topic Re: I,P, and B Frame VC1Encoder and VC1Decoder example(s) in Intel® Integrated Performance Primitives

I,P, and B Frame VC1Encoder and VC1Decoder example(s)

Dennis_W — Thu, 01 Oct 2009 14:38:55 GMT

Is there any example code of how to encode and decode a set of frames as a block set up to use a non-zero number of B Frames, preferably with the VC1Encoder and VC1Decoder, but an example with any of the UMC encoder/decoder classes would be helpful?

Thanks
Dennis

Re: I,P, and B Frame VC1Encoder and VC1Decoder example(s)

oxydius — Thu, 01 Oct 2009 17:58:33 GMT

Quoting - Dennis W

Something like this would work :

UMC::VC1EncoderParams videoParams;
videoParams.profile = UMC_VC1_ENCODER::VC1_ENC_PROFILE_A;
videoParams.info.clip_info.width = 1920;
videoParams.info.clip_info.height = 1080;
videoParams.info.color_format = UMC::YUV420;
videoParams.m_uiGOPLength = 12;
videoParams.m_uiBFrmLength = n;

UMC::VC1VideoEncoder encoder;
encoder.Init(&videoParams);

Re: I,P, and B Frame VC1Encoder and VC1Decoder example(s)

Dennis_W — Thu, 01 Oct 2009 18:05:27 GMT

I understand the basic settings of the parameters. I was looking for an broader example of the flow of calls to the encoder and decoder GetFrame methods

Re: I,P, and B Frame VC1Encoder and VC1Decoder example(s)

oxydius — Thu, 01 Oct 2009 21:37:33 GMT

After that encoder Init call, you must do the following :

UMC::VideoData input;
input.Init(1920, 1080, 3, 8);
input.SetColorFormat(UMC::YUV420);
...
input.SetPlanePointer && input.SetPlanePitch
...

UMC::MediaData vc1;

UMC::Status status = encoder.GetFrame(&input, &vc1);

status may be UMC::UMC_ERR_NOT_ENOUGH_BUFFER if you didn't make vc1 large enough.
it may be UMC::UMC_OK if there's an output frame in the vc1 buffer.
otherwise, it may be UMC::UMC_ERR_NOT_ENOUGH_DATA if it needs more input frames before outputting your B frames.
At the end of the encode process, you'll need to call encoder.GetFrame(0, &vc1) to flush such pending frames.

Re: I,P, and B Frame VC1Encoder and VC1Decoder example(s)

Dennis_W — Sat, 03 Oct 2009 18:02:30 GMT

Quoting - oxydius

After that encoder Init call, you must do the following : ...

I have the encoder working and it basically works with the VC1Decoder when I couple it with a splitter.

I set the decoder params to basically the same params as the encoder including the advanced profile.

The problem with the splitter is that basically there is only a FileReader that comes as is with Intel samples.
Basically what I want to do issimulate where I use GOP and BFrames but I want to "stream' the frames to the decoder as they become available from the decoder. For example, I may have a 100 frame clip but I want to process it in groups of 10 frames with 5 BFrames but I want to "stream" individual frames to the decoder as they are available from the encoder, and as frames are available from the decoder, render them.

The issue that I ran into is that a FileWriter with the encoder and a FileReader with the decoder splitter cannot write and read in tandem so I have to encode in blocks of 10 and then decode and render in blocks of 10.

For the long term I can develop a StreamWriter and StreamReader, but the for the short term I would like to attempt to use the existing UMC classes. Is it necessary to use a VC1Splitter with the VC1Decoder in advanced profile? I get an exception when I tried. Can I decode each individual output from the encoder when using GOP and BFrames and how do I do it?

Thanks

Dennis

Re: I,P, and B Frame VC1Encoder and VC1Decoder example(s)

oxydius — Tue, 06 Oct 2009 06:52:04 GMT

Each UMC::MediaData element is a standalone frame which can be passed to the decoder as is. Basically, if you refer to the VC-1 standard, they are raw BDU's, or Bitstream Data Units with their own header.

You could do :
encoder.GetFrame(&in, &out);
decoder.GetFrame(&out, &in);

without ever using the splitter or file I/O classes.

Re: I,P, and B Frame VC1Encoder and VC1Decoder example(s)

Dennis_W — Tue, 06 Oct 2009 17:22:35 GMT

Quoting - oxydius

Will that work in the advanced profile, I thought that in that case the decoder used the splitter to possibly reconstruct frames from more that one MediaData instance? But I will try it.

Re: I,P, and B Frame VC1Encoder and VC1Decoder example(s)

oxydius — Tue, 06 Oct 2009 20:34:40 GMT

Quoting - Dennis W

Will that work in the advanced profile, I thought that in that case the decoder used the splitter to possibly reconstruct frames from more that one MediaData instance? But I will try it.

Advanced profile is the VC-1 standard; other profiles are actually WMV and do not lend themselves to being used outside a proprietary ASF container. In the advanced/VC-1 mode, each MediaData has its own 001 delimiter and frame header, telling you whether it's a I, P or B frame.

The only reconstruction done by the decoder is holding on to B frames until the subsequent MediaData is fed to transform the B frame into a raw YUV image.

Re: I,P, and B Frame VC1Encoder and VC1Decoder example(s)

Dennis_W — Wed, 07 Oct 2009 00:23:38 GMT

Quoting - oxydius

The only reconstruction done by the decoder is holding on to B frames until the subsequent MediaData is fed to transform the B frame into a raw YUV image.

When I tried to pass the MediaData output from theVC1Encoder::GetFrame to the VC1Decoder::GetFrame only UMC_ERR_NOT_ENOUGH_DATA was returned, never UMC::UMC_OK. I initialized the encoder and decoder with the same parameters when I was trying the splitter which worked.

Re: I,P, and B Frame VC1Encoder and VC1Decoder example(s)

Dennis_W — Wed, 07 Oct 2009 02:16:10 GMT

Quoting - Dennis W

I solved that with some code changes, and I get through the clip, however, I am getting some flickering from frame to frame which varies on the value of GOPLength and BFrmLength. This flickers between complete frames, blocky frames and purple background. It appears I have something out of sync in my code.

Re: I,P, and B Frame VC1Encoder and VC1Decoder example(s)

Dennis_W — Wed, 07 Oct 2009 03:57:49 GMT

Quoting - Dennis W

It appears that the flickering problem goes away when the BFrmLength is one less than GOPLength, say GOPLength=4 and BFrmLength=3. I need to read the VC-1 specification to become more familiar with the options.