PDA

View Full Version : Parameters for banding elimination?


martinfrombern
30th April 2009, 21:37
it seems that which ever settings i choose, i cannot remove this ugly banding. i know it's a high motion scene, and that it wouldn't be near this obvious when playing.
however, i would still like to reduce it as much as possible. any particular settings you can suggest?

i used this command line for the screenshot below

x264.exe --pass 2 --bitrate 1750 --stats "test1a.stats" --ref 16 --mixed-refs --no-fast-pskip --bframes 4 --b-adapt 2 --b-pyramid --weightb --direct auto --deblock -1:-1 --subme 9 --trellis 2 --partitions all --8x8dct --me umh --merange 32 --threads 6 --aq-strength 1.0 --psy-rd 1.0:0.0 --thread-input --progress --no-dct-decimate --no-psnr --no-ssim --sar 64/45 --frames 500 --fps 25 --output test1a.h264

screenshots (thumbnails)
source (after avs filtering)
http://hdimage.org/images/9evpyz9nyagav1lub33_source_thumb.png (http://hdimage.org/viewer.php?file=9evpyz9nyagav1lub33_source.png)

encode
http://hdimage.org/images/8nfc135o5lqphcqy6w6q_test1a_thumb.png (http://hdimage.org/viewer.php?file=8nfc135o5lqphcqy6w6q_test1a.png)

i've tested a dozen of other tests but none of them was satisfactory enough. parameters i've been changing are psy-rdo (from 0.6 to 1.5), psy-trellis (0.0 to 0.4), aq (0.6 to 1.0). i've also tried lower deblocking. no help either. source seems to be good and compressable enough to achieve better results with this bitrate (1750).

Blue_MiSfit
30th April 2009, 21:43
What you're seeing is not banding, but fringing / chromatic aberration.

It's not fixable by tweaking encoding parameters, but you may have luck with some AviSynth filters.

Not sure - but I think by moving the chroma planes around you could get a good result.

~MiSfit

martinfrombern
30th April 2009, 21:52
thanks for your answer
however, i'm not talking about that blue/purple area, i know it's fringing. i'm talking about the banding immediately below it which is present in encode, but not in source

Sagekilla
30th April 2009, 22:57
I see the banding hes' talking about. It's very subtle, but it's there. I didn't notice until I blew up the image.

neuron2
1st May 2009, 00:30
What is the input source material and what is your script?

martinfrombern
1st May 2009, 02:20
source is dvd pal, 720x576 anamorphic, and avs script is followingLoadPlugin("C:\Program Files (x86)\AviSynth 2.5\plugins\TDeint1.1\TDeint.dll")
LoadPlugin("C:\Program Files (x86)\AviSynth 2.5\plugins\NNEDI 1.3\nnedi.dll")
LoadPlugin("C:\Program Files (x86)\AviSynth 2.5\plugins\MaskTools2.0.a36\mt_masktools-26.dll")
LoadPlugin("C:\Program Files (x86)\AviSynth 2.5\plugins\RemoveGrain1.0\Repair.dll")
LoadPlugin("C:\Program Files (x86)\AviSynth 2.5\plugins\RemoveGrain1.0\RemoveGrainSSE3.dll")
LoadPlugin("C:\Program Files (x86)\AviSynth 2.5\plugins\MVTools2.3.1\mvtools2.dll")
Import("C:\Program Files (x86)\AviSynth 2.5\plugins\TempGaussMC_beta1mod\TempGaussMC_beta1mod.avsi")
MPEG2Source("D:\encoding\current\earth1.d2v",cpu=0)

SelectRangeEvery(1500,50)
Crop(2, 2, -2, -2)

TempGaussMC_beta1mod(EdiMode="NNEDI")
SelectEven()
please, keep in mind that source screenshot is already avs filtered, so avs script is definitely not creating the banding

poisondeathray
1st May 2009, 02:37
Just so others aren't as confused as I was at first (I couldn't see the banding), if you open those links in a new tab, the full image is there (those screenshots are the thumbnail view)

Did you try a higher bitrate, or even zones for that sequence?

martinfrombern
1st May 2009, 03:25
Just so others aren't as confused as I was at first (I couldn't see the banding), if you open those links in a new tab, the full image is there (those screenshots are the thumbnail view)you are right, added note in the first post
Did you try a higher bitrate, or even zones for that sequence?tried higher bitrate, but banding was present even with 2M. didn't try zones because this clip is 9 minutes total, and this scene is not the only one with this problem :(

poisondeathray
1st May 2009, 03:30
Well it's hard to tell from the single screenshot, but this looks to me like a sequence with lots of motion chasing deer in a jeep, so it's entirely possible that 2000kbps might be too low for 576p25...

Sagekilla
1st May 2009, 04:16
If his sequence is very fast moving, I'd say he probably wouldn't notice banding if this were actually playing out. If it were a static scene, yeah. Otherwise, unless you're a fan of slideshow movies, you probably don't need to worry.

Adub
1st May 2009, 08:23
For help with banding, using a combination of x264's AQ and LaTo's Gradfun2DB function should work wonders.

martinfrombern
3rd May 2009, 14:49
tried that but no help :( seems i'm just a bit too nitpicky about the quality and transparency to the source

Sharktooth
3rd May 2009, 15:14
try --psy-rd 1.0:0.7

martinfrombern
3rd May 2009, 16:53
try --psy-rd 1.0:0.7

tried, here comes the screenshot

http://hdimage.org/images/irpaaub2yvxfht3rsfk_3_thumb.png (http://hdimage.org/viewer.php?file=irpaaub2yvxfht3rsfk_3.png)

it's still visible, but also it's definitely reduced.
it would be very helpful if i knew "logical" meaning of psy-trellis. I mean... higher aq means "more to background, less to edges", higher psy-rdo means "more to details". I'm lacking some kind of this simplified explanation for psy-trellis. tried searching it but no help

Emulgator
3rd May 2009, 19:14
You may try and add the line Addgrain(x,0,0) to you input .avs script, where you may move x from 3 to 6.
This is what I do when I shot stage material on 1080i HDV and want to have a smooth haze without banding after encoding.
Just dithering. Sonic Cinevision charges ~17k for a "grain management" plugin. I try with less ;-)

martinfrombern
3rd May 2009, 22:10
don't know if it matters, but forgot to mention that my test encodes are only 10 sec long