Improve how we handle denoising in multi-resolution encoding

[GoogleCodeExporter]

We should look into how the temporal denoiser is triggered when running in 
multi-resolution encoding mode and make sure we do the best thing there. First 
we should make sure the denoising code is actually triggered at all 
resolutions. We should also make sure good motion vectors are chosen for 
denoising.

Original issue reported on code.google.com by hol...@google.com on 11 Jun 2012 at 1:47

[GoogleCodeExporter]

Original comment by albe...@google.com on 19 Jul 2012 at 10:08

• Changed state: Assigned
• Added labels: Mstone-12

[GoogleCodeExporter]

If frame dropping is turned on in current multi-resolution encoding, we could 
end up with wrong motion vectors. The patch
 https://gerrit.chromium.org/gerrit/#change,24875,patchset=7
is to add frame dropping support in code, which will ensure the correct motion 
vectors are chosen. This should also make denoising work in multi-res encoding. 
I am testing it now.
When the denoising is turned on for all resolutions (noise-sensitivity is set 
to 1). I can see the encoding time increases by ~30%. Further optimization of 
the denoising part is needed, or there should be a strategy to decide when to 
turn on/off denoising. What do you think?

Original comment by yunqingw...@google.com on 19 Jul 2012 at 11:25

• Changed state: Started