转自：http://www.cnitblog.com/luofuchong/archive/2014/11/28/89869.html
https://www.suninf.net/2018/02/ffmpeg-timestamps.html
https://www.cnblogs.com/loveclover/archive/2011/03/23/1993065.html
ffmpeg存在多个时间基准(time_base)，对应不同的阶段(结构体)，每个time_base具体的值不一样，ffmpeg提供函数在各个time_base中进行切换。搞清楚各个time_base的来源，对于阅读ffmpeg的代码很重要。
一、time_base
1、AVStream(libavformat/avformat.h)
typedef struct AVStream {
/**
* This is the fundamental unit of time (in seconds) in terms
* of which frame timestamps are represented.
*
* decoding: set by libavformat
* encoding: May be set by the caller before avformat_write_header() to
* provide a hint to the muxer about the desired timebase. In
* avformat_write_header(), the muxer will overwrite this field
* with the timebase that will actually be used for the timestamps
* written into the file (which may or may not be related to the
* user-provided one, depending on the format).
*/
AVRational time_base;
/**
* Decoding: pts of the first frame of the stream in presentation order, in stream time base.
* Only set this if you are absolutely 100% sure that the value you set
* it to really is the pts of the first frame.
* This may be undefined (AV_NOPTS_VALUE).
* @note The ASF header does NOT contain a correct start_time the ASF
* demuxer must NOT set this.
*/
int64_t start_time;
/**
* Decoding: duration of the stream, in stream time base.
* If a source file does not specify a duration, but does specify
* a bitrate, this value will be estimated from bitrate and file size.
*/
int64_t duration;
从上面的信息可以看到，AVStream->time_base单位为秒。
那AVStream->time_base具体的值是多少呢？下面以mpegts_demuxer为例：
static int mpegts_set_stream_info(AVStream *st, PESContext *pes,
uint32_t stream_type, uint32_t prog_reg_desc)
{
avpriv_set_pts_info(st, 33, 1, 90000);
void avpriv_set_pts_info(AVStream *s, int pts_wrap_bits,
unsigned int pts_num, unsigned int pts_den)
{
AVRational new_tb;
if (av_reduce(&new_tb.num, &new_tb.den, pts_num, pts_den, INT_MAX)) {
if (new_tb.num != pts_num)
av_log(NULL, AV_LOG_DEBUG,
"st:%d removing common factor %d from timebase\n",
s->index, pts_num / new_tb.num);
} else
av_log(NULL, AV_LOG_WARNING,
"st:%d has too large timebase, reducing\n", s->index);
if (new_tb.num <= 0 || new_tb.den <= 0) {
av_log(NULL, AV_LOG_ERROR,
"Ignoring attempt to set invalid timebase %d/%d for st:%d\n",
new_tb.num, new_tb.den,
s->index);
return;
}
s->time_base = new_tb;
av_codec_set_pkt_timebase(s->codec, new_tb);
s->pts_wrap_bits = pts_wrap_bits;
}
通过avpriv_set_pts_info(st, 33, 1, 90000)函数，设置AVStream->time_base为1/90000。为什么是90000？因为mpeg的pts、dts都是以90kHz来采样的，所以采样间隔为1/90000秒。
2、AVCodecContext
typedef struct AVCodecContext {
/**
* This is the fundamental unit of time (in seconds) in terms
* of which frame timestamps are represented. For fixed-fps content,
* timebase should be 1/framerate and timestamp increments should be
* identically 1.
* - encoding: MUST be set by user.
* - decoding: Set by libavcodec.
*/
AVRational time_base;
从上面的信息可以看到，AVCodecContext->time_base单位同样为秒，不过精度没有AVStream->time_base高，大小为1/framerate。
下面以ffmpeg转码工具为例：
static int transcode_init(void)
{
if (enc_ctx->codec_type == AVMEDIA_TYPE_VIDEO) {
if (ost->filter && !ost->frame_rate.num)
ost->frame_rate = av_buffersink_get_frame_rate(ost->filter->filter);
if (ist && !ost->frame_rate.num)
ost->frame_rate = ist->framerate;
if (ist && !ost->frame_rate.num)
ost->frame_rate = ist->st->r_frame_rate;
if (ist && !ost->frame_rate.num) {
ost->frame_rate = (AVRational){25, 1};
av_log(NULL, AV_LOG_WARNING,
"No information "
"about the input framerate is available. Falling "
"back to a default value of 25fps for output stream #%d:%d. Use the -r option "
"if you want a different framerate.\n",
ost->file_index, ost->index);
}
// ost->frame_rate = ist->st->avg_frame_rate.num ? ist->st->avg_frame_rate : (AVRational){25, 1};
if (ost->enc && ost->enc->supported_framerates && !ost->force_fps) {
int idx = av_find_nearest_q_idx(ost->frame_rate, ost->enc->supported_framerates);
ost->frame_rate = ost->enc->supported_framerates[idx];
}
if (enc_ctx->codec_id == AV_CODEC_ID_MPEG4) {
av_reduce(&ost->frame_rate.num, &ost->frame_rate.den,
ost->frame_rate.num, ost->frame_rate.den, 65535);
}
}
switch (enc_ctx->codec_type) {
case AVMEDIA_TYPE_VIDEO:
enc_ctx->time_base = av_inv_q(ost->frame_rate);
if (ost->filter && !(enc_ctx->time_base.num && enc_ctx->time_base.den))
enc_ctx->time_base = ost->filter->filter->inputs[0]->time_base;
if ( av_q2d(enc_ctx->time_base) < 0.001 && video_sync_method != VSYNC_PASSTHROUGH
&& (video_sync_method == VSYNC_CFR || video_sync_method == VSYNC_VSCFR || (video_sync_method == VSYNC_AUTO && !(oc->oformat->flags & AVFMT_VARIABLE_FPS)))){
av_log(oc, AV_LOG_WARNING, "Frame rate very high for a muxer not efficiently supporting it.\n"
"Please consider specifying a lower framerate, a different muxer or -vsync 2\n");
}
首先获取ost->frame_rate，然后计算enc_ctx->time_base = 1/ost->frame_rate。
总结：
AVStream->time_base比AVCodecContext->time_base精度要高(数值要小)，比如AVStream->time_base为1/90000，而AVCodecContext->time_base为1/30(假设frame_rate为30)；同样的pts和dts，以AVStream->time_base为单位，数值要比以AVCodecContext->time_base为单位要大。
二、pts、dts
那各个结构下，pts和dts使用哪个time_base来表示呢？
1、AVPacket
typedef struct AVPacket {
/**
* Presentation timestamp in AVStream->time_base units; the time at which
* the decompressed packet will be presented to the user.
* Can be AV_NOPTS_VALUE if it is not stored in the file.
* pts MUST be larger or equal to dts as presentation cannot happen before
* decompression, unless one wants to view hex dumps. Some formats misuse
* the terms dts and pts/cts to mean something different. Such timestamps
* must be converted to true pts/dts before they are stored in AVPacket.
*/
int64_t pts;
/**
* Decompression timestamp in AVStream->time_base units; the time at which
* the packet is decompressed.
* Can be AV_NOPTS_VALUE if it is not stored in the file.
*/
int64_t dts;
从上面可以看到，AVPacket下的pts和dts以AVStream->time_base为单位(数值比较大)。这也很容易理解，根据mpeg的协议，压缩后或解压前的数据，pts和dts是90kHz时钟的采样值，时间间隔就是AVStream->time_base。
2、AVFrame
typedef struct AVFrame {
/**
* Presentation timestamp in time_base units (time when frame should be shown to user).
*/
int64_t pts;
/**
* PTS copied from the AVPacket that was decoded to produce this frame.
*/
int64_t pkt_pts;
/**
* DTS copied from the AVPacket that triggered returning this frame. (if frame threading isn't used)
* This is also the Presentation time of this AVFrame calculated from
* only AVPacket.dts values without pts values.
*/
int64_t pkt_dts;
注意：
AVFrame里面的pkt_pts和pkt_dts是拷贝自AVPacket，同样以AVStream->time_base为单位；而pts是为输出(显示)准备的，以AVCodecContex->time_base为单位)。//FIXME
3、InputStream
typedef struct InputStream {
int file_index;
AVStream *st;
AVCodecContext *dec_ctx;
int64_t start; /* time when read started */
/* predicted dts of the next packet read for this stream or (when there are
* several frames in a packet) of the next frame in current packet (in AV_TIME_BASE units) */
int64_t next_dts;
int64_t dts; ///< dts of the last packet read for this stream (in AV_TIME_BASE units)
int64_t next_pts; ///< synthetic pts for the next decode frame (in AV_TIME_BASE units)
int64_t pts; ///< current pts of the decoded frame (in AV_TIME_BASE units)
从上面可以看到，InputStream下的pts和dts以AV_TIME_BASE为单位(微秒)，至于为什么要转化为微妙，可能是为了避免使用浮点数。
4、OutputStream
typedef struct OutputStream {
int file_index; /* file index */
int index; /* stream index in the output file */
int source_index; /* InputStream index */
AVStream *st; /* stream in the output file */
int encoding_needed; /* true if encoding needed for this stream */
int frame_number;
/* input pts and corresponding output pts
for A/V sync */
struct InputStream *sync_ist; /* input stream to sync against */
int64_t sync_opts; /* output frame counter, could be changed to some true timestamp */ // FIXME look at frame_number
/* pts of the first frame encoded for this stream, used for limiting
* recording time */
int64_t first_pts;
/* dts of the last packet sent to the muxer */
int64_t last_mux_dts;
AVBitStreamFilterContext *bitstream_filters;
AVCodecContext *enc_ctx;
AVCodec *enc;
int64_t max_frames;
AVFrame *filtered_frame;
OutputStream涉及音视频同步，结构和InputStream不同，暂时只作记录，不分析。
三、各个time_base之间转换
ffmpeg提供av_rescale_q函数用于time_base之间转换，av_rescale_q(a,b,c)作用相当于执行a*b/c，通过设置b,c的值，可以很方便的实现time_base之间转换。
例如：
1、InputStream(AV_TIME_BASE)到AVPacket(AVStream->time_base)
static int decode_video(InputStream *ist, AVPacket *pkt, int *got_output)
{
pkt->dts = av_rescale_q(ist->dts, AV_TIME_BASE_Q, ist->st->time_base);
2、AVPacket(AVStream->time_base)到InputStream(AV_TIME_BASE)
static int process_input_packet(InputStream *ist, const AVPacket *pkt)
{
if (pkt->dts != AV_NOPTS_VALUE) {
ist->next_dts = ist->dts = av_rescale_q(pkt->dts, ist->st->time_base, AV_TIME_BASE_Q);
四、后记：
AVFrame->pts和AVPacket->pts、AVPacket->dts的值，在解码/编码后，会经历短暂的time_base不匹配的情况：
1、解码后
static int decode_video(InputStream *ist, AVPacket *pkt, int *got_output)
{
decoded_frame = ist->decoded_frame;
pkt->dts = av_rescale_q(ist->dts, AV_TIME_BASE_Q, ist->st->time_base);
update_benchmark(NULL);
ret = avcodec_decode_video2(ist->dec_ctx,
decoded_frame, got_output, pkt);
best_effort_timestamp= av_frame_get_best_effort_timestamp(decoded_frame);
if(best_effort_timestamp != AV_NOPTS_VALUE)
ist->next_pts = ist->pts = av_rescale_q(decoded_frame->pts = best_effort_timestamp, ist->st->time_base, AV_TIME_BASE_Q);
解码后，decoded_frame->pts的值使用AVStream->time_base为单位，后在AVFilter里面转换成以AVCodecContext->time_base为单位。 //FIXME
2、编码后
static void do_video_out(AVFormatContext *s,
OutputStream *ost,
AVFrame *in_picture)
{
ret = avcodec_encode_video2(enc, &pkt, in_picture, &got_packet);
if (got_packet) {
if (debug_ts) {
av_log(NULL, AV_LOG_INFO, "encoder -> type:video "
"pkt_pts:%s pkt_pts_time:%s pkt_dts:%s pkt_dts_time:%s\n",
av_ts2str(pkt.pts), av_ts2timestr(pkt.pts, &enc->time_base),
av_ts2str(pkt.dts), av_ts2timestr(pkt.dts, &enc->time_base));
}
if (pkt.pts == AV_NOPTS_VALUE && !(enc->codec->capabilities & CODEC_CAP_DELAY))
pkt.pts = ost->sync_opts;
av_packet_rescale_ts(&pkt, enc->time_base, ost->st->time_base);
if (debug_ts) {
av_log(NULL, AV_LOG_INFO, "encoder -> type:video "
"pkt_pts:%s pkt_pts_time:%s pkt_dts:%s pkt_dts_time:%s\n",
av_ts2str(pkt.pts), av_ts2timestr(pkt.pts, &ost->st->time_base),
av_ts2str(pkt.dts), av_ts2timestr(pkt.dts, &ost->st->time_base));
}
frame_size = pkt.size;
write_frame(s, &pkt, ost);
/* if two pass, output log */
if (ost->logfile && enc->stats_out) {
fprintf(ost->logfile, "%s", enc->stats_out);
}
}
编码后，pkt.pts、pkt.dts使用AVCodecContext->time_base为单位，后通过调用"av_packet_rescale_ts"转换为AVStream->time_base为单位。
第二篇：
FFmpeg时间戳整理
@SUNINF · FEB 4, 2018
Catalog
使用FFmpeg库进行音视频处理时，经常需要处理视频时间、编解码的dts/pts时间戳，时间的准确处理对音视频处理至关重要。
time_base
时间基(time_base)是FFmpeg中作为时间单位的概念。不同的音视频封装格式，可以使用不同的time_base。
FFmpeg内部的时间都会基于AV_TIME_BASE作为时间单位，比如AVStream中的duration表示这个流的长度为duration个AV_TIME_BASE。
/**
* Internal time base represented as integer
*/
#define AV_TIME_BASE 1000000AVStream中的time_base的定义：
typedef struct AVStream {
/**
* This is the fundamental unit of time (in seconds) in terms
* of which frame timestamps are represented.
*
* decoding: set by libavformat
* encoding: May be set by the caller before avformat_write_header() to
* provide a hint to the muxer about the desired timebase. In
* avformat_write_header(), the muxer will overwrite this field
* with the timebase that will actually be used for the timestamps
* written into the file (which may or may not be related to the
* user-provided one, depending on the format).
*/
AVRational time_base;
/**
* Decoding: duration of the stream, in stream time base.
* If a source file does not specify a duration, but does specify
* a bitrate, this value will be estimated from bitrate and file size.
*
* Encoding: May be set by the caller before avformat_write_header() to
* provide a hint to the muxer about the estimated duration.
*/
int64_t duration;
// ...
};为了时间计算的精度，FFmpeg引入有理数的时间基形式 AV_TIME_BASE_Q
/**
* Rational number (pair of numerator and denominator).
*/
typedef struct AVRational{
int num; ///< Numerator
int den; ///< Denominator
} AVRational;
/**
* Internal time base represented as fractional value
*/
#define AV_TIME_BASE_Q (AVRational){1, AV_TIME_BASE}ffmpeg提供了一个把AVRatioal有理数结构转换成double的函数：
/**
* Convert an AVRational to a `double`.
* @param a AVRational to convert
* @return `a` in floating-point form
* @see av_d2q()
*/
static inline double av_q2d(AVRational a){
return a.num / (double) a.den;
}时间戳换算
比如计算视频的总时长：
AVFormatContext *ifmt_ctx = NULL;
avformat_open_input(&ifmt_ctx, filename, NULL, NULL);
double totle_seconds = ifmt_ctx->duration * av_q2d(AV_TIME_BASE_Q);根据pts计算一帧在视频中对应的秒数位置：
double sec = enc_pkt.pts * av_q2d(ofmt_ctx->streams[stream_index]->time_base);ffmpeg内部的时间戳与标准的时间转换方法
timestamp(ffmpeg内部时间戳) = AV_TIME_BASE * time(秒)
time(秒) = AV_TIME_BASE_Q * timestamp(ffmpeg内部时间戳)
当需要把视频跳转到N秒的时候：
int64_t timestamp = N * AV_TIME_BASE;
av_seek_frame(fmtctx, index_of_video, timestamp, AVSEEK_FLAG_BACKWARD);时间基转换函数
另外，FFmpeg为我们提供了不同时间基之间的转换函数，把时间戳从一个时基调整到另外一个时基：
/**
* Rescale a 64-bit integer by 2 rational numbers.
*
* The operation is mathematically equivalent to `a * bq / cq`.
*
* This function is equivalent to av_rescale_q_rnd() with #AV_ROUND_NEAR_INF.
*
* @see av_rescale(), av_rescale_rnd(), av_rescale_q_rnd()
*/
int64_t av_rescale_q(int64_t a, AVRational bq, AVRational cq) av_const;dts/pts计算
- dts: decoding time stamp （解码时间戳）
- pts: present time stamp （显示时间戳）
转码流程中的dts/pts转换：
输入
AVPacket packet;
av_read_frame(ifmt_ctx, &packet)
解码
AVFrame *frame = NULL;
// 转换为解码器的时间戳
av_packet_rescale_ts(&packet,
ifmt_ctx->streams[stream_index]->time_base,
ifmt_ctx->streams[stream_index]->codec->time_base);
dec_func = (type == AVMEDIA_TYPE_VIDEO) ? avcodec_decode_video2 :
avcodec_decode_audio4;
ret = dec_func(ifmt_ctx->streams[stream_index]->codec, frame,
&got_frame, &packet);
编码
AVPacket enc_pkt;
int (*enc_func)(AVCodecContext *, AVPacket *, const AVFrame *, int *) =
(ifmt_ctx->streams[stream_index]->codec->codec_type ==
AVMEDIA_TYPE_VIDEO) ? avcodec_encode_video2 : avcodec_encode_audio2;
/* encode filtered frame */
enc_pkt.data = NULL;
enc_pkt.size = 0;
av_init_packet(&enc_pkt);
ret = enc_func(ofmt_ctx->streams[stream_index]->codec, &enc_pkt,
filt_frame, got_frame);
输出
// 编码后的enc_pkt的时间戳以AVCodecContext->time_base为单位
// 编码器时间戳转为输出流AVStream时间戳
av_packet_rescale_ts(&enc_pkt,
ofmt_ctx->streams[stream_index]->codec->time_base,
ofmt_ctx->streams[stream_index]->time_base);
av_interleaved_write_frame(ofmt_ctx, &enc_pkt);
参考
- http://blog.csdn.net/leixiaohua1020/article/details/28114081
- http://www.cnblogs.com/yinxiangpei/articles/3892982.html
- http://blog.csdn.net/TopsLuo/article/details/76239136
- https://www.jianshu.com/p/bf323cee3b8e
- http://blog.csdn.net/leixiaohua1020/article/details/15811977
第三篇
ffmpeg 时间戳问题汇总
| 大家好
A:RFC3984 规定采用 90000 Hz 的时钟，因此如果编码帧频是 30，那么时间戳间隔就该是 90000 / 30 = 3000，根据抓包来看，似乎时间戳间隔的确是 3000。 时间戳的 间隔不固定，比如有的时间戳间隔是 2990 有的是 3002，会导致解析出来的视频快播的效果么 Q：各位大侠好： A；时间戳顺序累加就行了，每次加1 Q:最近做了一个捕捉摄像头并保存FLV的小东西，发现转换完毕后的FLV文件，用播放器播放的时候，速度特别快，大概是正常速度的4倍。请问这是怎么回事？网上搜了一下，说是时间戳的问题，可是PTS我跟了，AVPacket的PTS是每帧增长40，time_base为: 25/s.。DTS是个无效值。PTS的计算是根据ffmpeg的例子写的。 A:我也做摄像头捕捉，跟你出现一样的问题，我自己分析的话，应该是捕捉摄像头的图像的速度只有10帧每秒，但是保存成视频25帧每秒的话播放看起来就非常快，但是我摄像头捕捉设定的是25帧每秒，难道是速度达不到？ 谢谢。暂时认为是摄像头捕捉速率问题。换了一个高清无驱摄像头就好了 Q:在每个音视频数据包中都含有PTS和DTS，一个数据包中应该含有多个数据帧以及音频数据，那么这里的PTS和DTS它是如何来标识数据帧的？PTS和DTS的单位是什么？视频的最小单位是帧，可通过PTS来指定它何时播放，那音频的最小单位是什么？这里的PTS对音频而言它标识的是什么？是这个时间点采样点吗？ A: audio_timebase = av_q2d(fmtctx->streams[audio_index]->time_base); Q:谢谢kf701的回复，看后明白了不少 A:上面的代码是每display一个picture,就与audio的PTS比较一下, Q:如何利用AVPacket包里的pts,dts实现音视频同步？声频播放是只管自己播放，视频有一个初始化播放帧率，如何根据AVPacket里的pts,dts还实现两者的同步？ A:DTS：decoding time stamp Q:关于b帧和时间戳的问题 我从mpeg2视频中用av_read_frame()读取视频帧并解码，顺序是IPBBPBB... A:你不能这么做，编码器会给你这么做的。如果你有B帧，那么所有的B帧都会被放在缓冲区里直到下一个I/P帧到来 例如：你的输入序列是IBBPBBPBBI 那么输出的序列是 输入I，编码I，输出I 输入B 输入B 输入P，编码P，输出P 编码B，输出B 编码B，输出B 输入P,编码P，输出P 。。。。。。 在解码端所有的P帧都会被放在缓冲力直到下一个I/P真的到来 如：解码I，输出I 解码P,放入缓冲P 解码B，输出B 解码B，输出B 解码P，输出上一次P帧 Q:解码出来的图片的时间戳问题 MPEG一个包中包含有时间戳, 而可能几个包才能解码出一张图象, 也可能一个包能解码出几张图, 请问包中的时间戳与解码出来的图象如何对应上? A:在ffmpeg中通过parser部件把从avformat部件取下来的原始包重新“合成”为有仅包含一个完整帧的包。从MPEG2部份的代码中看出，如果“几个包才能解码出一张图象”的话，会取第一个包的PTS和DTS，如果“也可能一个包能解码出几张图”，则会跟据这个包的PTS和DTS通过帧频推算出其它帧的DTS。 Q: ffmpeg的avcodec_decode_video 函数解码时间戳问题？在 VLC 中调用 avcodec_decode_video() 函数进行解码时，AVFrame->pts 时间戳不对，导致我的图像不能够显示？ 请问有谁知道它的解码原理，这个 PTS 怎么得出的吗？还是外部传入的？ A: /* NOTE: ipts is the PTS of the _first_ picture beginning in if( (decoder_reorder_pts || pkt->dts == AV_NOPTS_VALUE) Q:我贴下 VLC 的代码，（vlc-0.9.8a/modules/codec/avcodec/video.c 文件中）
|
转自:http://hi.baidu.com/ilovejoy/blog/item/3da717eca51d7b302797917a.html/cmtid/a50c1c1837ec330935fa419b
8887
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
