[Opensvcdecoder-support] help-Some questions about the error concealment

[Yippee L. <yip...@gm...>]

Hi there,

First really thank you for the great staff you've done.

Now I have some questions about the error concealment in the opensvcdecoder.

It seems that this software has a ec mode, but not work very well. It can
only make the decoding continue but can not recover the loss frames.

I use the loss simulator in JVT-Q069, and make some stream with the loss
rate of 20% in enhance layer,
then I use the opensvcdecoder to write(yes, I also write a short code to
make opensvcdecoder output the yuv) a yuv file.

But the result is that, I lose x frames after loss simulation(suppose x is
20), then the reconstructed stream also loses 20 frames.

Besides that, the quality of the reconstructed file is not very good, or
very bad.

Now comes my questions:
1) I hope to know if you want to make better error control in the future
version?
2) I want to know if I can put the only three error concealment methods to
the opensvcdecoder(from JSVM 9.8, Frame Copy, Temperal Direct, BaseLayer
Skip), I've seen many papers, and all these three are almost the only
methods to deal with the svc frame loss. But it's not an easy job, you know
the JSVM is written in C++, with many new C++ feature than C. So if I want
to do some transplant, I guess it's really a difficult job.

3) I want to know whether the error concealment is worthy as a research
point or a project part. I've asked many people doing the coding job, they
told me that in the real project, if the decoder encounter an error, they
just leave it there and waiting for the next IDR frame. So I want to make
sure the error concealment in the decoder is worthy or not.


Thank you for your help.

Yates, China
June-2012