Could this be deblocked?

[jeffy]

A small 2MB sample, can anyone please help me to deblock the fire or is this just impossible?
http://www.mediafire.com/?sqimzj9ymmw

:thanks:

[Dark Shikari]

I only see blocking in a single frame (frame 15)...

[jeffy]

frame 15 (VirtualDub 1.8.0): slight blocking at the top, frame 16 (the trees and the fire) full of blocks, frame 18 (blocking in the fire), frame 19 the same (zoom: 150%).

script:
MPEG2Source("example_fire.demuxed.d2v", cpu=0)

Is there something that could help? Thank you.

[thetoof]

deblock_qed

[jeffy]

deblock_qed

Have you really tried it? If so, could you please post the screenshot?

[Blue_MiSfit]

erm... why don't you try it?

It's a simple script, and is very effective..

~MiSfit

[Didée]

... and just in case you happen to actually try it, a suggestion for starting out:

Mpeg2Source("example_fire.demuxed.d2v")

# when activated, it clears the lacing artefacts:
# vinverse(sstr=2.0,amnt=7)

# deblocking seems quite okay like so:
deblock_qed(quant1=24,quant2=24,aOff1=16,bOff1=20,aOff2=20,bOff2=24)

[jeffy]

Didée, thank you very much, I did tried it, but I replaced deblock_qed above with deblock_qed_mt2(quant1=24,quant2=24,aOff1=16,bOff1=20,aOff2=20,bOff2=24)

This is the result, is this the max?
http://img261.imageshack.us/img261/5553/stillblockyhm6.jpg

[Didée]

No. The max is with quant1/2=51, even 40 should suffice. But the blocking in that frame is very extreme, and you probably don't want such aggressive settings throughout the source. Overall, IMO you're better off with the suggested settings.
Consider producing a 2nd clip with extreme settings, and fiddle it in manually with trim()+trim()... or stickboy's ReplaceFramesSimple() for the few problematic frames.

[jeffy]

The trim idea looks reasonable. I'm sorry I didn't think of raising the quantizers myself. 40/40 for that 1 frame looks IMHO so-so, so I am considering 24/24, as suggested.

Thank you very much!