[Bf-blender-cvs] [c0367b19e20] master: Fix: VSE search in mpegts files would fail
Sebastian Parborg
noreply at git.blender.org
Fri Jun 11 14:05:26 CEST 2021
Commit: c0367b19e20e7bd7ffe25588dee6466c1a8157e8
Author: Sebastian Parborg
Date: Thu Jun 10 02:38:28 2021 +0200
Branches: master
https://developer.blender.org/rBc0367b19e20e7bd7ffe25588dee6466c1a8157e8
Fix: VSE search in mpegts files would fail
ffmpeg_generic_seek_workaround did work properly and our start pts
calculation was wrong.
Reviewed By: Richard Antalik
Differential Revision: http://developer.blender.org/D11562
===================================================================
M intern/ffmpeg/ffmpeg_compat.h
M source/blender/imbuf/intern/anim_movie.c
===================================================================
diff --git a/intern/ffmpeg/ffmpeg_compat.h b/intern/ffmpeg/ffmpeg_compat.h
index ce4bfc7d5e2..17e2a7dc05b 100644
--- a/intern/ffmpeg/ffmpeg_compat.h
+++ b/intern/ffmpeg/ffmpeg_compat.h
@@ -92,26 +92,6 @@ void my_guess_pkt_duration(AVFormatContext *s, AVStream *st, AVPacket *pkt)
}
#endif
-FFMPEG_INLINE
-void my_update_cur_dts(AVFormatContext *s, AVStream *ref_st, int64_t timestamp)
-{
- int i;
-
- for (i = 0; i < s->nb_streams; i++) {
- AVStream *st = s->streams[i];
-
- st->cur_dts = av_rescale(timestamp,
- st->time_base.den * (int64_t)ref_st->time_base.num,
- st->time_base.num * (int64_t)ref_st->time_base.den);
- }
-}
-
-FFMPEG_INLINE
-void av_update_cur_dts(AVFormatContext *s, AVStream *ref_st, int64_t timestamp)
-{
- my_update_cur_dts(s, ref_st, timestamp);
-}
-
FFMPEG_INLINE
int64_t timestamp_from_pts_or_dts(int64_t pts, int64_t dts)
{
diff --git a/source/blender/imbuf/intern/anim_movie.c b/source/blender/imbuf/intern/anim_movie.c
index 739ec988121..796f7469fb3 100644
--- a/source/blender/imbuf/intern/anim_movie.c
+++ b/source/blender/imbuf/intern/anim_movie.c
@@ -1048,25 +1048,31 @@ static int64_t ffmpeg_get_seek_pos(struct anim *anim, int position)
return pos;
}
+/* This gives us an estimate of which pts our requested frame will have.
+ * Note that this might be off a bit in certain video files, but it should still be close enough.
+ */
static int64_t ffmpeg_get_pts_to_search(struct anim *anim,
struct anim_index *tc_index,
int position)
{
int64_t pts_to_search;
- int64_t st_time = anim->pFormatCtx->start_time;
- AVStream *v_st = anim->pFormatCtx->streams[anim->videoStream];
- double frame_rate = av_q2d(av_guess_frame_rate(anim->pFormatCtx, v_st, NULL));
- double pts_time_base = av_q2d(v_st->time_base);
if (tc_index) {
int new_frame_index = IMB_indexer_get_frame_index(tc_index, position);
pts_to_search = IMB_indexer_get_pts(tc_index, new_frame_index);
}
else {
- pts_to_search = (long long)floor(((double)position) / pts_time_base / frame_rate + 0.5);
+ int64_t st_time = anim->pFormatCtx->start_time;
+ AVStream *v_st = anim->pFormatCtx->streams[anim->videoStream];
+ AVRational frame_rate = av_guess_frame_rate(anim->pFormatCtx, v_st, NULL);
+ AVRational time_base = v_st->time_base;
+
+ int64_t steps_per_frame = (frame_rate.den * time_base.den) / (frame_rate.num * time_base.num);
+ pts_to_search = position * steps_per_frame;
- if (st_time != AV_NOPTS_VALUE) {
- pts_to_search += st_time / pts_time_base / AV_TIME_BASE;
+ if (st_time != AV_NOPTS_VALUE && st_time != 0) {
+ int64_t start_frame = (double)st_time / AV_TIME_BASE * av_q2d(frame_rate);
+ pts_to_search += start_frame * steps_per_frame;
}
}
return pts_to_search;
@@ -1100,6 +1106,7 @@ static void ffmpeg_decode_video_frame_scan(struct anim *anim, int64_t pts_to_sea
(int64_t)pts_to_search);
int64_t start_gop_frame = anim->cur_key_frame_pts;
+ bool scan_fuzzy = false;
while (anim->cur_pts < pts_to_search) {
av_log(anim->pFormatCtx,
@@ -1114,6 +1121,14 @@ static void ffmpeg_decode_video_frame_scan(struct anim *anim, int64_t pts_to_sea
if (start_gop_frame != anim->cur_key_frame_pts) {
break;
}
+
+ if (anim->cur_pts < pts_to_search &&
+ anim->cur_pts + anim->pFrame->pkt_duration > pts_to_search) {
+ /* Our estimate of the pts was a bit off, but we have the frame we want. */
+ av_log(anim->pFormatCtx, AV_LOG_DEBUG, "SCAN fuzzy frame match\n");
+ scan_fuzzy = true;
+ break;
+ }
}
if (start_gop_frame != anim->cur_key_frame_pts) {
@@ -1126,7 +1141,8 @@ static void ffmpeg_decode_video_frame_scan(struct anim *anim, int64_t pts_to_sea
(int64_t)anim->cur_pts,
(int64_t)pts_to_search);
}
- if (anim->cur_pts == pts_to_search) {
+
+ if (scan_fuzzy || anim->cur_pts == pts_to_search) {
av_log(anim->pFormatCtx, AV_LOG_DEBUG, "SCAN HAPPY: we found our PTS!\n");
}
else {
@@ -1139,17 +1155,19 @@ static void ffmpeg_decode_video_frame_scan(struct anim *anim, int64_t pts_to_sea
* necessary I-frame is not honored. It is not even guaranteed that I-frame, that must be
* decoded will be read. See https://trac.ffmpeg.org/ticket/1607 and
* https://developer.blender.org/T86944. */
-static int ffmpeg_generic_seek_workaround(struct anim *anim, int64_t *requested_pos)
+static int ffmpeg_generic_seek_workaround(struct anim *anim,
+ int64_t *requested_pos,
+ int64_t pts_to_search)
{
AVStream *v_st = anim->pFormatCtx->streams[anim->videoStream];
double frame_rate = av_q2d(av_guess_frame_rate(anim->pFormatCtx, v_st, NULL));
int64_t current_pos = *requested_pos;
+ int64_t offset = 0;
+
+ int64_t cur_pts, prev_pts = -1;
- /* This time offset maximum limit is arbitrary. If some files fails to decode it may be
- * increased. Seek performance will be negatively affected. Small initial offset is
- * necessary because encoder can re-arrange frames as it needs but within it's delay, which
- * is usually small. */
- for (int offset = 5; offset < 25; offset++) {
+ /* Step backward frame by frame until we find the key frame we are looking for. */
+ while (current_pos != 0) {
current_pos = *requested_pos - ((int64_t)(offset)*AV_TIME_BASE / frame_rate);
current_pos = max_ii(current_pos, 0);
@@ -1161,18 +1179,34 @@ static int ffmpeg_generic_seek_workaround(struct anim *anim, int64_t *requested_
/* Read first video stream packet. */
AVPacket *read_packet = av_packet_alloc();
while (av_read_frame(anim->pFormatCtx, read_packet) >= 0) {
- if (anim->cur_packet->stream_index == anim->videoStream) {
+ if (read_packet->stream_index == anim->videoStream) {
break;
}
av_packet_unref(read_packet);
}
- /* If this packet contains I-frame, exit loop. This should be the frame that we need. */
+ /* If this packet contains an I-frame, this could be the frame that we need. */
bool is_key_frame = read_packet->flags & AV_PKT_FLAG_KEY;
+ /* We need to check the packet timestamp as the key frame could be for a GOP forward in the the
+ * video stream. So if it has a larger timestamp than the frame we want, ignore it.
+ */
+ cur_pts = timestamp_from_pts_or_dts(read_packet->pts, read_packet->dts);
av_packet_free(&read_packet);
+
if (is_key_frame) {
- break;
+ if (cur_pts <= pts_to_search) {
+ /* We found the I-frame we were looking for! */
+ break;
+ }
+ if (cur_pts == prev_pts) {
+ /* We got the same key frame packet twice.
+ * This probably means that we have hit the beginning of the stream. */
+ break;
+ }
}
+
+ prev_pts = cur_pts;
+ offset++;
}
*requested_pos = current_pos;
@@ -1182,10 +1216,11 @@ static int ffmpeg_generic_seek_workaround(struct anim *anim, int64_t *requested_
}
/* Seek to last necessary key frame. */
-static int ffmpeg_seek_to_key_frame(struct anim *anim, int position, struct anim_index *tc_index)
+static int ffmpeg_seek_to_key_frame(struct anim *anim,
+ int position,
+ struct anim_index *tc_index,
+ int64_t pts_to_search)
{
- AVStream *v_st = anim->pFormatCtx->streams[anim->videoStream];
-
int64_t pos;
int ret;
@@ -1215,7 +1250,6 @@ static int ffmpeg_seek_to_key_frame(struct anim *anim, int position, struct anim
av_log(anim->pFormatCtx, AV_LOG_DEBUG, "... using BYTE pos\n");
ret = av_seek_frame(anim->pFormatCtx, -1, pos, AVSEEK_FLAG_BYTE);
- av_update_cur_dts(anim->pFormatCtx, v_st, dts);
}
else {
av_log(anim->pFormatCtx, AV_LOG_DEBUG, "... using PTS pos\n");
@@ -1235,7 +1269,7 @@ static int ffmpeg_seek_to_key_frame(struct anim *anim, int position, struct anim
ret = av_seek_frame(anim->pFormatCtx, -1, pos, AVSEEK_FLAG_BACKWARD);
}
else {
- ret = ffmpeg_generic_seek_workaround(anim, &pos);
+ ret = ffmpeg_generic_seek_workaround(anim, &pos, pts_to_search);
av_log(anim->pFormatCtx, AV_LOG_DEBUG, "Adjusted final seek pos = %" PRId64 "\n", pos);
}
@@ -1277,7 +1311,6 @@ static int ffmpeg_seek_to_key_frame(struct anim *anim, int position, struct anim
}
if (ret < 0) {
- int64_t pts_to_search = ffmpeg_get_pts_to_search(anim, tc_index, position);
av_log(anim->pFormatCtx,
AV_LOG_ERROR,
"FETCH: "
@@ -1340,7 +1373,7 @@ static ImBuf *ffmpeg_fetchibuf(struct anim *anim, int position, IMB_Timecode_Typ
av_log(anim->pFormatCtx, AV_LOG_DEBUG, "FETCH: no seek necessary, just continue...\n");
ffmpeg_decode_video_frame(anim);
}
- else if (ffmpeg_seek_to_key_frame(anim, position, tc_index) >= 0) {
+ else if (ffmpeg_seek_to_key_frame(anim, position, tc_index, pts_to_search) >= 0) {
ffmpeg_decode_video_frame_scan(anim, pts_to_search);
}
More information about the Bf-blender-cvs
mailing list