#include <vector>
#include "bmusb/bmusb.h"
-#include "ffmpeg_raii.h"
+#include "shared/ffmpeg_raii.h"
#include "ffmpeg_util.h"
#include "flags.h"
#include "image_input.h"
#include "ref_counted_frame.h"
-#include "timebase.h"
+#include "shared/timebase.h"
#define FRAME_SIZE (8 << 20) // 8 MB.
return av_get_pix_fmt(best_format);
}
-YCbCrFormat decode_ycbcr_format(const AVPixFmtDescriptor *desc, const AVFrame *frame)
+YCbCrFormat decode_ycbcr_format(const AVPixFmtDescriptor *desc, const AVFrame *frame, bool is_mjpeg)
{
YCbCrFormat format;
- AVColorSpace colorspace = av_frame_get_colorspace(frame);
+ AVColorSpace colorspace = frame->colorspace;
switch (colorspace) {
case AVCOL_SPC_BT709:
format.luma_coefficients = YCBCR_REC_709;
break;
}
+ if (is_mjpeg && !format.full_range) {
+ // Limited-range MJPEG is only detected by FFmpeg whenever a special
+ // JPEG comment is set, which means that in practice, the stream is
+ // almost certainly generated by Futatabi. Override FFmpeg's forced
+ // MJPEG defaults (it disregards the values set in the mux) with what
+ // Futatabi sets.
+ format.luma_coefficients = YCBCR_REC_709;
+ format.cb_x_position = 0.0;
+ format.cb_y_position = 0.5;
+ }
+
format.cr_x_position = format.cb_x_position;
format.cr_y_position = format.cb_y_position;
return format;
if (has_dequeue_callbacks) {
dequeue_cleanup_callback();
}
- avresample_free(&resampler);
+ swr_free(&resampler);
}
void FFmpegCapture::configure_card()
last_modified = buf.st_mtim;
}
- AVDictionary *opts = nullptr;
- av_dict_set(&opts, "fflags", "nobuffer", 0);
-
- auto format_ctx = avformat_open_input_unique(pathname.c_str(), nullptr, &opts, AVIOInterruptCB{ &FFmpegCapture::interrupt_cb_thunk, this });
+ auto format_ctx = avformat_open_input_unique(pathname.c_str(), nullptr, nullptr, AVIOInterruptCB{ &FFmpegCapture::interrupt_cb_thunk, this });
if (format_ctx == nullptr) {
fprintf(stderr, "%s: Error opening file\n", pathname.c_str());
return false;
}
int audio_stream_index = find_stream_index(format_ctx.get(), AVMEDIA_TYPE_AUDIO);
+ int subtitle_stream_index = find_stream_index(format_ctx.get(), AVMEDIA_TYPE_SUBTITLE);
+ has_last_subtitle = false;
// Open video decoder.
const AVCodecParameters *video_codecpar = format_ctx->streams[video_stream_index]->codecpar;
unique_ptr<AVCodecContext, decltype(avcodec_close)*> video_codec_ctx_cleanup(
video_codec_ctx.get(), avcodec_close);
+ // Used in decode_ycbcr_format().
+ is_mjpeg = video_codecpar->codec_id == AV_CODEC_ID_MJPEG;
+
// Open audio decoder, if we have audio.
AVCodecContextWithDeleter audio_codec_ctx;
if (audio_stream_index != -1) {
int64_t audio_pts;
bool error;
AVFrameWithDeleter frame = decode_frame(format_ctx.get(), video_codec_ctx.get(), audio_codec_ctx.get(),
- pathname, video_stream_index, audio_stream_index, audio_frame.get(), &audio_format, &audio_pts, &error);
+ pathname, video_stream_index, audio_stream_index, subtitle_stream_index, audio_frame.get(), &audio_format, &audio_pts, &error);
if (error) {
return false;
}
} // namespace
AVFrameWithDeleter FFmpegCapture::decode_frame(AVFormatContext *format_ctx, AVCodecContext *video_codec_ctx, AVCodecContext *audio_codec_ctx,
- const std::string &pathname, int video_stream_index, int audio_stream_index,
+ const std::string &pathname, int video_stream_index, int audio_stream_index, int subtitle_stream_index,
FrameAllocator::Frame *audio_frame, AudioFormat *audio_format, int64_t *audio_pts, bool *error)
{
*error = false;
*error = true;
return AVFrameWithDeleter(nullptr);
}
+ } else if (pkt.stream_index == subtitle_stream_index) {
+ last_subtitle = string(reinterpret_cast<const char *>(pkt.data), pkt.size);
+ has_last_subtitle = true;
}
} else {
eof = true; // Or error, but ignore that for the time being.
audio_avframe->format != last_src_format ||
dst_format != last_dst_format ||
channel_layout != last_channel_layout ||
- av_frame_get_sample_rate(audio_avframe) != last_sample_rate) {
- avresample_free(&resampler);
- resampler = avresample_alloc_context();
+ audio_avframe->sample_rate != last_sample_rate) {
+ swr_free(&resampler);
+ resampler = swr_alloc_set_opts(nullptr,
+ /*out_ch_layout=*/AV_CH_LAYOUT_STEREO_DOWNMIX,
+ /*out_sample_fmt=*/dst_format,
+ /*out_sample_rate=*/OUTPUT_FREQUENCY,
+ /*in_ch_layout=*/channel_layout,
+ /*in_sample_fmt=*/AVSampleFormat(audio_avframe->format),
+ /*in_sample_rate=*/audio_avframe->sample_rate,
+ /*log_offset=*/0,
+ /*log_ctx=*/nullptr);
+
if (resampler == nullptr) {
fprintf(stderr, "Allocating resampler failed.\n");
exit(1);
}
- av_opt_set_int(resampler, "in_channel_layout", channel_layout, 0);
- av_opt_set_int(resampler, "out_channel_layout", AV_CH_LAYOUT_STEREO_DOWNMIX, 0);
- av_opt_set_int(resampler, "in_sample_rate", av_frame_get_sample_rate(audio_avframe), 0);
- av_opt_set_int(resampler, "out_sample_rate", OUTPUT_FREQUENCY, 0);
- av_opt_set_int(resampler, "in_sample_fmt", audio_avframe->format, 0);
- av_opt_set_int(resampler, "out_sample_fmt", dst_format, 0);
-
- if (avresample_open(resampler) < 0) {
+ if (swr_init(resampler) < 0) {
fprintf(stderr, "Could not open resample context.\n");
exit(1);
}
last_src_format = AVSampleFormat(audio_avframe->format);
last_dst_format = dst_format;
last_channel_layout = channel_layout;
- last_sample_rate = av_frame_get_sample_rate(audio_avframe);
+ last_sample_rate = audio_avframe->sample_rate;
}
size_t bytes_per_sample = (audio_format->bits_per_sample / 8) * 2;
size_t num_samples_room = (audio_frame->size - audio_frame->len) / bytes_per_sample;
uint8_t *data = audio_frame->data + audio_frame->len;
- int out_samples = avresample_convert(resampler, &data, 0, num_samples_room,
- const_cast<uint8_t **>(audio_avframe->data), audio_avframe->linesize[0], audio_avframe->nb_samples);
+ int out_samples = swr_convert(resampler, &data, num_samples_room,
+ const_cast<const uint8_t **>(audio_avframe->data), audio_avframe->nb_samples);
if (out_samples < 0) {
fprintf(stderr, "Audio conversion failed.\n");
exit(1);
video_format.stride = width;
}
video_format.frame_rate_nom = video_timebase.den;
- video_format.frame_rate_den = av_frame_get_pkt_duration(frame) * video_timebase.num;
+ video_format.frame_rate_den = frame->pkt_duration * video_timebase.num;
if (video_format.frame_rate_nom == 0 || video_format.frame_rate_den == 0) {
// Invalid frame rate.
video_format.frame_rate_nom = 60;
video_frame->len = (width * 2) * height;
const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(sws_dst_format);
- current_frame_ycbcr_format = decode_ycbcr_format(desc, frame);
+ current_frame_ycbcr_format = decode_ycbcr_format(desc, frame, is_mjpeg);
} else {
assert(pixel_format == bmusb::PixelFormat_8BitYCbCrPlanar);
const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(sws_dst_format);
video_frame->len = width * height + 2 * chroma_width * chroma_height;
- current_frame_ycbcr_format = decode_ycbcr_format(desc, frame);
+ current_frame_ycbcr_format = decode_ycbcr_format(desc, frame, is_mjpeg);
}
sws_scale(sws_ctx.get(), frame->data, frame->linesize, 0, frame->height, pic_data, linesizes);
projects / nageru / blobdiff