原帖地址:http://blog.csdn.net/leixiaohua1020/article/details/12678577
ffmpeg中的av_read_frame()的作用是读取码流中的音频若干帧或者视频一帧。例如,解码视频的时候,每解码一个视频帧,需要先调用 av_read_frame()获得一帧视频的压缩数据,然后才能对该数据进行解码(例如H.264中一帧压缩数据通常对应一个NAL)。
对该函数源代码的分析是很久之前做的了,现在翻出来,用博客记录一下。
上代码之前,先参考了其他人对av_read_frame()的解释,在此做一个参考:
通过av_read_packet(***),读取一个包,需要说明的是此函数必须是包含整数帧的,不存在半帧的情况,以ts流为例,是读取一个完整的PES包(一个完整pes包包含若干视频或音频es包),读取完毕后,通过av_parser_parse2(***)分析出视频一帧(或音频若干帧),返回,下次进入循环的时候,如果上次的数据没有完全取完,则st = s->cur_st;不会是NULL,即再此进入av_parser_parse2(***)流程,而不是下面的av_read_packet(**)流程,这样就保证了,如果读取一次包含了N帧视频数据(以视频为例),则调用av_read_frame(***)N次都不会去读数据,而是返回第一次读取的数据,直到全部解析完毕。
av_read_frame()的源代码如下:
int av_read_frame(AVFormatContext *s, AVPacket *pkt)
{
const int genpts = s->flags & AVFMT_FLAG_GENPTS;
int eof = 0;
int ret;
AVStream *st;
if (!genpts) {
ret = s->packet_buffer ?
read_from_packet_buffer(&s->packet_buffer, &s->packet_buffer_end, pkt) :
read_frame_internal(s, pkt);
if (ret < 0)
return ret;
goto return_packet;
}
for (;;) {
AVPacketList *pktl = s->packet_buffer;
if (pktl) {
AVPacket *next_pkt = &pktl->pkt;
if (next_pkt->dts != AV_NOPTS_VALUE) {
int wrap_bits = s->streams[next_pkt->stream_index]->pts_wrap_bits;
// last dts seen for this stream. if any of packets following
// current one had no dts, we will set this to AV_NOPTS_VALUE.
int64_t last_dts = next_pkt->dts;
while (pktl && next_pkt->pts == AV_NOPTS_VALUE) {
if (pktl->pkt.stream_index == next_pkt->stream_index &&
(av_compare_mod(next_pkt->dts, pktl->pkt.dts, 2LL << (wrap_bits - 1)) < 0)) {
if (av_compare_mod(pktl->pkt.pts, pktl->pkt.dts, 2LL << (wrap_bits - 1))) { //not b frame
next_pkt->pts = pktl->pkt.dts;
}
if (last_dts != AV_NOPTS_VALUE) {
// Once last dts was set to AV_NOPTS_VALUE, we don't change it.
last_dts = pktl->pkt.dts;
}
}
pktl = pktl->next;
}
if (eof && next_pkt->pts == AV_NOPTS_VALUE && last_dts != AV_NOPTS_VALUE) {
// Fixing the last reference frame had none pts issue (For MXF etc).
// We only do this when
// 1. eof.
// 2. we are not able to resolve a pts value for current packet.
// 3. the packets for this stream at the end of the files had valid dts.
next_pkt->pts = last_dts + next_pkt->duration;
}
pktl = s->packet_buffer;
}
/* read packet from packet buffer, if there is data */
if (!(next_pkt->pts == AV_NOPTS_VALUE &&
next_pkt->dts != AV_NOPTS_VALUE && !eof)) {
ret = read_from_packet_buffer(&s->packet_buffer,
&s->packet_buffer_end, pkt);
goto return_packet;
}
}
ret = read_frame_internal(s, pkt);
if (ret < 0) {
if (pktl && ret != AVERROR(EAGAIN)) {
eof = 1;
continue;
} else
return ret;
}
if (av_dup_packet(add_to_pktbuf(&s->packet_buffer, pkt,
&s->packet_buffer_end)) < 0)
return AVERROR(ENOMEM);
}
return_packet:
st = s->streams[pkt->stream_index];
if (st->skip_samples) {
uint8_t *p = av_packet_new_side_data(pkt, AV_PKT_DATA_SKIP_SAMPLES, 10);
if (p) {
AV_WL32(p, st->skip_samples);
av_log(s, AV_LOG_DEBUG, "demuxer injecting skip %d\n", st->skip_samples);
}
st->skip_samples = 0;
}
if ((s->iformat->flags & AVFMT_GENERIC_INDEX) && pkt->flags & AV_PKT_FLAG_KEY) {
ff_reduce_index(s, st->index);
av_add_index_entry(st, pkt->pos, pkt->dts, 0, 0, AVINDEX_KEYFRAME);
}
if (is_relative(pkt->dts))
pkt->dts -= RELATIVE_TS_BASE;
if (is_relative(pkt->pts))
pkt->pts -= RELATIVE_TS_BASE;
return ret;
}
一些说明:
1、新版本的ffmpeg用的是av_read_frame,而老版本的是av_read_packet。区别是av_read_packet读出的是包,它可能是半帧或多帧,不保证帧的完整性。av_read_frame对av_read_packet进行了封装,使读出的数据总是完整的帧。
2、一般情况下,av_read_frame会调用read_frame_internal(),其代码如下所示:
static int read_frame_internal(AVFormatContext *s, AVPacket *pkt)
{
int ret = 0, i, got_packet = 0;
av_init_packet(pkt);
while (!got_packet && !s->parse_queue)
{
AVStream *st;
AVPacket cur_pkt;
/* read next packet */
ret = ff_read_packet(s, &cur_pkt);
if (ret < 0)
{
if (ret == AVERROR(EAGAIN))
return ret;
/* flush the parsers */
for(i = 0; i < s->nb_streams; i++)
{
st = s->streams[i];
if (st->parser && st->need_parsing)
parse_packet(s, NULL, st->index);
}
/* all remaining packets are now in parse_queue =>
* really terminate parsing */
break;
}
ret = 0;
st = s->streams[cur_pkt.stream_index];
if (cur_pkt.pts != AV_NOPTS_VALUE &&
cur_pkt.dts != AV_NOPTS_VALUE &&
cur_pkt.pts < cur_pkt.dts)
{
av_log(s, AV_LOG_WARNING, "Invalid timestamps stream=%d, pts=%s, dts=%s, size=%d\n",
cur_pkt.stream_index,
av_ts2str(cur_pkt.pts),
av_ts2str(cur_pkt.dts),
cur_pkt.size);
}
if (s->debug & FF_FDEBUG_TS)
av_log(s, AV_LOG_DEBUG, "ff_read_packet stream=%d, pts=%s, dts=%s, size=%d, duration=%d, flags=%d\n",
cur_pkt.stream_index,
av_ts2str(cur_pkt.pts),
av_ts2str(cur_pkt.dts),
cur_pkt.size,
cur_pkt.duration,
cur_pkt.flags);
if (st->need_parsing && !st->parser && !(s->flags & AVFMT_FLAG_NOPARSE))
{
st->parser = av_parser_init(st->codec->codec_id);
if (!st->parser)
{
av_log(s, AV_LOG_VERBOSE, "parser not found for codec "
"%s, packets or times may be invalid.\n",
avcodec_get_name(st->codec->codec_id));
/* no parser available: just output the raw packets */
st->need_parsing = AVSTREAM_PARSE_NONE;
} else if(st->need_parsing == AVSTREAM_PARSE_HEADERS)
{
st->parser->flags |= PARSER_FLAG_COMPLETE_FRAMES;
} else if(st->need_parsing == AVSTREAM_PARSE_FULL_ONCE)
{
st->parser->flags |= PARSER_FLAG_ONCE;
} else if(st->need_parsing == AVSTREAM_PARSE_FULL_RAW)
{
st->parser->flags |= PARSER_FLAG_USE_CODEC_TS;
}
}
if (!st->need_parsing || !st->parser)
{
/* no parsing needed: we just output the packet as is */
*pkt = cur_pkt;
compute_pkt_fields(s, st, NULL, pkt);
if ((s->iformat->flags & AVFMT_GENERIC_INDEX) &&
(pkt->flags & AV_PKT_FLAG_KEY) && pkt->dts != AV_NOPTS_VALUE)
{
ff_reduce_index(s, st->index);
av_add_index_entry(st, pkt->pos, pkt->dts, 0, 0, AVINDEX_KEYFRAME);
}
got_packet = 1;
} else if (st->discard < AVDISCARD_ALL)
{
if ((ret = parse_packet(s, &cur_pkt, cur_pkt.stream_index)) < 0)
return ret;
} else
{
/* free packet */
av_free_packet(&cur_pkt);
}
if (pkt->flags & AV_PKT_FLAG_KEY)
st->skip_to_keyframe = 0;
if (st->skip_to_keyframe)
{
av_free_packet(&cur_pkt);
if (got_packet) {
*pkt = cur_pkt;
}
got_packet = 0;
}
}
if (!got_packet && s->parse_queue)
ret = read_from_packet_buffer(&s->parse_queue, &s->parse_queue_end, pkt);
if(s->debug & FF_FDEBUG_TS)
av_log(s, AV_LOG_DEBUG, "read_frame_internal stream=%d, pts=%s, dts=%s, size=%d, duration=%d, flags=%d\n",
pkt->stream_index,
av_ts2str(pkt->pts),
av_ts2str(pkt->dts),
pkt->size,
pkt->duration,
pkt->flags);
return ret;
}
更新:ffmpeg 2.1中,原文中提到的
av_paser_parse2()已经被舍弃,新采用的方法随着学习的深入慢慢分析。