avformat/mov: fix seeking with HEVC open GOP files

This was tested with medias recorded from an iPhone XR and an iPhone 13.

Here is how a typical stream looks like in coding order:

    ┌────────┬─────┬─────┬──────────┐
    │ sample | PTS | DTS | keyframe |
    ├────────┼─────┼─────┼──────────┤
    ┊        ┊     ┊     ┊          ┊
    │   53   │ 560 │ 510 │    No    │
    │   54   │ 540 │ 520 │    No    │
    │   55   │ 530 │ 530 │    No    │
    │   56   │ 550 │ 540 │    No    │
    │   57   │ 600 │ 550 │    Yes   │
    │ * 58   │ 580 │ 560 │    No    │
    │ * 59   │ 570 │ 570 │    No    │
    │ * 60   │ 590 │ 580 │    No    │
    │   61   │ 640 │ 590 │    No    │
    │   62   │ 620 │ 600 │    No    │
    ┊        ┊     ┊     ┊          ┊

In composition/display order:

    ┌────────┬─────┬─────┬──────────┐
    │ sample | PTS | DTS | keyframe |
    ├────────┼─────┼─────┼──────────┤
    ┊        ┊     ┊     ┊          ┊
    │   55   │ 530 │ 530 │    No    │
    │   54   │ 540 │ 520 │    No    │
    │   56   │ 550 │ 540 │    No    │
    │   53   │ 560 │ 510 │    No    │
    │ * 59   │ 570 │ 570 │    No    │
    │ * 58   │ 580 │ 560 │    No    │
    │ * 60   │ 590 │ 580 │    No    │
    │   57   │ 600 │ 550 │    Yes   │
    │   63   │ 610 │ 610 │    No    │
    │   62   │ 620 │ 600 │    No    │
    ┊        ┊     ┊     ┊          ┊

Sample/frame 58, 59 and 60 are B-frames which actually depends on the
key frame (57). Here the key frame is not an IDR but a "CRA" (Clean
Random Access).

Initially, I thought I could rely on the sdtp box (independent and
disposable samples), but unfortunately:

    sdtp[54] is_leading:0 sample_depends_on:1 sample_is_depended_on:0 sample_has_redundancy:0
    sdtp[55] is_leading:0 sample_depends_on:1 sample_is_depended_on:2 sample_has_redundancy:0
    sdtp[56] is_leading:0 sample_depends_on:1 sample_is_depended_on:2 sample_has_redundancy:0
    sdtp[57] is_leading:0 sample_depends_on:2 sample_is_depended_on:0 sample_has_redundancy:0
    sdtp[58] is_leading:0 sample_depends_on:1 sample_is_depended_on:0 sample_has_redundancy:0
    sdtp[59] is_leading:0 sample_depends_on:1 sample_is_depended_on:2 sample_has_redundancy:0
    sdtp[60] is_leading:0 sample_depends_on:1 sample_is_depended_on:2 sample_has_redundancy:0
    sdtp[61] is_leading:0 sample_depends_on:1 sample_is_depended_on:0 sample_has_redundancy:0
    sdtp[62] is_leading:0 sample_depends_on:1 sample_is_depended_on:0 sample_has_redundancy:0

The information that might have been useful here would have been
is_leading, but all the samples are set to 0 so this was unusable.

Instead, we need to rely on sgpd/sbgp tables. In my case the video track
contained 3 sgpd tables with the following grouping types: tscl, sync
and tsas. In the sync table we have the following 2 entries (only):

    sgpd.sync[1]: sync nal_unit_type:0x14
    sgpd.sync[2]: sync nal_unit_type:0x15

(The count starts at 1 because 0 carries the undefined semantic, we'll
see that later in the reference table).

The NAL unit types presented here correspond to:

    libavcodec/hevc.h:    HEVC_NAL_IDR_N_LP       = 20,
    libavcodec/hevc.h:    HEVC_NAL_CRA_NUT        = 21,

In parallel, the sbgp sync table contains the following:

    ┌────┬───────┬─────┐
    │ id │ count │ gdi │
    ├────┼───────┼─────┤
    │  0 │   1   │  1  │
    │  1 │   56  │  0  │
    │  2 │   1   │  2  │
    │  3 │   59  │  0  │
    │  4 │   1   │  2  │
    │  5 │   59  │  0  │
    │  6 │   1   │  2  │
    │  7 │   59  │  0  │
    │  8 │   1   │  2  │
    │  9 │   59  │  0  │
    │ 10 │   1   │  2  │
    │ 11 │   11  │  0  │
    └────┴───────┴─────┘

The gdi column (group description index) directly refers to the index in
the sgpd.sync table. This means the first frame is an IDR, then we have
batches of undefined frames interlaced with CRA frames. No IDR ever
appears again (tried on a 30+ seconds sample).

With that information, we can build an heuristic using the presentation
order.

A few things needed to be introduced in this commit:

1. min_sample_duration is extracted from the stts: we need the minimal
   step between sample in order to PTS-step backward to a valid point
2. In order to avoid a loop over the ctts table systematically during a
   seek, we build an expanded list of sample offsets which will be used
   to translate from DTS to PTS
3. An open_key_samples index to keep track of all the non-IDR key
   frames; for now it only supports HEVC CRA frames. We should probably
   add BLA frames as well, but I don't have any sample so I prefered to
   leave that for later

It is entirely possible I missed something obvious in my approach, but I
couldn't come up with a better solution. Also, as mentioned in the diff,
we could optimize is_open_key_sample(), but the linear scaling overhead
should be fine for now since it only happens in seek events.

Fixing this issue prevents sending broken packets to the decoder. With
FFmpeg hevc decoder the frames are skipped, with VideoToolbox the frames
are glitching.
This commit is contained in:
Clément Bœsch 2022-02-18 23:02:20 +01:00
parent e05e4398c3
commit ab77b878f1
2 changed files with 125 additions and 0 deletions

View File

@ -222,6 +222,11 @@ typedef struct MOVStreamContext {
MOVSbgp *sync_group;
uint8_t *sgpd_sync;
uint32_t sgpd_sync_count;
int32_t *sample_offsets;
int sample_offsets_count;
int *open_key_samples;
int open_key_samples_count;
uint32_t min_sample_duration;
int nb_frames_for_fps;
int64_t duration_for_fps;

View File

@ -50,6 +50,7 @@
#include "libavutil/dovi_meta.h"
#include "libavcodec/ac3tab.h"
#include "libavcodec/flac.h"
#include "libavcodec/hevc.h"
#include "libavcodec/mpegaudiodecheader.h"
#include "libavcodec/mlp_parse.h"
#include "avformat.h"
@ -3882,6 +3883,69 @@ static void mov_fix_index(MOVContext *mov, AVStream *st)
msc->current_index = msc->index_ranges[0].start;
}
static uint32_t get_sgpd_sync_index(const MOVStreamContext *sc, int nal_unit_type)
{
for (uint32_t i = 0; i < sc->sgpd_sync_count; i++)
if (sc->sgpd_sync[i] == HEVC_NAL_CRA_NUT)
return i + 1;
return 0;
}
static int build_open_gop_key_points(AVStream *st)
{
int k;
int sample_id = 0;
uint32_t cra_index;
MOVStreamContext *sc = st->priv_data;
if (st->codecpar->codec_id != AV_CODEC_ID_HEVC || !sc->sync_group_count)
return 0;
/* Build an unrolled index of the samples */
sc->sample_offsets_count = 0;
for (uint32_t i = 0; i < sc->ctts_count; i++)
sc->sample_offsets_count += sc->ctts_data[i].count;
av_freep(&sc->sample_offsets);
sc->sample_offsets = av_calloc(sc->sample_offsets_count, sizeof(*sc->sample_offsets));
if (!sc->sample_offsets)
return AVERROR(ENOMEM);
k = 0;
for (uint32_t i = 0; i < sc->ctts_count; i++)
for (int j = 0; j < sc->ctts_data[i].count; j++)
sc->sample_offsets[k++] = sc->ctts_data[i].duration;
/* The following HEVC NAL type reveal the use of open GOP sync points
* (TODO: BLA types may also be concerned) */
cra_index = get_sgpd_sync_index(sc, HEVC_NAL_CRA_NUT); /* Clean Random Access */
if (!cra_index)
return 0;
/* Build a list of open-GOP key samples */
sc->open_key_samples_count = 0;
for (uint32_t i = 0; i < sc->sync_group_count; i++)
if (sc->sync_group[i].index == cra_index)
sc->open_key_samples_count += sc->sync_group[i].count;
av_freep(&sc->open_key_samples);
sc->open_key_samples = av_calloc(sc->open_key_samples_count, sizeof(*sc->open_key_samples));
if (!sc->open_key_samples)
return AVERROR(ENOMEM);
k = 0;
for (uint32_t i = 0; i < sc->sync_group_count; i++) {
const MOVSbgp *sg = &sc->sync_group[i];
if (sg->index == cra_index)
for (uint32_t j = 0; j < sg->count; j++)
sc->open_key_samples[k++] = sample_id;
sample_id += sg->count;
}
/* Identify the minimal time step between samples */
sc->min_sample_duration = UINT_MAX;
for (uint32_t i = 0; i < sc->stts_count; i++)
sc->min_sample_duration = FFMIN(sc->min_sample_duration, sc->stts_data[i].duration);
return 0;
}
static void mov_build_index(MOVContext *mov, AVStream *st)
{
MOVStreamContext *sc = st->priv_data;
@ -3897,6 +3961,10 @@ static void mov_build_index(MOVContext *mov, AVStream *st)
MOVCtts *ctts_data_old = sc->ctts_data;
unsigned int ctts_count_old = sc->ctts_count;
int ret = build_open_gop_key_points(st);
if (ret < 0)
return;
if (sc->elst_count) {
int i, edit_start_index = 0, multiple_edits = 0;
int64_t empty_duration = 0; // empty duration of the first edit list entry
@ -7774,6 +7842,8 @@ static int mov_read_close(AVFormatContext *s)
av_freep(&sc->rap_group);
av_freep(&sc->sync_group);
av_freep(&sc->sgpd_sync);
av_freep(&sc->sample_offsets);
av_freep(&sc->open_key_samples);
av_freep(&sc->display_matrix);
av_freep(&sc->index_ranges);
@ -8446,6 +8516,49 @@ static int mov_seek_fragment(AVFormatContext *s, AVStream *st, int64_t timestamp
return 0;
}
static int is_open_key_sample(const MOVStreamContext *sc, int sample)
{
// TODO: a bisect search would scale much better
for (int i = 0; i < sc->open_key_samples_count; i++) {
const int oks = sc->open_key_samples[i];
if (oks == sample)
return 1;
if (oks > sample) /* list is monotically increasing so we can stop early */
break;
}
return 0;
}
/*
* Some key sample may be key frames but not IDR frames, so a random access to
* them may not be allowed.
*/
static int can_seek_to_key_sample(AVStream *st, int sample, int64_t requested_pts)
{
MOVStreamContext *sc = st->priv_data;
FFStream *const sti = ffstream(st);
int64_t key_sample_dts, key_sample_pts;
if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
return 1;
if (sample >= sc->sample_offsets_count)
return 1;
key_sample_dts = sti->index_entries[sample].timestamp;
key_sample_pts = key_sample_dts + sc->sample_offsets[sample] + sc->dts_shift;
/*
* If the sample needs to be presented before an open key sample, they may
* not be decodable properly, even though they come after in decoding
* order.
*/
if (is_open_key_sample(sc, sample) && key_sample_pts > requested_pts)
return 0;
return 1;
}
static int mov_seek_stream(AVFormatContext *s, AVStream *st, int64_t timestamp, int flags)
{
MOVStreamContext *sc = st->priv_data;
@ -8461,12 +8574,19 @@ static int mov_seek_stream(AVFormatContext *s, AVStream *st, int64_t timestamp,
if (ret < 0)
return ret;
for (;;) {
sample = av_index_search_timestamp(st, timestamp, flags);
av_log(s, AV_LOG_TRACE, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
if (sample < 0 && sti->nb_index_entries && timestamp < sti->index_entries[0].timestamp)
sample = 0;
if (sample < 0) /* not sure what to do */
return AVERROR_INVALIDDATA;
if (!sample || can_seek_to_key_sample(st, sample, timestamp))
break;
timestamp -= FFMAX(sc->min_sample_duration, 1);
}
mov_current_sample_set(sc, sample);
av_log(s, AV_LOG_TRACE, "stream %d, found sample %d\n", st->index, sc->current_sample);
/* adjust ctts index */