2 Copyright (C) 2013-2015 Carl Hetherington <cth@carlh.net>
4 This file is part of DCP-o-matic.
6 DCP-o-matic is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or
9 (at your option) any later version.
11 DCP-o-matic is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with DCP-o-matic. If not, see <http://www.gnu.org/licenses/>.
21 #include "dcpomatic_log.h"
23 #include <libavcodec/avcodec.h>
24 #include <libavformat/avformat.h>
25 #include <libavutil/pixfmt.h>
26 #include <libavutil/pixdesc.h>
27 #include <libavutil/eval.h>
28 #include <libavutil/display.h>
30 #include "ffmpeg_examiner.h"
31 #include "ffmpeg_content.h"
33 #include "ffmpeg_audio_stream.h"
34 #include "ffmpeg_subtitle_stream.h"
36 #include <boost/foreach.hpp>
45 using boost::shared_ptr;
46 using boost::optional;
48 /* This is how many frames from the start of any video that we will examine to see if we
49 * can spot soft 2:3 pull-down ("telecine").
51 static const int PULLDOWN_CHECK_FRAMES = 16;
54 /** @param job job that the examiner is operating in, or 0 */
55 FFmpegExaminer::FFmpegExaminer (shared_ptr<const FFmpegContent> c, shared_ptr<Job> job)
58 , _need_video_length (false)
61 /* Find audio and subtitle streams */
63 for (uint32_t i = 0; i < _format_context->nb_streams; ++i) {
64 AVStream* s = _format_context->streams[i];
65 if (s->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
67 /* This is a hack; sometimes it seems that _audio_codec_context->channel_layout isn't set up,
68 so bodge it here. No idea why we should have to do this.
71 if (s->codec->channel_layout == 0) {
72 s->codec->channel_layout = av_get_default_channel_layout (s->codec->channels);
75 DCPOMATIC_ASSERT (_format_context->duration != AV_NOPTS_VALUE);
76 DCPOMATIC_ASSERT (s->codec->codec);
77 DCPOMATIC_ASSERT (s->codec->codec->name);
79 _audio_streams.push_back (
80 shared_ptr<FFmpegAudioStream> (
81 new FFmpegAudioStream (
83 s->codec->codec->name,
85 s->codec->sample_rate,
86 llrint ((double (_format_context->duration) / AV_TIME_BASE) * s->codec->sample_rate),
92 } else if (s->codec->codec_type == AVMEDIA_TYPE_SUBTITLE) {
93 _subtitle_streams.push_back (shared_ptr<FFmpegSubtitleStream> (new FFmpegSubtitleStream (subtitle_stream_name (s), s->id)));
98 /* See if the header has duration information in it */
99 _need_video_length = _format_context->duration == AV_NOPTS_VALUE;
100 if (!_need_video_length) {
101 _video_length = llrint ((double (_format_context->duration) / AV_TIME_BASE) * video_frame_rate().get());
105 if (job && _need_video_length) {
106 job->sub (_("Finding length"));
109 /* Run through until we find:
111 * - the first audio for each stream.
112 * - the top-field-first and repeat-first-frame values ("temporal_reference") for the first PULLDOWN_CHECK_FRAMES video frames.
115 int64_t const len = _file_group.length ();
116 /* A string which we build up to describe the top-field-first and repeat-first-frame values for the first few frames.
117 * It would be nicer to use something like vector<bool> here but we want to search the array for a pattern later,
118 * and a string seems a reasonably neat way to do that.
120 string temporal_reference;
122 int r = av_read_frame (_format_context, &_packet);
129 job->set_progress (float (_format_context->pb->pos) / len);
131 job->set_progress_unknown ();
135 AVCodecContext* context = _format_context->streams[_packet.stream_index]->codec;
137 if (_video_stream && _packet.stream_index == _video_stream.get()) {
138 video_packet (context, temporal_reference);
141 bool got_all_audio = true;
143 for (size_t i = 0; i < _audio_streams.size(); ++i) {
144 if (_audio_streams[i]->uses_index (_format_context, _packet.stream_index)) {
145 audio_packet (context, _audio_streams[i]);
147 if (!_audio_streams[i]->first_audio) {
148 got_all_audio = false;
152 av_packet_unref (&_packet);
154 if (_first_video && got_all_audio && temporal_reference.size() >= (PULLDOWN_CHECK_FRAMES * 2)) {
161 /* This code taken from get_rotation() in ffmpeg:cmdutils.c */
162 AVStream* stream = _format_context->streams[*_video_stream];
163 AVDictionaryEntry* rotate_tag = av_dict_get (stream->metadata, "rotate", 0, 0);
164 uint8_t* displaymatrix = av_stream_get_side_data (stream, AV_PKT_DATA_DISPLAYMATRIX, 0);
167 if (rotate_tag && *rotate_tag->value && strcmp(rotate_tag->value, "0")) {
169 _rotation = av_strtod (rotate_tag->value, &tail);
175 if (displaymatrix && !_rotation) {
176 _rotation = - av_display_rotation_get ((int32_t*) displaymatrix);
179 _rotation = *_rotation - 360 * floor (*_rotation / 360 + 0.9 / 360);
182 LOG_GENERAL("Temporal reference was %1", temporal_reference);
183 if (temporal_reference.find("T2T3B2B3T2T3B2B3") != string::npos || temporal_reference.find("B2B3T2T3B2B3T2T3") != string::npos) {
184 /* The magical sequence (taken from mediainfo) suggests that 2:3 pull-down is in use */
186 LOG_GENERAL_NC("Suggest that this may be 2:3 pull-down (soft telecine)");
189 #ifdef DCPOMATIC_VARIANT_SWAROOP
190 AVDictionaryEntry* e = av_dict_get (_format_context->metadata, SWAROOP_ID_TAG, 0, 0);
198 /** @param temporal_reference A string to which we should add two characters per frame;
199 * the first is T or B depending on whether it's top- or bottom-field first,
200 * ths seconds is 3 or 2 depending on whether "repeat_pict" is true or not.
203 FFmpegExaminer::video_packet (AVCodecContext* context, string& temporal_reference)
205 DCPOMATIC_ASSERT (_video_stream);
207 if (_first_video && !_need_video_length && temporal_reference.size() >= (PULLDOWN_CHECK_FRAMES * 2)) {
212 if (avcodec_decode_video2 (context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
214 _first_video = frame_time (_format_context->streams[_video_stream.get()]);
216 if (_need_video_length) {
217 _video_length = frame_time (
218 _format_context->streams[_video_stream.get()]
219 ).get_value_or (ContentTime ()).frames_round (video_frame_rate().get ());
221 if (temporal_reference.size() < (PULLDOWN_CHECK_FRAMES * 2)) {
222 temporal_reference += (_frame->top_field_first ? "T" : "B");
223 temporal_reference += (_frame->repeat_pict ? "3" : "2");
229 FFmpegExaminer::audio_packet (AVCodecContext* context, shared_ptr<FFmpegAudioStream> stream)
231 if (stream->first_audio) {
236 if (avcodec_decode_audio4 (context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
237 stream->first_audio = frame_time (stream->stream (_format_context));
241 optional<ContentTime>
242 FFmpegExaminer::frame_time (AVStream* s) const
244 optional<ContentTime> t;
246 int64_t const bet = av_frame_get_best_effort_timestamp (_frame);
247 if (bet != AV_NOPTS_VALUE) {
248 t = ContentTime::from_seconds (bet * av_q2d (s->time_base));
255 FFmpegExaminer::video_frame_rate () const
257 DCPOMATIC_ASSERT (_video_stream);
258 return av_q2d(av_guess_frame_rate(_format_context, _format_context->streams[_video_stream.get()], 0));
262 FFmpegExaminer::video_size () const
264 return dcp::Size (video_codec_context()->width, video_codec_context()->height);
267 /** @return Length according to our content's header */
269 FFmpegExaminer::video_length () const
271 return max (Frame (1), _video_length);
275 FFmpegExaminer::sample_aspect_ratio () const
277 DCPOMATIC_ASSERT (_video_stream);
278 AVRational sar = av_guess_sample_aspect_ratio (_format_context, _format_context->streams[_video_stream.get()], 0);
280 /* I assume this means that we don't know */
281 return optional<double> ();
283 return double (sar.num) / sar.den;
287 FFmpegExaminer::subtitle_stream_name (AVStream* s) const
289 string n = stream_name (s);
299 FFmpegExaminer::stream_name (AVStream* s) const
304 AVDictionaryEntry const * lang = av_dict_get (s->metadata, "language", 0, 0);
309 AVDictionaryEntry const * title = av_dict_get (s->metadata, "title", 0, 0);
322 FFmpegExaminer::bits_per_pixel () const
324 if (video_codec_context()->pix_fmt == -1) {
325 return optional<int>();
328 AVPixFmtDescriptor const * d = av_pix_fmt_desc_get (video_codec_context()->pix_fmt);
329 DCPOMATIC_ASSERT (d);
330 return av_get_bits_per_pixel (d);
334 FFmpegExaminer::yuv () const
336 switch (video_codec_context()->pix_fmt) {
337 case AV_PIX_FMT_YUV420P:
338 case AV_PIX_FMT_YUYV422:
339 case AV_PIX_FMT_YUV422P:
340 case AV_PIX_FMT_YUV444P:
341 case AV_PIX_FMT_YUV410P:
342 case AV_PIX_FMT_YUV411P:
343 case AV_PIX_FMT_YUVJ420P:
344 case AV_PIX_FMT_YUVJ422P:
345 case AV_PIX_FMT_YUVJ444P:
346 case AV_PIX_FMT_UYVY422:
347 case AV_PIX_FMT_UYYVYY411:
348 case AV_PIX_FMT_NV12:
349 case AV_PIX_FMT_NV21:
350 case AV_PIX_FMT_YUV440P:
351 case AV_PIX_FMT_YUVJ440P:
352 case AV_PIX_FMT_YUVA420P:
353 case AV_PIX_FMT_YUV420P16LE:
354 case AV_PIX_FMT_YUV420P16BE:
355 case AV_PIX_FMT_YUV422P16LE:
356 case AV_PIX_FMT_YUV422P16BE:
357 case AV_PIX_FMT_YUV444P16LE:
358 case AV_PIX_FMT_YUV444P16BE:
359 case AV_PIX_FMT_YUV420P9BE:
360 case AV_PIX_FMT_YUV420P9LE:
361 case AV_PIX_FMT_YUV420P10BE:
362 case AV_PIX_FMT_YUV420P10LE:
363 case AV_PIX_FMT_YUV422P10BE:
364 case AV_PIX_FMT_YUV422P10LE:
365 case AV_PIX_FMT_YUV444P9BE:
366 case AV_PIX_FMT_YUV444P9LE:
367 case AV_PIX_FMT_YUV444P10BE:
368 case AV_PIX_FMT_YUV444P10LE:
369 case AV_PIX_FMT_YUV422P9BE:
370 case AV_PIX_FMT_YUV422P9LE:
371 case AV_PIX_FMT_YUVA420P9BE:
372 case AV_PIX_FMT_YUVA420P9LE:
373 case AV_PIX_FMT_YUVA422P9BE:
374 case AV_PIX_FMT_YUVA422P9LE:
375 case AV_PIX_FMT_YUVA444P9BE:
376 case AV_PIX_FMT_YUVA444P9LE:
377 case AV_PIX_FMT_YUVA420P10BE:
378 case AV_PIX_FMT_YUVA420P10LE:
379 case AV_PIX_FMT_YUVA422P10BE:
380 case AV_PIX_FMT_YUVA422P10LE:
381 case AV_PIX_FMT_YUVA444P10BE:
382 case AV_PIX_FMT_YUVA444P10LE:
383 case AV_PIX_FMT_YUVA420P16BE:
384 case AV_PIX_FMT_YUVA420P16LE:
385 case AV_PIX_FMT_YUVA422P16BE:
386 case AV_PIX_FMT_YUVA422P16LE:
387 case AV_PIX_FMT_YUVA444P16BE:
388 case AV_PIX_FMT_YUVA444P16LE:
389 case AV_PIX_FMT_NV16:
390 case AV_PIX_FMT_NV20LE:
391 case AV_PIX_FMT_NV20BE:
392 case AV_PIX_FMT_YVYU422:
393 case AV_PIX_FMT_YUVA444P:
394 case AV_PIX_FMT_YUVA422P:
395 case AV_PIX_FMT_YUV420P12BE:
396 case AV_PIX_FMT_YUV420P12LE:
397 case AV_PIX_FMT_YUV420P14BE:
398 case AV_PIX_FMT_YUV420P14LE:
399 case AV_PIX_FMT_YUV422P12BE:
400 case AV_PIX_FMT_YUV422P12LE:
401 case AV_PIX_FMT_YUV422P14BE:
402 case AV_PIX_FMT_YUV422P14LE:
403 case AV_PIX_FMT_YUV444P12BE:
404 case AV_PIX_FMT_YUV444P12LE:
405 case AV_PIX_FMT_YUV444P14BE:
406 case AV_PIX_FMT_YUV444P14LE:
407 case AV_PIX_FMT_YUVJ411P:
415 FFmpegExaminer::has_video () const
417 return static_cast<bool> (_video_stream);