X-Git-Url: https://main.carlh.net/gitweb/?a=blobdiff_plain;f=src%2Flib%2Fffmpeg_examiner.cc;h=d439b0eb68ba54e27b3ce779071d3096c6967177;hb=75820f6fe6d1d2de7546694c6c875c24b40478bf;hp=6ed7b06fea050cc33f04763978283025456fbb9e;hpb=ada1c702da6a1d892286a8665d9e0e93dac4c166;p=dcpomatic.git diff --git a/src/lib/ffmpeg_examiner.cc b/src/lib/ffmpeg_examiner.cc index 6ed7b06fe..d439b0eb6 100644 --- a/src/lib/ffmpeg_examiner.cc +++ b/src/lib/ffmpeg_examiner.cc @@ -1,27 +1,31 @@ /* Copyright (C) 2013-2015 Carl Hetherington - This program is free software; you can redistribute it and/or modify + This file is part of DCP-o-matic. + + DCP-o-matic is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. - This program is distributed in the hope that it will be useful, + DCP-o-matic 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 General Public License for more details. You should have received a copy of the GNU General Public License - along with this program; if not, write to the Free Software - Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + along with DCP-o-matic. If not, see . */ +#include "dcpomatic_log.h" extern "C" { #include #include #include #include +#include +#include } #include "ffmpeg_examiner.h" #include "ffmpeg_content.h" @@ -29,21 +33,30 @@ extern "C" { #include "ffmpeg_audio_stream.h" #include "ffmpeg_subtitle_stream.h" #include "util.h" -#include "safe_stringstream.h" +#include +#include #include "i18n.h" using std::string; using std::cout; using std::max; +using std::vector; using boost::shared_ptr; using boost::optional; +/* This is how many frames from the start of any video that we will examine to see if we + * can spot soft 2:3 pull-down ("telecine"). + */ +static const int PULLDOWN_CHECK_FRAMES = 16; + + /** @param job job that the examiner is operating in, or 0 */ FFmpegExaminer::FFmpegExaminer (shared_ptr c, shared_ptr job) : FFmpeg (c) , _video_length (0) , _need_video_length (false) + , _pulldown (false) { /* Find audio and subtitle streams */ @@ -59,9 +72,20 @@ FFmpegExaminer::FFmpegExaminer (shared_ptr c, shared_ptrcodec->channel_layout = av_get_default_channel_layout (s->codec->channels); } + DCPOMATIC_ASSERT (_format_context->duration != AV_NOPTS_VALUE); + DCPOMATIC_ASSERT (s->codec->codec); + DCPOMATIC_ASSERT (s->codec->codec->name); + _audio_streams.push_back ( shared_ptr ( - new FFmpegAudioStream (audio_stream_name (s), s->id, s->codec->sample_rate, s->codec->channels) + new FFmpegAudioStream ( + stream_name (s), + s->codec->codec->name, + s->id, + s->codec->sample_rate, + llrint ((double (_format_context->duration) / AV_TIME_BASE) * s->codec->sample_rate), + s->codec->channels + ) ) ); @@ -70,29 +94,30 @@ FFmpegExaminer::FFmpegExaminer (shared_ptr c, shared_ptrduration == AV_NOPTS_VALUE; - if (!_need_video_length) { - _video_length = (double (_format_context->duration) / AV_TIME_BASE) * video_frame_rate().get (); - } else if (job) { - job->sub (_("Finding length")); - job->set_progress_unknown (); + if (has_video ()) { + /* See if the header has duration information in it */ + _need_video_length = _format_context->duration == AV_NOPTS_VALUE; + if (!_need_video_length) { + _video_length = llrint ((double (_format_context->duration) / AV_TIME_BASE) * video_frame_rate().get()); + } } - if (job) { - job->sub (_("Finding subtitles")); + if (job && _need_video_length) { + job->sub (_("Finding length")); } /* Run through until we find: * - the first video. * - the first audio for each stream. - * - the subtitle periods for each stream. - * - * We have to note subtitle periods as otherwise we have no way of knowing - * where we should look for subtitles (video and audio are always present, - * so they are ok). + * - the top-field-first and repeat-first-frame values ("temporal_reference") for the first PULLDOWN_CHECK_FRAMES video frames. */ + int64_t const len = _file_group.length (); + /* A string which we build up to describe the top-field-first and repeat-first-frame values for the first few frames. + * It would be nicer to use something like vector here but we want to search the array for a pattern later, + * and a string seems a reasonably neat way to do that. + */ + string temporal_reference; while (true) { int r = av_read_frame (_format_context, &_packet); if (r < 0) { @@ -100,13 +125,17 @@ FFmpegExaminer::FFmpegExaminer (shared_ptr c, shared_ptrset_progress_unknown (); + if (len > 0) { + job->set_progress (float (_format_context->pb->pos) / len); + } else { + job->set_progress_unknown (); + } } AVCodecContext* context = _format_context->streams[_packet.stream_index]->codec; - if (_packet.stream_index == _video_stream) { - video_packet (context); + if (_video_stream && _packet.stream_index == _video_stream.get()) { + video_packet (context, temporal_reference); } bool got_all_audio = true; @@ -120,38 +149,81 @@ FFmpegExaminer::FFmpegExaminer (shared_ptr c, shared_ptruses_index (_format_context, _packet.stream_index)) { - subtitle_packet (context, _subtitle_streams[i]); - } - } + av_packet_unref (&_packet); - av_free_packet (&_packet); - - if (_first_video && got_all_audio && _subtitle_streams.empty ()) { + if (_first_video && got_all_audio && temporal_reference.size() >= PULLDOWN_CHECK_FRAMES) { /* All done */ break; } } + + if (_video_stream) { + /* This code taken from get_rotation() in ffmpeg:cmdutils.c */ + AVStream* stream = _format_context->streams[*_video_stream]; + AVDictionaryEntry* rotate_tag = av_dict_get (stream->metadata, "rotate", 0, 0); + uint8_t* displaymatrix = av_stream_get_side_data (stream, AV_PKT_DATA_DISPLAYMATRIX, 0); + _rotation = 0; + + if (rotate_tag && *rotate_tag->value && strcmp(rotate_tag->value, "0")) { + char *tail; + _rotation = av_strtod (rotate_tag->value, &tail); + if (*tail) { + _rotation = 0; + } + } + + if (displaymatrix && !_rotation) { + _rotation = - av_display_rotation_get ((int32_t*) displaymatrix); + } + + _rotation = *_rotation - 360 * floor (*_rotation / 360 + 0.9 / 360); + + DCPOMATIC_ASSERT (fabs (*_rotation - 90 * round (*_rotation / 90)) < 2); + } + + LOG_GENERAL("Temporal reference was %1", temporal_reference); + if (temporal_reference.find("T2T3B2B3T2T3B2B3") != string::npos || temporal_reference.find("B2B3T2T3B2B3T2T3") != string::npos) { + /* The magical sequence (taken from mediainfo) suggests that 2:3 pull-down is in use */ + _pulldown = true; + LOG_GENERAL_NC("Suggest that this may be 2:3 pull-down (soft telecine)"); + } + +#ifdef DCPOMATIC_VARIANT_SWAROOP + AVDictionaryEntry* e = av_dict_get (_format_context->metadata, SWAROOP_ID_TAG, 0, 0); + if (e) { + _id = e->value; + } +#endif } + +/** @param temporal_reference A string to which we should add two characters per frame; + * the first is T or B depending on whether it's top- or bottom-field first, + * ths seconds is 3 or 2 depending on whether "repeat_pict" is true or not. + */ void -FFmpegExaminer::video_packet (AVCodecContext* context) +FFmpegExaminer::video_packet (AVCodecContext* context, string& temporal_reference) { - if (_first_video && !_need_video_length) { + DCPOMATIC_ASSERT (_video_stream); + + if (_first_video && !_need_video_length && temporal_reference.size() >= PULLDOWN_CHECK_FRAMES) { return; } int frame_finished; if (avcodec_decode_video2 (context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) { if (!_first_video) { - _first_video = frame_time (_format_context->streams[_video_stream]); + _first_video = frame_time (_format_context->streams[_video_stream.get()]); } if (_need_video_length) { _video_length = frame_time ( - _format_context->streams[_video_stream] + _format_context->streams[_video_stream.get()] ).get_value_or (ContentTime ()).frames_round (video_frame_rate().get ()); } + if (temporal_reference.size() < PULLDOWN_CHECK_FRAMES) { + temporal_reference += (_frame->top_field_first ? "T" : "B"); + temporal_reference += (_frame->repeat_pict ? "3" : "2"); + } } } @@ -168,27 +240,6 @@ FFmpegExaminer::audio_packet (AVCodecContext* context, shared_ptr stream) -{ - int frame_finished; - AVSubtitle sub; - if (avcodec_decode_subtitle2 (context, &sub, &frame_finished, &_packet) >= 0 && frame_finished) { - FFmpegSubtitlePeriod const period = subtitle_period (sub); - if (sub.num_rects <= 0 && _last_subtitle_start) { - stream->add_subtitle (ContentTimePeriod (_last_subtitle_start.get (), period.from)); - _last_subtitle_start = optional (); - } else if (sub.num_rects == 1) { - if (period.to) { - stream->add_subtitle (ContentTimePeriod (period.from, period.to.get ())); - } else { - _last_subtitle_start = period.from; - } - } - avsubtitle_free (&sub); - } -} - optional FFmpegExaminer::frame_time (AVStream* s) const { @@ -205,11 +256,8 @@ FFmpegExaminer::frame_time (AVStream* s) const optional FFmpegExaminer::video_frame_rate () const { - /* This use of r_frame_rate is debateable; there's a few different - * frame rates in the format context, but this one seems to be the most - * reliable. - */ - return av_q2d (av_stream_get_r_frame_rate (_format_context->streams[_video_stream])); + DCPOMATIC_ASSERT (_video_stream); + return av_q2d(av_guess_frame_rate(_format_context, _format_context->streams[_video_stream.get()], 0)); } dcp::Size @@ -228,7 +276,8 @@ FFmpegExaminer::video_length () const optional FFmpegExaminer::sample_aspect_ratio () const { - AVRational sar = av_guess_sample_aspect_ratio (_format_context, _format_context->streams[_video_stream], 0); + DCPOMATIC_ASSERT (_video_stream); + AVRational sar = av_guess_sample_aspect_ratio (_format_context, _format_context->streams[_video_stream.get()], 0); if (sar.num == 0) { /* I assume this means that we don't know */ return optional (); @@ -236,61 +285,136 @@ FFmpegExaminer::sample_aspect_ratio () const return double (sar.num) / sar.den; } -string -FFmpegExaminer::audio_stream_name (AVStream* s) const -{ - SafeStringStream n; - - n << stream_name (s); - - if (!n.str().empty()) { - n << "; "; - } - - n << s->codec->channels << " channels"; - - return n.str (); -} - string FFmpegExaminer::subtitle_stream_name (AVStream* s) const { - SafeStringStream n; + string n = stream_name (s); - n << stream_name (s); - - if (n.str().empty()) { - n << _("unknown"); + if (n.empty()) { + n = _("unknown"); } - return n.str (); + return n; } string FFmpegExaminer::stream_name (AVStream* s) const { - SafeStringStream n; + string n; if (s->metadata) { AVDictionaryEntry const * lang = av_dict_get (s->metadata, "language", 0, 0); if (lang) { - n << lang->value; + n = lang->value; } AVDictionaryEntry const * title = av_dict_get (s->metadata, "title", 0, 0); if (title) { - if (!n.str().empty()) { - n << " "; + if (!n.empty()) { + n += " "; } - n << title->value; + n += title->value; } } - return n.str (); + return n; } -int +optional FFmpegExaminer::bits_per_pixel () const { - return av_get_bits_per_pixel (av_pix_fmt_desc_get (video_codec_context()->pix_fmt)); + if (video_codec_context()->pix_fmt == -1) { + return optional(); + } + + AVPixFmtDescriptor const * d = av_pix_fmt_desc_get (video_codec_context()->pix_fmt); + DCPOMATIC_ASSERT (d); + return av_get_bits_per_pixel (d); +} + +bool +FFmpegExaminer::yuv () const +{ + switch (video_codec_context()->pix_fmt) { + case AV_PIX_FMT_YUV420P: + case AV_PIX_FMT_YUYV422: + case AV_PIX_FMT_YUV422P: + case AV_PIX_FMT_YUV444P: + case AV_PIX_FMT_YUV410P: + case AV_PIX_FMT_YUV411P: + case AV_PIX_FMT_YUVJ420P: + case AV_PIX_FMT_YUVJ422P: + case AV_PIX_FMT_YUVJ444P: + case AV_PIX_FMT_UYVY422: + case AV_PIX_FMT_UYYVYY411: + case AV_PIX_FMT_NV12: + case AV_PIX_FMT_NV21: + case AV_PIX_FMT_YUV440P: + case AV_PIX_FMT_YUVJ440P: + case AV_PIX_FMT_YUVA420P: + case AV_PIX_FMT_YUV420P16LE: + case AV_PIX_FMT_YUV420P16BE: + case AV_PIX_FMT_YUV422P16LE: + case AV_PIX_FMT_YUV422P16BE: + case AV_PIX_FMT_YUV444P16LE: + case AV_PIX_FMT_YUV444P16BE: + case AV_PIX_FMT_YUV420P9BE: + case AV_PIX_FMT_YUV420P9LE: + case AV_PIX_FMT_YUV420P10BE: + case AV_PIX_FMT_YUV420P10LE: + case AV_PIX_FMT_YUV422P10BE: + case AV_PIX_FMT_YUV422P10LE: + case AV_PIX_FMT_YUV444P9BE: + case AV_PIX_FMT_YUV444P9LE: + case AV_PIX_FMT_YUV444P10BE: + case AV_PIX_FMT_YUV444P10LE: + case AV_PIX_FMT_YUV422P9BE: + case AV_PIX_FMT_YUV422P9LE: + case AV_PIX_FMT_YUVA420P9BE: + case AV_PIX_FMT_YUVA420P9LE: + case AV_PIX_FMT_YUVA422P9BE: + case AV_PIX_FMT_YUVA422P9LE: + case AV_PIX_FMT_YUVA444P9BE: + case AV_PIX_FMT_YUVA444P9LE: + case AV_PIX_FMT_YUVA420P10BE: + case AV_PIX_FMT_YUVA420P10LE: + case AV_PIX_FMT_YUVA422P10BE: + case AV_PIX_FMT_YUVA422P10LE: + case AV_PIX_FMT_YUVA444P10BE: + case AV_PIX_FMT_YUVA444P10LE: + case AV_PIX_FMT_YUVA420P16BE: + case AV_PIX_FMT_YUVA420P16LE: + case AV_PIX_FMT_YUVA422P16BE: + case AV_PIX_FMT_YUVA422P16LE: + case AV_PIX_FMT_YUVA444P16BE: + case AV_PIX_FMT_YUVA444P16LE: + case AV_PIX_FMT_NV16: + case AV_PIX_FMT_NV20LE: + case AV_PIX_FMT_NV20BE: + case AV_PIX_FMT_YVYU422: + case AV_PIX_FMT_YUVA444P: + case AV_PIX_FMT_YUVA422P: + case AV_PIX_FMT_YUV420P12BE: + case AV_PIX_FMT_YUV420P12LE: + case AV_PIX_FMT_YUV420P14BE: + case AV_PIX_FMT_YUV420P14LE: + case AV_PIX_FMT_YUV422P12BE: + case AV_PIX_FMT_YUV422P12LE: + case AV_PIX_FMT_YUV422P14BE: + case AV_PIX_FMT_YUV422P14LE: + case AV_PIX_FMT_YUV444P12BE: + case AV_PIX_FMT_YUV444P12LE: + case AV_PIX_FMT_YUV444P14BE: + case AV_PIX_FMT_YUV444P14LE: + case AV_PIX_FMT_YUVJ411P: + return true; + default: + return false; + } +} + +bool +FFmpegExaminer::has_video () const +{ + return static_cast (_video_stream); }