背景
最近用ffmpeg对mp4文件进行转码开发,无论是用ffmpeg命令,亦或是ffmpeg所提供的api,转码后的mp4文件,通过Elecard StreamEye系列工具查看,在首个IDR帧前用SEI携带了x264的版本信息。
但是,如果携带了x264的版本信息,对于mp4的首开延迟有一定影响,因为加载了多余的数据。
此处需要优化掉。
问题
用ffmpeg api转码后的mp4文件,用Elecard StreamEye查看,如下:
用StreamEye4工具查看:
以上工具均显示用ffmpeg编码后,在首个IDR帧前用SEI携带了x264的版本信息。
解决方案
回过头来,翻阅x264的源码,发现在编码时的确强制写入了sei,携带了x264的版本信息:
static int encode( x264_param_t *param, cli_opt_t *opt )
{
x264_t *h = NULL;
x264_picture_t pic;
cli_pic_t cli_pic;
const cli_pulldown_t *pulldown = NULL; // shut up gcc
int i_frame = 0;
int i_frame_output = 0;
int64_t i_end, i_previous = 0, i_start = 0;
int64_t i_file = 0;
int i_frame_size;
int64_t last_dts = 0;
int64_t prev_dts = 0;
int64_t first_dts = 0;
# define MAX_PTS_WARNING 3 /* arbitrary */
int pts_warning_cnt = 0;
int64_t largest_pts = -1;
int64_t second_largest_pts = -1;
int64_t ticks_per_frame;
double duration;
double pulldown_pts = 0;
int retval = 0;
opt->b_progress &= param->i_log_level < X264_LOG_DEBUG;
/* set up pulldown */
if( opt->i_pulldown && !param->b_vfr_input )
{
param->b_pulldown = 1;
param->b_pic_struct = 1;
pulldown = &pulldown_values[opt->i_pulldown];
param->i_timebase_num = param->i_fps_den;
FAIL_IF_ERROR2( fmod( param->i_fps_num * pulldown->fps_factor, 1 ),
"unsupported framerate for chosen pulldown\n" );
param->i_timebase_den = param->i_fps_num * pulldown->fps_factor;
}
h = x264_encoder_open( param );
FAIL_IF_ERROR2( !h, "x264_encoder_open failed\n" );
x264_encoder_parameters( h, param );
FAIL_IF_ERROR2( cli_output.set_param( opt->hout, param ), "can't set outfile param\n" );
i_start = x264_mdate();
/* ticks/frame = ticks/second / frames/second */
ticks_per_frame = (int64_t)param->i_timebase_den * param->i_fps_den / param->i_timebase_num / param->i_fps_num;
FAIL_IF_ERROR2( ticks_per_frame < 1 && !param->b_vfr_input, "ticks_per_frame invalid: %"PRId64"\n", ticks_per_frame );
ticks_per_frame = X264_MAX( ticks_per_frame, 1 );
if( !param->b_repeat_headers )
{
// Write SPS/PPS/SEI
x264_nal_t *headers;
int i_nal;
FAIL_IF_ERROR2( x264_encoder_headers( h, &headers, &i_nal ) < 0, "x264_encoder_headers failed\n" );
FAIL_IF_ERROR2( (i_file = cli_output.write_headers( opt->hout, headers )) < 0, "error writing headers to output file\n" );
}
if( opt->tcfile_out )
fprintf( opt->tcfile_out, "# timecode format v2\n" );
/* Encode frames */
for( ; !b_ctrl_c && (i_frame < param->i_frame_total || !param->i_frame_total); i_frame++ )
{
if( filter.get_frame( opt->hin, &cli_pic, i_frame + opt->i_seek ) )
break;
x264_picture_init( &pic );
convert_cli_to_lib_pic( &pic, &cli_pic );
if( !param->b_vfr_input )
pic.i_pts = i_frame;
if( opt->i_pulldown && !param->b_vfr_input )
{
pic.i_pic_struct = pulldown->pattern[ i_frame % pulldown->mod ];
pic.i_pts = (int64_t)( pulldown_pts + 0.5 );
pulldown_pts += pulldown_frame_duration[pic.i_pic_struct];
}
else if( opt->timebase_convert_multiplier )
pic.i_pts = (int64_t)( pic.i_pts * opt->timebase_convert_multiplier + 0.5 );
if( pic.i_pts <= largest_pts )
{
if( cli_log_level >= X264_LOG_DEBUG || pts_warning_cnt < MAX_PTS_WARNING )
x264_cli_log( "x264", X264_LOG_WARNING, "non-strictly-monotonic pts at frame %d (%"PRId64" <= %"PRId64")\n",
i_frame, pic.i_pts, largest_pts );
else if( pts_warning_cnt == MAX_PTS_WARNING )
x264_cli_log( "x264", X264_LOG_WARNING, "too many nonmonotonic pts warnings, suppressing further ones\n" );
pts_warning_cnt++;
pic.i_pts = largest_pts + ticks_per_frame;
}
second_largest_pts = largest_pts;
largest_pts = pic.i_pts;
if( opt->tcfile_out )
fprintf( opt->tcfile_out, "%.6f\n", pic.i_pts * ((double)param->i_timebase_num / param->i_timebase_den) * 1e3 );
if( opt->qpfile )
parse_qpfile( opt, &pic, i_frame + opt->i_seek );
prev_dts = last_dts;
i_frame_size = encode_frame( h, opt->hout, &pic, &last_dts );
if( i_frame_size < 0 )
{
b_ctrl_c = 1; /* lie to exit the loop */
retval = -1;
}
else if( i_frame_size )
{
i_file += i_frame_size;
i_frame_output++;
if( i_frame_output == 1 )
first_dts = prev_dts = last_dts;
}
if( filter.release_frame( opt->hin, &cli_pic, i_frame + opt->i_seek ) )
break;
/* update status line (up to 1000 times per input file) */
if( opt->b_progress && i_frame_output )
i_previous = print_status( i_start, i_previous, i_frame_output, param->i_frame_total, i_file, param, 2 * last_dts - prev_dts - first_dts );
}
/* Flush delayed frames */
while( !b_ctrl_c && x264_encoder_delayed_frames( h ) )
{
prev_dts = last_dts;
i_frame_size = encode_frame( h, opt->hout, NULL, &last_dts );
if( i_frame_size < 0 )
{
b_ctrl_c = 1; /* lie to exit the loop */
retval = -1;
}
else if( i_frame_size )
{
i_file += i_frame_size;
i_frame_output++;
if( i_frame_output == 1 )
first_dts = prev_dts = last_dts;
}
if( opt->b_progress && i_frame_output )
i_previous = print_status( i_start, i_previous, i_frame_output, param->i_frame_total, i_file, param, 2 * last_dts - prev_dts - first_dts );
}
fail:
if( pts_warning_cnt >= MAX_PTS_WARNING && cli_log_level < X264_LOG_DEBUG )
x264_cli_log( "x264", X264_LOG_WARNING, "%d suppressed nonmonotonic pts warnings\n", pts_warning_cnt-MAX_PTS_WARNING );
/* duration algorithm fails when only 1 frame is output */
if( i_frame_output == 1 )
duration = (double)param->i_fps_den / param->i_fps_num;
else if( b_ctrl_c )
duration = (double)(2 * last_dts - prev_dts - first_dts) * param->i_timebase_num / param->i_timebase_den;
else
duration = (double)(2 * largest_pts - second_largest_pts) * param->i_timebase_num / param->i_timebase_den;
i_end = x264_mdate();
/* Erase progress indicator before printing encoding stats. */
if( opt->b_progress )
fprintf( stderr, " \r" );
if( h )
x264_encoder_close( h );
fprintf( stderr, "\n" );
if( b_ctrl_c )
fprintf( stderr, "aborted at input frame %d, output frame %d\n", opt->i_seek + i_frame, i_frame_output );
cli_output.close_file( opt->hout, largest_pts, second_largest_pts );
opt->hout = NULL;
if( i_frame_output > 0 )
{
double fps = (double)i_frame_output * (double)1000000 /
(double)( i_end - i_start );
fprintf( stderr, "encoded %d frames, %.2f fps, %.2f kb/s\n", i_frame_output, fps,
(double) i_file * 8 / ( 1000 * duration ) );
}
return retval;
}
在x264.c文件static int encode( x264_param_t *param, cli_opt_t *opt )方法中,调用x264_encoder_headers方法写入sps/pps/sei:
if( !param->b_repeat_headers )
{
// Write SPS/PPS/SEI
x264_nal_t *headers;
int i_nal;
FAIL_IF_ERROR2( x264_encoder_headers( h, &headers, &i_nal ) < 0, "x264_encoder_headers failed\n" );
FAIL_IF_ERROR2( (i_file = cli_output.write_headers( opt->hout, headers )) < 0, "error writing headers to output file\n" );
}
接下来再看 x264_encoder_headers方法源码:
/****************************************************************************
* x264_encoder_headers:
****************************************************************************/
int x264_encoder_headers( x264_t *h, x264_nal_t **pp_nal, int *pi_nal )
{
int frame_size = 0;
/* init bitstream context */
h->out.i_nal = 0;
bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
/* Write SEI, SPS and PPS. */
/* generate sequence parameters */
nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
x264_sps_write( &h->out.bs, h->sps );
if( nal_end( h ) )
return -1;
/* generate picture parameters */
nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
x264_pps_write( &h->out.bs, h->sps, h->pps );
if( nal_end( h ) )
return -1;
/* identify ourselves */
nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
if( x264_sei_version_write( h, &h->out.bs ) )
return -1;
if( nal_end( h ) )
return -1;
frame_size = encoder_encapsulate_nals( h, 0 );
if( frame_size < 0 )
return -1;
/* now set output*/
*pi_nal = h->out.i_nal;
*pp_nal = &h->out.nal[0];
h->out.i_nal = 0;
return frame_size;
}
可以看到, x264_encoder_headers方法在写入了sps/pps后,调用x264_sei_version_write方法强行用sei携带了x264的版本信息。
分析到此,原因已很明了了,解决办法也很简单,不调用x264_sei_version_write方法,不用sei携带x264的版本信息即可。
经验证,去掉x264的版本信息,是不会影响播放的,满足需求。