 
        pad.name = av_asprintf("output%d", i);
        if (!pad.name)
            return AVERROR(ENOMEM);
        pad.type = ctx->filter->inputs[0].type;
        if ((ret = ff_append_outpad_free_name(ctx, &pad)) < 0)
            return ret;
    }
 
    return 0;
}
 
#define INTERLACE_TYPE_P 0
#define INTERLACE_TYPE_T 1
#define INTERLACE_TYPE_B 2
 
static int config_input(AVFilterLink *inlink)
{
    SelectContext *select = inlink->dst->priv;
    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
    int is_yuv = !(desc->flags & AV_PIX_FMT_FLAG_RGB) &&
                 (desc->flags & AV_PIX_FMT_FLAG_PLANAR) &&
                 desc->nb_components >= 3;
 
    select->bitdepth = desc->comp[0].depth;
    select->nb_planes = is_yuv ? 1 : av_pix_fmt_count_planes(inlink->format);
 
    for (int plane = 0; plane < select->nb_planes; plane++) {
        ptrdiff_t line_size = av_image_get_linesize(inlink->format, inlink->w, plane);
        int vsub = desc->log2_chroma_h;
 
        select->width[plane] = line_size >> (select->bitdepth > 8);
        select->height[plane] = plane == 1 || plane == 2 ?  AV_CEIL_RSHIFT(inlink->h, vsub) : inlink->h;
    }
 
    select->var_values[VAR_N]          = 0.0;
    select->var_values[VAR_SELECTED_N] = 0.0;
 
    select->var_values[VAR_TB] = av_q2d(inlink->time_base);
 
    select->var_values[VAR_PREV_SELECTED_N]   = NAN;
    select->var_values[VAR_PREV_PTS]          = NAN;
    select->var_values[VAR_PREV_SELECTED_PTS] = NAN;
    select->var_values[VAR_PREV_SELECTED_T]   = NAN;
    select->var_values[VAR_PREV_T]            = NAN;
    select->var_values[VAR_START_PTS]         = NAN;
    select->var_values[VAR_START_T]           = NAN;
 
    select->var_values[VAR_I]  = AV_PICTURE_TYPE_I;
    select->var_values[VAR_P]  = AV_PICTURE_TYPE_P;
    select->var_values[VAR_B]  = AV_PICTURE_TYPE_B;
    select->var_values[VAR_SI] = AV_PICTURE_TYPE_SI;
    select->var_values[VAR_SP] = AV_PICTURE_TYPE_SP;
    select->var_values[VAR_BI] = AV_PICTURE_TYPE_BI;
    select->var_values[VAR_PICT_TYPE_I]  = AV_PICTURE_TYPE_I;
    select->var_values[VAR_PICT_TYPE_P]  = AV_PICTURE_TYPE_P;
    select->var_values[VAR_PICT_TYPE_B]  = AV_PICTURE_TYPE_B;
    select->var_values[VAR_PICT_TYPE_SI] = AV_PICTURE_TYPE_SI;
    select->var_values[VAR_PICT_TYPE_SP] = AV_PICTURE_TYPE_SP;
    select->var_values[VAR_PICT_TYPE_BI] = AV_PICTURE_TYPE_BI;
 
    select->var_values[VAR_INTERLACE_TYPE_P] = INTERLACE_TYPE_P;
    select->var_values[VAR_INTERLACE_TYPE_T] = INTERLACE_TYPE_T;
    select->var_values[VAR_INTERLACE_TYPE_B] = INTERLACE_TYPE_B;
 
    select->var_values[VAR_PICT_TYPE]         = NAN;
    select->var_values[VAR_INTERLACE_TYPE]    = NAN;
    select->var_values[VAR_SCENE]             = NAN;
    select->var_values[VAR_CONSUMED_SAMPLES_N] = NAN;
    select->var_values[VAR_SAMPLES_N]          = NAN;
 
    select->var_values[VAR_IH] = NAN;
    select->var_values[VAR_IW] = NAN;
 
    select->var_values[VAR_SAMPLE_RATE] =
        inlink->type == AVMEDIA_TYPE_AUDIO ? inlink->sample_rate : NAN;
 
    if (CONFIG_SELECT_FILTER && select->do_scene_detect) {
        select->sad = ff_scene_sad_get_fn(select->bitdepth);
        if (!select->sad)
            return AVERROR(EINVAL);
    }
    return 0;
}
 
static double get_scene_score(AVFilterContext *ctx, AVFrame *frame)
{
    double ret = 0;
    SelectContext *select = ctx->priv;
    AVFrame *prev_picref = select->prev_picref;
 
    if (prev_picref &&
        frame->height == prev_picref->height &&
        frame->width  == prev_picref->width) {
        uint64_t sad = 0;
        double mafd, diff;
        uint64_t count = 0;
 
        for (int plane = 0; plane < select->nb_planes; plane++) {
            uint64_t plane_sad;
            select->sad(prev_picref->data[plane], prev_picref->linesize[plane],
                    frame->data[plane], frame->linesize[plane],
                    select->width[plane], select->height[plane], &plane_sad);
            sad += plane_sad;
            count += select->width[plane] * select->height[plane];
        }
 
        mafd = (double)sad / count / (1ULL << (select->bitdepth - 8));
        diff = fabs(mafd - select->prev_mafd);
        ret  = av_clipf(FFMIN(mafd, diff) / 100., 0, 1);
        select->prev_mafd = mafd;
        av_frame_free(&prev_picref);
    }
    select->prev_picref = av_frame_clone(frame);
    return ret;
}
 
static double get_concatdec_select(AVFrame *frame, int64_t pts)
{
    AVDictionary *metadata = frame->metadata;
    AVDictionaryEntry *start_time_entry = av_dict_get(metadata, "lavf.concatdec.start_time", NULL, 0);
    AVDictionaryEntry *duration_entry = av_dict_get(metadata, "lavf.concatdec.duration", NULL, 0);
    if (start_time_entry) {
        int64_t start_time = strtoll(start_time_entry->value, NULL, 10);
        if (pts >= start_time) {
            if (duration_entry) {
              int64_t duration = strtoll(duration_entry->value, NULL, 10);
              if (pts < start_time + duration)
                  return -1;
              else
                  return 0;
            }
            return -1;
        }
        return 0;
    }
    return NAN;
}
 
static void select_frame(AVFilterContext *ctx, AVFrame *frame)
{
    SelectContext *select = ctx->priv;
    AVFilterLink *inlink = ctx->inputs[0];
    FilterLink      *inl = ff_filter_link(inlink);
    const AVFrameSideData *sd;
    double res;
 
    if (isnan(select->var_values[VAR_START_PTS]))
        select->var_values[VAR_START_PTS] = TS2D(frame->pts);
    if (isnan(select->var_values[VAR_START_T]))
        select->var_values[VAR_START_T] = TS2D(frame->pts) * av_q2d(inlink->time_base);
 
    select->var_values[VAR_N  ] = inl->frame_count_out - 1;
    select->var_values[VAR_PTS] = TS2D(frame->pts);
    select->var_values[VAR_T  ] = TS2D(frame->pts) * av_q2d(inlink->time_base);
    select->var_values[VAR_KEY] = !!(frame->flags & AV_FRAME_FLAG_KEY);
    select->var_values[VAR_CONCATDEC_SELECT] = get_concatdec_select(frame, av_rescale_q(frame->pts, inlink->time_base, AV_TIME_BASE_Q));
 
    switch (inlink->type) {
    case AVMEDIA_TYPE_AUDIO:
        select->var_values[VAR_SAMPLES_N] = frame->nb_samples;
        break;
 
    case AVMEDIA_TYPE_VIDEO:
        select->var_values[VAR_IH] = frame->height;
        select->var_values[VAR_IW] = frame->width;
 
        select->var_values[VAR_INTERLACE_TYPE] =
            !(frame->flags & AV_FRAME_FLAG_INTERLACED) ? INTERLACE_TYPE_P :
        (frame->flags & AV_FRAME_FLAG_TOP_FIELD_FIRST) ? INTERLACE_TYPE_T : INTERLACE_TYPE_B;
        select->var_values[VAR_PICT_TYPE] = frame->pict_type;
        if (select->do_scene_detect) {
            char buf[32];
            select->var_values[VAR_SCENE] = get_scene_score(ctx, frame);
            // TODO: document metadata
            snprintf(buf, sizeof(buf), "%f", select->var_values[VAR_SCENE]);
            av_dict_set(&frame->metadata, "lavfi.scene_score", buf, 0);
        }
 
        sd = av_frame_side_data_get(frame->side_data, frame->nb_side_data,
                                    AV_FRAME_DATA_VIEW_ID);
        select->var_values[VAR_VIEW] = sd ? *(int*)sd->data : NAN;
        break;
    }
 
    select->select = res = av_expr_eval(select->expr, select->var_values, NULL);
    av_log(inlink->dst, AV_LOG_DEBUG,
           "n:%f pts:%f t:%f key:%d",
           select->var_values[VAR_N],
           select->var_values[VAR_PTS],
           select->var_values[VAR_T],
           !!(frame->flags & AV_FRAME_FLAG_KEY));
 
    switch (inlink->type) {
    case AVMEDIA_TYPE_VIDEO:
        av_log(inlink->dst, AV_LOG_DEBUG, " interlace_type:%c pict_type:%c scene:%f",
               !(frame->flags & AV_FRAME_FLAG_INTERLACED)     ? 'P' :
               (frame->flags & AV_FRAME_FLAG_TOP_FIELD_FIRST) ? 'T' : 'B',
               av_get_picture_type_char(frame->pict_type),
               select->var_values[VAR_SCENE]);
        break;
    case AVMEDIA_TYPE_AUDIO:
        av_log(inlink->dst, AV_LOG_DEBUG, " samples_n:%d consumed_samples_n:%f",
               frame->nb_samples,
               select->var_values[VAR_CONSUMED_SAMPLES_N]);
        break;
    }
 
    if (res == 0) {
        select->select_out = -1; /* drop */
    } else if (isnan(res) || res < 0) {
        select->select_out = 0; /* first output */
    } else {
        select->select_out = FFMIN(ceilf(res)-1, select->nb_outputs-1); /* other outputs */
    }
 
    av_log(inlink->dst, AV_LOG_DEBUG, " -> select:%f select_out:%d\n", res, select->select_out);
 
    if (res) {
        select->var_values[VAR_PREV_SELECTED_N]   = select->var_values[VAR_N];
        select->var_values[VAR_PREV_SELECTED_PTS] = select->var_values[VAR_PTS];
        select->var_values[VAR_PREV_SELECTED_T]   = select->var_values[VAR_T];
        select->var_values[VAR_SELECTED_N] += 1.0;
        if (inlink->type == AVMEDIA_TYPE_AUDIO)
            select->var_values[VAR_CONSUMED_SAMPLES_N] += frame->nb_samples;
    }
 
    select->var_values[VAR_PREV_PTS] = select->var_values[VAR_PTS];
    select->var_values[VAR_PREV_T]   = select->var_values[VAR_T];
}
 
static int activate(AVFilterContext *ctx)
{
    SelectContext *select = ctx->priv;
    AVFilterLink *inlink = ctx->inputs[0];
    AVFrame *in;
    int nb_eofs = 0;
    int ret;
 
    for (int i = 0; i < ctx->nb_outputs; i++)
        nb_eofs += ff_outlink_get_status(ctx->outputs[i]) == AVERROR_EOF;
 
    if (nb_eofs == ctx->nb_outputs) {
        ff_inlink_set_status(inlink, AVERROR_EOF);
        return 0;
    }
 
    while ((ret = ff_inlink_consume_frame(inlink, &in))) {
        if (ret < 0)
            return ret;
        select_frame(ctx, in);
        if (select->select && !ff_outlink_get_status(ctx->outputs[select->select_out]))
            return ff_filter_frame(ctx->outputs[select->select_out], in);
        av_frame_free(&in);
    };
 
    FF_FILTER_FORWARD_STATUS_ALL(inlink, ctx);
    FF_FILTER_FORWARD_WANTED_ANY(ctx, inlink);
 
    return FFERROR_NOT_READY;
}
 
static av_cold void uninit(AVFilterContext *ctx)
{
    SelectContext *select = ctx->priv;
 
    av_expr_free(select->expr);
    select->expr = NULL;
 
    if (select->do_scene_detect) {
        av_frame_free(&select->prev_picref);
    }
}
 
#if CONFIG_ASELECT_FILTER
 
DEFINE_OPTIONS(aselect, AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM);
AVFILTER_DEFINE_CLASS(aselect);
 
static av_cold int aselect_init(AVFilterContext *ctx)
{
    SelectContext *select = ctx->priv;
    int ret;
 
    if ((ret = init(ctx)) < 0)
        return ret;
 
    if (select->do_scene_detect) {
        av_log(ctx, AV_LOG_ERROR, "Scene detection is ignored in aselect filter\n");
        return AVERROR(EINVAL);
    }
 
    return 0;
}
 
static const AVFilterPad avfilter_af_aselect_inputs[] = {
    {
        .name         = "default",
        .type         = AVMEDIA_TYPE_AUDIO,
        .config_props = config_input,
    },
};
 
const FFFilter ff_af_aselect = {
    .p.name        = "aselect",
    .p.description = NULL_IF_CONFIG_SMALL("Select audio frames to pass in output."),
    .p.priv_class  = &aselect_class,
    .p.flags       = AVFILTER_FLAG_DYNAMIC_OUTPUTS,
    .init        = aselect_init,
    .uninit      = uninit,
    .priv_size   = sizeof(SelectContext),
    FILTER_INPUTS(avfilter_af_aselect_inputs),
};
#endif /* CONFIG_ASELECT_FILTER */
 
#if CONFIG_SELECT_FILTER
 
static int query_formats(const AVFilterContext *ctx,
                         AVFilterFormatsConfig **cfg_in,
                         AVFilterFormatsConfig **cfg_out)
{
    const SelectContext *select = ctx->priv;
 
    if (select->do_scene_detect) {
        static const enum AVPixelFormat pix_fmts[] = {
            AV_PIX_FMT_RGB24, AV_PIX_FMT_BGR24, AV_PIX_FMT_RGBA,
            AV_PIX_FMT_ABGR, AV_PIX_FMT_BGRA, AV_PIX_FMT_GRAY8,
            AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUVJ420P,
            AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUVJ422P,
            AV_PIX_FMT_YUV420P10,
            AV_PIX_FMT_NONE
        };
        return ff_set_common_formats_from_list2(ctx, cfg_in, cfg_out, pix_fmts);
    }
    return 0;
}
 
DEFINE_OPTIONS(select, AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM);
AVFILTER_DEFINE_CLASS(select);
 
static av_cold int select_init(AVFilterContext *ctx)
{
    int ret;
 
    if ((ret = init(ctx)) < 0)
        return ret;
 
    return 0;
}
 
static const AVFilterPad avfilter_vf_select_inputs[] = {
    {
        .name         = "default",
        .type         = AVMEDIA_TYPE_VIDEO,
        .config_props = config_input,
    },
};
 
const FFFilter ff_vf_select = {
    .p.name        = "select",
    .p.description = NULL_IF_CONFIG_SMALL("Select video frames to pass in output."),
    .p.priv_class  = &select_class,
    .p.flags       = AVFILTER_FLAG_DYNAMIC_OUTPUTS | AVFILTER_FLAG_METADATA_ONLY,
    .init          = select_init,
    .uninit        = uninit,
    .activate      = activate,
    .priv_size     = sizeof(SelectContext),
    FILTER_INPUTS(avfilter_vf_select_inputs),
    FILTER_QUERY_FUNC2(query_formats),
};
#endif /* CONFIG_SELECT_FILTER */