C ++-在ffmpeg中应用过滤器

Question

I'm trying to deinterlace a frame using ffmpeg (latest release). Related with this question, I can get the filter I want using this sentence:

AVFilter *filter = avfilter_get_by_name("yadif");

After that, I open the filter context as:

AVFilterContext *filter_ctx;
avfilter_open(&filter_ctx, filter, NULL);

My first question is about this function. Visual Studio warns me about avfilter_open is deprecated. Which is the alternative?

After that, I do:

avfilter_init_str(filter_ctx, "yadif=1:-1");

And always fails. I've tried " 1:-1 " instead " yadif=1:-1 ", but always fails too, what parameter I should use?

EDIT: A value of " 1 " or " 2 ", for example, it works. Debuging it, I found that with one of this values, the function uses mode=1 or mode=2 . The explanation of those values is in this link .

Then, I have a AVFrame *frame that is the frame I want to deinterlace. When the last sentence work, I'll have the filter and his context init. How do I apply this filter to my frame?

Thanks for your help.

Answer 1

I undertsnad your question is over a year now but recently I had to work with interlaced DVB-TS streams so I might be able to help anybody else coming across this subject.

These snippets are from a finished player I've written

Initialise the filter graph:

void VideoManager::init_filter_graph(AVFrame *frame) {
    if (filter_initialised) return;

    int result;

    AVFilter *buffer_src   = avfilter_get_by_name("buffer");
    AVFilter *buffer_sink  = avfilter_get_by_name("buffersink");
    AVFilterInOut *inputs  = avfilter_inout_alloc();
    AVFilterInOut *outputs = avfilter_inout_alloc();

    AVCodecContext *ctx = ffmpeg.vid_stream.context;
    char args[512];

    int frame_fix = 0; // fix bad width on some streams
    if (frame->width < 704) frame_fix = 2;
    else if (frame->width > 704) frame_fix = -16;

    snprintf(args, sizeof(args),
         "video_size=%dx%d:pix_fmt=%d:time_base=%d/%d:pixel_aspect=%d/%d",
         frame->width + frame_fix,
         frame->height,
         frame->format,// ctx->pix_fmt,
         ctx->time_base.num,
         ctx->time_base.den,
         ctx->sample_aspect_ratio.num,
         ctx->sample_aspect_ratio.den);

    const char *description = "yadif=1:-1:0";

    LOGD("Filter: %s - Settings: %s", description, args);

    filter_graph = avfilter_graph_alloc();
    result = avfilter_graph_create_filter(&filter_src_ctx, buffer_src, "in", args, NULL, filter_graph);
    if (result < 0) {
        LOGI("Filter graph - Unable to create buffer source");
        return;
    }

    AVBufferSinkParams *params = av_buffersink_params_alloc();
    enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_GRAY8, AV_PIX_FMT_NONE };

    params->pixel_fmts = pix_fmts;
    result = avfilter_graph_create_filter(&filter_sink_ctx, buffer_sink, "out", NULL, params, filter_graph);
    av_free(params);
    if (result < 0) {
        LOGI("Filter graph - Unable to create buffer sink");
        return;
    }

    inputs->name        = av_strdup("out");
    inputs->filter_ctx  = filter_sink_ctx;
    inputs->pad_idx     = 0;
    inputs->next        = NULL;

    outputs->name       = av_strdup("in");
    outputs->filter_ctx = filter_src_ctx;
    outputs->pad_idx    = 0;
    outputs->next       = NULL;

    result = avfilter_graph_parse_ptr(filter_graph, description, &inputs, &outputs, NULL);
    if (result < 0) LOGI("avfilter_graph_parse_ptr ERROR");

    result = avfilter_graph_config(filter_graph, NULL);
    if (result < 0) LOGI("avfilter_graph_config ERROR");

    filter_initialised = true;
}

When processing the video packets from the stream, check if it is an interlaced frame and send the frame off to the filter. The filter will then return the de-interlaced frames back to you.

void FFMPEG::process_video_packet(AVPacket *pkt) {
    int got;
    AVFrame *frame = vid_stream.frame;
    avcodec_decode_video2(vid_stream.context, frame, &got, pkt);

    if (got) {
        if (!frame->interlaced_frame) {     // not interlaced
            Video.add_av_frame(frame, 0);
        } else {
            if (!Video.filter_initialised) {
                Video.init_filter_graph(frame);
            }

            av_buffersrc_add_frame_flags(Video.filter_src_ctx, frame, AV_BUFFERSRC_FLAG_KEEP_REF);
            int c = 0;

            while (true) {
                AVFrame *filter_frame = ffmpeg.vid_stream.filter_frame;

                int result = av_buffersink_get_frame(Video.filter_sink_ctx, filter_frame);

                if (result == AVERROR(EAGAIN) || result == AVERROR(AVERROR_EOF)) break;
                if (result < 0) return;

                Video.add_av_frame(filter_frame, c++);
                av_frame_unref(filter_frame);
            }
        }
    }
}

Hope this helps anyone because finding information about ffmpeg is tough going.