I, P and B frames in MediaSDK h264 encoder are controlled with the following two parameters passed to encoder Init method (from MediaSDK manual):
GopPicSize
Number
of pictures within the current GOP (Group of Pictures); if GopPicSize=0,
then the GOP size is unspecified. If GopPicSize=1,
only I-frames are used. See Example 12 for pseudo-code that demonstrates how SDK uses this
parameter.
GopRefDist
Distance
between I- or P- key frames; if it is zero, the GOP structure is unspecified.
Note: If GopRefDist
= 1, there are no B-frames used. See Example 12 for pseudo-code that demonstrates how SDK uses this
parameter.
So for example if you specify GopRefDist = 1 you get only I and P frames. It is always better to specify reasonable values for these two parameters because you never know what the default values are. For example if you want two B frames between P you need to specify GopRefDist = 3.
H264 encoder must use B frames ?
I've built a transcoder (mpeg2/mp3 -> h264/aac) using microsoft direct show filters for the decoding, and the intel sdk direct show filters for the encoder.
The encoded video and audio, are then muxed (using intel mp2_mux filter), and I get a beautiful ts file.
I've tried in both ways: either through C++ (yes, it takes some effort
to build), either through graphedit. The flow is : TS file, Microsoft
MPEG2 demux => Microsoft Mp3 Audio + Microsoft MPEG2 Video decoders
=> Intel AAC + Intel H264 encoder => Intel MPEG2 mux => file
write.
All is good, and nice. VLC can play the resulted ts file, everythings is ok.
However, Windows Media Player no, unless I use B frames for the h264 encoder (only audio, and black video).
I've tried all sort of configurations (I have some experience with transcoders and video standards), different profiles, slices, levels, etc etc. The only common parameter making the Windows Media Player capable to play the obtained TS file, is to use B frames. Unfortunately, this is out of spec for Baseline.
Also, the FFMPEG TS segmenter complains about the TS when not using B frames.
So, it might be either some problems with timestamps into the encoder, or into the mp2 muxer, or I'm doing something wrong.
I use the latest 1.6 Gold edition of media SDK.
So, the questions are:
Is this a known problem? Are there some magic parameters to give to the h264 encoder direct show filter (with B frames zero) that will make Windows Media Player support it?
Thanks.