Log in

View Full Version : Dirt after encode


ravil
25th October 2010, 07:42
Hi all, sorry for the (probably) lame question...
I have this video as source, screen:
http://i5.fastpic.ru/big/2010/1025/0a/263f266ba2dc0964bef363db6673500a.png

(screenshot from the .y4m)

And picture after encode looks like this:
http://i5.fastpic.ru/big/2010/1025/2f/d7b84c8d3889cd9e6ed4275c66e2052f.png

Obviously it's an encode problem, not source or filtering. Can you tell me, what can I do to prevent appearing this dirt?

Coded w/ crf=16, ~6mbps in summary.

x264 settings:
cabac=1 / ref=4 / deblock=1:-2:-2 / analyse=0x3:0x133 / me=umh / subme=9 / psy=1 / psy_rd=0.85:0.00 / mixed_ref=1 / me_range=24 / chroma_me=1 / trellis=2 / 8x8dct=1 / cqm=0 / deadzone=21,11 / fast_pskip=1 / chroma_qp_offset=-2 / threads=6 / sliced_threads=0 / nr=0 / decimate=0 / interlaced=0 / constrained_intra=0 / bframes=8 / b_pyramid=2 / b_adapt=2 / b_bias=0 / direct=3 / weightb=1 / open_gop=0 / weightp=2 / keyint=250 / keyint_min=23 / scenecut=40 / intra_refresh=0 / rc=2pass / mbtree=0 / bitrate=6010 / ratetol=1.0 / qcomp=0.70 / qpmin=9 / qpmax=26 / qpstep=4 / cplxblur=20.0 / qblur=0.5 / ip_ratio=1.40 / pb_ratio=1.30 / aq=1:0.90

nurbs
25th October 2010, 08:03
It would be interesting to know what the picture looks like after filtering, but before encoding. Apart from that why don't you use --tune animation? That would be --deblock 1:1 --psy-rd 0.4:0.0 --aq-strength 0.6 among other things.

Dark Shikari
25th October 2010, 08:09
Those aren't even the same frame.

ravil
25th October 2010, 08:42
Post updated, now frames are exactly the same and first screen is from the lossless source.

Didée
25th October 2010, 09:28
With a reasonable display setup, you would hardly see any problem there.

Many "problems" only arise because lots of today's typical TFTs are not capable of showing a true "black", but have only some medium/darkish gray instead.
Unreasonable gamma settings are another common problem.

The term "lossy compression" is for a reason. To get a perfectly artifact-free encode, use qp=0.

Blue_MiSfit
25th October 2010, 13:14
Yeah, I'm not seeing anything there worth complaining about, and that's even on a torched 42" LCD HDTV XD

Sharktooth
25th October 2010, 14:17
also colors are off... and however, the "dirt" is present in the source too.

ravil
25th October 2010, 14:38
Colors are off because I did screenshot from the lossless using AvsP (and used convertorgb() off course) so it's because of different colormatrix, colors from the .ts are identical to rip.

And I don't see dirt in the source.

kypec
25th October 2010, 14:42
Besides the slight colour shift (as Sharktooh has already mentioned) I do not see any "dirt" in the picture. What exactly are you referring to? Try to zoom the problematic area and mark it with big red circle or something like that. Otherwise we are all dirt-blind as it seems...

Sharktooth
25th October 2010, 14:56
i had to rise my monitor brightness to spot the "dirt"... and it's in the source too.

poisondeathray
25th October 2010, 15:29
^Yes, there's "dirt" in the source as well

Why wasn't the same method use for screenshots ?

Besides the rec601/rec709 mixup, there are different chroma upsampling algorithms being used. You can see how "blocky" or "stair steppy" the edges of red characters are

ravil
28th October 2010, 21:16
Another pair of screens: the first is filtered lossless, the second - crf15 ~8.5mbps 2-pass encode result of it:

http://i11.fastpic.ru/big/2010/1029/82/16a6b3d3f49929069188e2d26efe9482.png

http://i11.fastpic.ru/big/2010/1029/6d/55788d252dd8145ee2ab940c0840ef6d.png

Backside of the jacket on the chair is dirted by the x264 pretty well - are you sure I cannot do anything about it? Encode settings are the same as in the first message.

poisondeathray
28th October 2010, 21:41
The creases/folds in the black jacket seem to be obliterated as well

ravil
28th October 2010, 21:56
The creases/folds in the black jacket seem to be obliterated as well

You mean on the source picture too? It's a result of gradfun working, I just tried to remove some banding but was too lazy to make a scene filtering, so applied it on all clip... anyway it's not the point.

poisondeathray
28th October 2010, 21:59
You mean on the source picture too? It's a result of gradfun working, I just tried to remove some banding but was too lazy to make a scene filtering, so applied it on all clip... anyway it's not the point.


Not on the "filtered lossless" , on the encode. It looks almost as if you applied some filter after

I increased the gamma to make it more obvious (for both the dirt on chair, and jacket folds)

http://www.mediafire.com/?6tc9cfeawp4nx1v

poisondeathray
28th October 2010, 22:08
I edited my post above: I meant the encode, not the filtered lossless . It looks almost as if you added another filter to the "filtered lossless"

Did you try the tune animation settings suggested earlier?

ravil
28th October 2010, 22:15
I edited my post above: I meant the encode, not the filtered lossless . It looks almost as if you added another filter to the "filtered lossless"

Did you try the tune animation settings suggested earlier?

ohhhh crap, sorry man - I've uploaded the screenshot of the source BD by mistake ._. my bad, seems that I have to sleep more than 3 hours. I'll upload the picture of the filtered lossless later.