FFmpeg libavfilter使用示例-filter au

目录

1. 参考
2. 示例说明
3. 示例代码

1. 参考

• [1] ffmpeg.org/libavfilter documentation
• [2] ffmpeg.org/Filters Documentation
• [3] FFmpeg/doc/examples/filter_audio.c
• [4] FFmpeg libavfilter使用示例-filtering video

2. 示例说明

FFmpeg中的libavfilter提供了一个通用的音视频filter框架。使用avfilter可以对音视频数据做一些效果处理如去色调、模糊、水平翻转、裁剪、加方框、叠加文字等功能。

本示例将生成一个正弦的音频PCM数据，然后把PCM数据经过如下filterchain的处理。把输出的每一帧PCM数据的MD5值打印出来。示例来源于[3]。

(input) -> abuffer -> volume -> aformat -> abuffersink -> (output)

程序的流程图如下所示。

FFmpeg_filter_audio.png

关键函数说明：

• avfilter_graph_alloc_filter: 在filtergraph中创建一个filter实例。
• avfilter_init_str：使用提供的字符串参数初始化一个filter。
• avfilter_init_dict：使用提供的AVDictionary初始化一个filter。
• avfilter_link：把两个filter连接在一起。

3. 示例代码

以下的代码来源于[3]。

/**
 * @file
 * libavfilter API usage example.
 *
 * @example filter_audio.c
 * This example will generate a sine wave audio,
 * pass it through a simple filter chain, and then compute the MD5 checksum of
 * the output data.
 *
 * The filter chain it uses is:
 * (input) -> abuffer -> volume -> aformat -> abuffersink -> (output)
 *
 * abuffer: This provides the endpoint where you can feed the decoded samples.
 * volume: In this example we hardcode it to 0.90.
 * aformat: This converts the samples to the samplefreq, channel layout,
 *          and sample format required by the audio device.
 * abuffersink: This provides the endpoint where you can read the samples after
 *              they have passed through the filter chain.
 */

#include <inttypes.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>

#include "libavutil/channel_layout.h"
#include "libavutil/md5.h"
#include "libavutil/mem.h"
#include "libavutil/opt.h"
#include "libavutil/samplefmt.h"

#include "libavfilter/avfilter.h"
#include "libavfilter/buffersink.h"
#include "libavfilter/buffersrc.h"

#define INPUT_SAMPLERATE     48000
#define INPUT_FORMAT         AV_SAMPLE_FMT_FLTP
#define INPUT_CHANNEL_LAYOUT AV_CH_LAYOUT_5POINT0

#define VOLUME_VAL 0.90

static int init_filter_graph(AVFilterGraph **graph, AVFilterContext **src,
                             AVFilterContext **sink)
{
    AVFilterGraph *filter_graph;
    AVFilterContext *abuffer_ctx;
    const AVFilter  *abuffer;
    AVFilterContext *volume_ctx;
    const AVFilter  *volume;
    AVFilterContext *aformat_ctx;
    const AVFilter  *aformat;
    AVFilterContext *abuffersink_ctx;
    const AVFilter  *abuffersink;

    AVDictionary *options_dict = NULL;
    uint8_t options_str[1024];
    uint8_t ch_layout[64];

    int err;

    /* Create a new filtergraph, which will contain all the filters. */
    filter_graph = avfilter_graph_alloc();
    if (!filter_graph) {
        fprintf(stderr, "Unable to create filter graph.\n");
        return AVERROR(ENOMEM);
    }

    /* Create the abuffer filter;
     * it will be used for feeding the data into the graph. */
    abuffer = avfilter_get_by_name("abuffer");
    if (!abuffer) {
        fprintf(stderr, "Could not find the abuffer filter.\n");
        return AVERROR_FILTER_NOT_FOUND;
    }

    abuffer_ctx = avfilter_graph_alloc_filter(filter_graph, abuffer, "src");
    if (!abuffer_ctx) {
        fprintf(stderr, "Could not allocate the abuffer instance.\n");
        return AVERROR(ENOMEM);
    }

    /* Set the filter options through the AVOptions API. */
    av_get_channel_layout_string(ch_layout, sizeof(ch_layout), 0, INPUT_CHANNEL_LAYOUT);
    av_opt_set    (abuffer_ctx, "channel_layout", ch_layout,                            AV_OPT_SEARCH_CHILDREN);
    av_opt_set    (abuffer_ctx, "sample_fmt",     av_get_sample_fmt_name(INPUT_FORMAT), AV_OPT_SEARCH_CHILDREN);
    av_opt_set_q  (abuffer_ctx, "time_base",      (AVRational){ 1, INPUT_SAMPLERATE },  AV_OPT_SEARCH_CHILDREN);
    av_opt_set_int(abuffer_ctx, "sample_rate",    INPUT_SAMPLERATE,                     AV_OPT_SEARCH_CHILDREN);

    /* Now initialize the filter; we pass NULL options, since we have already
     * set all the options above. */
    err = avfilter_init_str(abuffer_ctx, NULL);
    if (err < 0) {
        fprintf(stderr, "Could not initialize the abuffer filter.\n");
        return err;
    }

    /* Create volume filter. */
    volume = avfilter_get_by_name("volume");
    if (!volume) {
        fprintf(stderr, "Could not find the volume filter.\n");
        return AVERROR_FILTER_NOT_FOUND;
    }

    volume_ctx = avfilter_graph_alloc_filter(filter_graph, volume, "volume");
    if (!volume_ctx) {
        fprintf(stderr, "Could not allocate the volume instance.\n");
        return AVERROR(ENOMEM);
    }

    /* A different way of passing the options is as key/value pairs in a
     * dictionary. */
    av_dict_set(&options_dict, "volume", AV_STRINGIFY(VOLUME_VAL), 0);
    err = avfilter_init_dict(volume_ctx, &options_dict);
    av_dict_free(&options_dict);
    if (err < 0) {
        fprintf(stderr, "Could not initialize the volume filter.\n");
        return err;
    }

    /* Create the aformat filter;
     * it ensures that the output is of the format we want. */
    aformat = avfilter_get_by_name("aformat");
    if (!aformat) {
        fprintf(stderr, "Could not find the aformat filter.\n");
        return AVERROR_FILTER_NOT_FOUND;
    }

    aformat_ctx = avfilter_graph_alloc_filter(filter_graph, aformat, "aformat");
    if (!aformat_ctx) {
        fprintf(stderr, "Could not allocate the aformat instance.\n");
        return AVERROR(ENOMEM);
    }

    /* A third way of passing the options is in a string of the form
     * key1=value1:key2=value2.... */
    snprintf(options_str, sizeof(options_str),
             "sample_fmts=%s:sample_rates=%d:channel_layouts=0x%"PRIx64,
             av_get_sample_fmt_name(AV_SAMPLE_FMT_S16), 44100,
             (uint64_t)AV_CH_LAYOUT_STEREO);
    err = avfilter_init_str(aformat_ctx, options_str);
    if (err < 0) {
        av_log(NULL, AV_LOG_ERROR, "Could not initialize the aformat filter.\n");
        return err;
    }

    /* Finally create the abuffersink filter;
     * it will be used to get the filtered data out of the graph. */
    abuffersink = avfilter_get_by_name("abuffersink");
    if (!abuffersink) {
        fprintf(stderr, "Could not find the abuffersink filter.\n");
        return AVERROR_FILTER_NOT_FOUND;
    }

    abuffersink_ctx = avfilter_graph_alloc_filter(filter_graph, abuffersink, "sink");
    if (!abuffersink_ctx) {
        fprintf(stderr, "Could not allocate the abuffersink instance.\n");
        return AVERROR(ENOMEM);
    }

    /* This filter takes no options. */
    err = avfilter_init_str(abuffersink_ctx, NULL);
    if (err < 0) {
        fprintf(stderr, "Could not initialize the abuffersink instance.\n");
        return err;
    }

    /* Connect the filters;
     * in this simple case the filters just form a linear chain. */
    err = avfilter_link(abuffer_ctx, 0, volume_ctx, 0);
    if (err >= 0)
        err = avfilter_link(volume_ctx, 0, aformat_ctx, 0);
    if (err >= 0)
        err = avfilter_link(aformat_ctx, 0, abuffersink_ctx, 0);
    if (err < 0) {
        fprintf(stderr, "Error connecting filters\n");
        return err;
    }

    /* Configure the graph. */
    err = avfilter_graph_config(filter_graph, NULL);
    if (err < 0) {
        av_log(NULL, AV_LOG_ERROR, "Error configuring the filter graph\n");
        return err;
    }

    *graph = filter_graph;
    *src   = abuffer_ctx;
    *sink  = abuffersink_ctx;

    return 0;
}

/* Do something useful with the filtered data: this simple
 * example just prints the MD5 checksum of each plane to stdout. */
static int process_output(struct AVMD5 *md5, AVFrame *frame)
{
    int planar     = av_sample_fmt_is_planar(frame->format);
    int channels   = av_get_channel_layout_nb_channels(frame->channel_layout);
    int planes     = planar ? channels : 1;
    int bps        = av_get_bytes_per_sample(frame->format);
    int plane_size = bps * frame->nb_samples * (planar ? 1 : channels);
    int i, j;

    for (i = 0; i < planes; i++) {
        uint8_t checksum[16];

        av_md5_init(md5);
        av_md5_sum(checksum, frame->extended_data[i], plane_size);

        fprintf(stdout, "plane %d: 0x", i);
        for (j = 0; j < sizeof(checksum); j++)
            fprintf(stdout, "%02X", checksum[j]);
        fprintf(stdout, "\n");
    }
    fprintf(stdout, "\n");

    return 0;
}

/* Construct a frame of audio data to be filtered;
 * this simple example just synthesizes a sine wave. */
static int get_input(AVFrame *frame, int frame_num)
{
    int err, i, j;

#define FRAME_SIZE 1024

    /* Set up the frame properties and allocate the buffer for the data. */
    frame->sample_rate    = INPUT_SAMPLERATE;
    frame->format         = INPUT_FORMAT;
    frame->channel_layout = INPUT_CHANNEL_LAYOUT;
    frame->nb_samples     = FRAME_SIZE;
    frame->pts            = frame_num * FRAME_SIZE;

    err = av_frame_get_buffer(frame, 0);
    if (err < 0)
        return err;

    /* Fill the data for each channel. */
    for (i = 0; i < 5; i++) {
        float *data = (float*)frame->extended_data[i];

        for (j = 0; j < frame->nb_samples; j++)
            data[j] = sin(2 * M_PI * (frame_num + j) * (i + 1) / FRAME_SIZE);
    }

    return 0;
}

int main(int argc, char *argv[])
{
    struct AVMD5 *md5;
    AVFilterGraph *graph;
    AVFilterContext *src, *sink;
    AVFrame *frame;
    uint8_t errstr[1024];
    float duration;
    int err, nb_frames, i;

    if (argc < 2) {
        fprintf(stderr, "Usage: %s <duration>\n", argv[0]);
        return 1;
    }

    duration  = atof(argv[1]);
    nb_frames = duration * INPUT_SAMPLERATE / FRAME_SIZE;
    if (nb_frames <= 0) {
        fprintf(stderr, "Invalid duration: %s\n", argv[1]);
        return 1;
    }

    /* Allocate the frame we will be using to store the data. */
    frame  = av_frame_alloc();
    if (!frame) {
        fprintf(stderr, "Error allocating the frame\n");
        return 1;
    }

    md5 = av_md5_alloc();
    if (!md5) {
        fprintf(stderr, "Error allocating the MD5 context\n");
        return 1;
    }

    /* Set up the filtergraph. */
    err = init_filter_graph(&graph, &src, &sink);
    if (err < 0) {
        fprintf(stderr, "Unable to init filter graph:");
        goto fail;
    }

    /* the main filtering loop */
    for (i = 0; i < nb_frames; i++) {
        /* get an input frame to be filtered */
        err = get_input(frame, i);
        if (err < 0) {
            fprintf(stderr, "Error generating input frame:");
            goto fail;
        }

        /* Send the frame to the input of the filtergraph. */
        err = av_buffersrc_add_frame(src, frame);
        if (err < 0) {
            av_frame_unref(frame);
            fprintf(stderr, "Error submitting the frame to the filtergraph:");
            goto fail;
        }

        /* Get all the filtered output that is available. */
        while ((err = av_buffersink_get_frame(sink, frame)) >= 0) {
            /* now do something with our filtered frame */
            err = process_output(md5, frame);
            if (err < 0) {
                fprintf(stderr, "Error processing the filtered frame:");
                goto fail;
            }
            av_frame_unref(frame);
        }

        if (err == AVERROR(EAGAIN)) {
            /* Need to feed more frames in. */
            continue;
        } else if (err == AVERROR_EOF) {
            /* Nothing more to do, finish. */
            break;
        } else if (err < 0) {
            /* An error occurred. */
            fprintf(stderr, "Error filtering the data:");
            goto fail;
        }
    }

    avfilter_graph_free(&graph);
    av_frame_free(&frame);
    av_freep(&md5);

    return 0;

fail:
    av_strerror(err, errstr, sizeof(errstr));
    fprintf(stderr, "%s\n", errstr);
    return 1;
}