topic UMC H.264 decoder in Intel® Integrated Performance Primitives

UMC H.264 decoder

Purvin_Pandit — Wed, 09 Jun 2010 15:09:33 GMT

I have a H.264 bitstream (fields only) with intra refresh (no I/IDR only P fields) + recovery point SEI. The bitstream does not start at frame_num = 0. It starts at frame_num = 201, now after the frame_num wraps and the next time a frame with frame_num = 201 arrive the old frame_num = 201 is still in the Decoded list. Not sure why it is not removed.

This bitstream can be decoded by JM and other decoders. Is there a known issue when a bitstream does not begin with frame_num = 0, has recovery point SEI?

Thanks,

-Purvin

UMC H.264 decoder

Purvin_Pandit — Wed, 09 Jun 2010 18:39:13 GMT

Digging a little more I found this code in:

bool TaskSupplier::AddSlice(H264MemoryPiece * pMem, MediaData * pSource)

if (NumShortTermRefs + NumLongTermRefs == 0)

{

....

.....

if (sliceHeader->field_pic_flag)

{

pFrame->setPicOrderCnt(m_PicOrderCnt,0);

pFrame->setPicOrderCnt(m_PicOrderCnt,1);

}

else

{

pFrame->setPicOrderCnt(m_TopFieldPOC, 0);

pFrame->setPicOrderCnt(m_BottomFieldPOC, 1);

}

Looks like this is done only for the 1st field (in my case) so it sets the top & bottom POC to be the same. This is causing the problem where the bottom field is never removed. This looks like a bug. Can someone confirm?

UMC H.264 decoder

Chao_Y_Intel — Sat, 12 Jun 2010 05:01:27 GMT

Purvin,

Would it be possible to attach the test stream? Our expert comments that this is not very clear on what is happeing from the description.

Thanks,
Chao

UMC H.264 decoder

Purvin_Pandit — Mon, 14 Jun 2010 13:07:21 GMT

Thanks Chao. I will try to get a bitstream to you as soon as I can.

-Purvin