[FFmpeg-devel] AVCHD/H.264 decoder: further development/corrections

[Ivan Schreter]

Ivan Schreter wrote:
> Michael Niedermayer wrote:
>   
>> you might want to look at
>> svn di -r12162:12161
>>   
>>     
> I will. But the problem is, even if av_read_frame() returns proper 
> timestamps for field pictures (interlaced video coded as sequence of 
> field pictures as opposed to interlaced frames), it doesn't fulfill its 
> contract of returning whole frames to the caller. I.e., the caller still 
> has to call av_read_frame() twice to get one interlaced frame.
>   
I did now. Any explanation, why it was removed? It looks like it was 
supposed to handle interlaced frames.

Anyway, I suppose, the H.264 parser should simply fill in missing 
timestamps (if possible) and don't let lavf do it for it. I'll look into it.

Independent of that, it seems like not only H.264 codec doesn't return 
whole frames from av_read_frame(). Maybe is the description of 
av_read_frame() incorrect and it should say it returns buffer with a 
coded picture containing either a complete frame or a complete field?

I now checked in MLT, how they use lavf. It seems like they handle the 
case of field pictures correctly. So adding proper timestamps and key 
flags should work, collation of fields in parser would be then unnecessary.

Regards,

Ivan