DECODE and VPP interaction

DECODE and VPP interaction

Hi,

I am a bit puzzled by interactions between DECODE and VIDEO parts in the Intel Media SDK. I manage to decode properly a video stream, and I would like to add deinterlacing support.

To decode, I call these functions :

MFXVideoDECODE_DecodeHeader

MFXVideoDECODE_Query

MFXVideoDECODE_QueryIOSurf

MFXVideoDECODE_Init

I try progressively to configure vpp in and out params. Just copying mfx.FrameInfo structure into vpp.In and vpp.Out in video params makes MFXVideoDECODE_Query fail (MFX_ERR_UNSUPPORTED). Am I supposed to handle VPP initializations after intializing DECODE structures ? Is there a specific order ? Why does the decoder look in vpp parts of the parameters structure ?

Thanks,

2 帖子 / 0 全新
最新文章
如需更全面地了解编译器优化,请参阅优化注意事项

You'll need to setup some of the output fields, sometimes the input fields as well.
At a minimum, the PicStruct of the output frame should be progressive, this triggers deinterlacing if the input is marked as interlaced.
You can set the output frame rate to double that of the source in order to receive all the frames from the deinterlacer (useful for high quality playback, less for transcoding).
Note that soft telecine is not supported and you'll have to do it yourself.

I've implemented VPP into my own open source project (Intel QuickSync Decoder). You can use the source code for reference on how to use VPP and you'll might want to look at the decode code as well. You can download or view sources from http://sourceforge.net/projects/qsdecoder/
Both FFDShow and LAV filters use this decoder internally so you can use them as well.

Eric Gur,
Intel QuickSync Decoder Author
Processor Client Application Engineer
Intel Corp.

发表评论

登录添加评论。还不是成员?立即加入