Use enum class for VideoRange.

[dcpomatic.git] / src / lib / ffmpeg_examiner.cc

diff --git a/src/lib/ffmpeg_examiner.cc b/src/lib/ffmpeg_examiner.cc
index cc6a62ff68e180af2118e0e0a23c30e0a6392e52..384c21e9c27288cbfc26b92c7c640985cfcf5a0f 100644 (file)
--- a/src/lib/ffmpeg_examiner.cc
+++ b/src/lib/ffmpeg_examiner.cc
@@ -18,6 +18,7 @@
 
 */
 
+#include "dcpomatic_log.h"
 #include "ffmpeg_examiner.h"
 #include "ffmpeg_content.h"
 #include "job.h"
@@ -35,7 +36,6 @@ extern "C" {
 #include <libavutil/display.h>
 }
 DCPOMATIC_ENABLE_WARNINGS
-#include <boost/foreach.hpp>
 #include <iostream>
 
 #include "i18n.h"
@@ -43,16 +43,24 @@ DCPOMATIC_ENABLE_WARNINGS
 using std::string;
 using std::cout;
 using std::max;
-using boost::shared_ptr;
+using std::vector;
+using std::shared_ptr;
 using boost::optional;
 using namespace dcpomatic;
 
 
+/* 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<const FFmpegContent> c, shared_ptr<Job> job)
        : FFmpeg (c)
        , _video_length (0)
        , _need_video_length (false)
+       , _pulldown (false)
 {
        /* Find audio and subtitle streams */
 
@@ -106,9 +114,15 @@ DCPOMATIC_DISABLE_WARNINGS
        /* Run through until we find:
         *   - the first video.
         *   - the first audio for each stream.
+        *   - 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<bool> 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) {
@@ -127,7 +141,7 @@ DCPOMATIC_DISABLE_WARNINGS
 DCPOMATIC_ENABLE_WARNINGS
 
                if (_video_stream && _packet.stream_index == _video_stream.get()) {
-                       video_packet (context);
+                       video_packet (context, temporal_reference);
                }
 
                bool got_all_audio = true;
@@ -143,7 +157,7 @@ DCPOMATIC_ENABLE_WARNINGS
 
                av_packet_unref (&_packet);
 
-               if (_first_video && got_all_audio) {
+               if (_first_video && got_all_audio && temporal_reference.size() >= (PULLDOWN_CHECK_FRAMES * 2)) {
                        /* All done */
                        break;
                }
@@ -169,25 +183,27 @@ DCPOMATIC_ENABLE_WARNINGS
                }
 
                _rotation = *_rotation - 360 * floor (*_rotation / 360 + 0.9 / 360);
-
-               DCPOMATIC_ASSERT (fabs (*_rotation - 90 * round (*_rotation / 90)) < 2);
        }
 
-#ifdef DCPOMATIC_VARIANT_SWAROOP
-       AVDictionaryEntry* e = av_dict_get (_format_context->metadata, SWAROOP_ID_TAG, 0, 0);
-       if (e) {
-               _id = e->value;
+       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)");
        }
-#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)
 {
        DCPOMATIC_ASSERT (_video_stream);
 
-       if (_first_video && !_need_video_length) {
+       if (_first_video && !_need_video_length && temporal_reference.size() >= (PULLDOWN_CHECK_FRAMES * 2)) {
                return;
        }
 
@@ -203,6 +219,10 @@ DCPOMATIC_ENABLE_WARNINGS
                                _format_context->streams[_video_stream.get()]
                                ).get_value_or (ContentTime ()).frames_round (video_frame_rate().get ());
                }
+               if (temporal_reference.size() < (PULLDOWN_CHECK_FRAMES * 2)) {
+                       temporal_reference += (_frame->top_field_first ? "T" : "B");
+                       temporal_reference += (_frame->repeat_pict ? "3" : "2");
+               }
        }
 }
 
@@ -411,9 +431,9 @@ FFmpegExaminer::range () const
        switch (color_range()) {
        case AVCOL_RANGE_MPEG:
        case AVCOL_RANGE_UNSPECIFIED:
-               return VIDEO_RANGE_VIDEO;
+               return VideoRange::VIDEO;
        case AVCOL_RANGE_JPEG:
        default:
-               return VIDEO_RANGE_FULL;
+               return VideoRange::FULL;
        }
 }