Video Artifacting ?

[gav1577]

Hi i does anyone have an idea what could be the cause of the artifact's i have when backing up bluray to avchd all i can describe it as is blocking when the lightning strikes the antenna here is a sample if anyone is interested
http://www.mediafire.com/?z9owvxl9wja

Build x264.956 patched
Source Retail bluray

Spline36Resize(1920,800).msharpen(10,15,true).AddBorders(0,140,0,140)

--pass 1 --bitrate 4597 --stats "D:\Temp\video\job1\job1.stats" --level 4.1 --ref 3 --mixed-refs --no-fast-pskip --bframes 3 --bime --direct auto --filter -1,-1 --subme 5 --partitions none --ipratio 1.1 --pbratio 1.1 --vbv-bufsize 14000 --vbv-maxrate 18000 --me dia --threads auto --thread-input --progress --no-dct-decimate --no-psnr --no-ssim --mvrange 511 --aud --nal-hrd --sar 1:1 --output "D:\Temp\video\video.264"

--pass 2 --bitrate 4597 --stats "D:\Temp\video\job1\job1.stats" --level 4.1 --ref 3 --mixed-refs --no-fast-pskip --bframes 3 --b-rdo --bime --direct auto --filter -1,-1 --subme 6 --trellis 2 --partitions p8x8,b8x8,i4x4,i8x8 --8x8dct --ipratio 1.1 --pbratio 1.1 --chroma-qp-offset -2 --vbv-bufsize 14000 --vbv-maxrate 18000 --me umh --merange 64 --threads auto --thread-input --progress --no-dct-decimate --no-psnr --no-ssim --psy-rd 0.0:0.0 --mvrange 511 --aud --nal-hrd --sar 1:1 --output "D:\Temp\video\video.264"

cabac=1 / ref=3 / deblock=1:-1:-1 / analyse=0x3:0x113 / me=umh / subme=6 / psy_rd=0.0:0.0 / brdo=1 / mixed_ref=1 / me_range=64 / chroma_me=1 / trellis=2 / 8x8dct=1 / cqm=0 / deadzone=21,11 / chroma_qp_offset=-2 / threads=6 / nr=0 / decimate=0 / mbaff=0 / bframes=3 / b_pyramid=0 / b_adapt=1 / b_bias=0 / direct=3 / wpredb=0 / bime=1 / keyint=250 / keyint_min=25 / scenecut=40(pre) / rc=2pass / bitrate=4597 / ratetol=1.0 / qcomp=1.00 / qpmin=10 / qpmax=51 / qpstep=4 / cplxblur=20.0 / qblur=0.5 / vbv_maxrate=18000 / vbv_bufsize=14000 / ip_ratio=1.10 / pb_ratio=1.10 / aq=2:1.00

[Sharktooth]

i didnt look at the sample but the x264 build r951 included in megui was screwed due to a wrong patch (my fault...well, not properly but however it's somewhat my fault...). try with the new build (r956 in the auto-update) and see if it's getting better.
just a suggestion: merange 64, no-dct-decimate, no-fast-pskip are an overkill... you will get more quality and maybe the same speed with merange 16, subme 7 and psy-rd set to the default values in place of them...

[gav1577]

i didnt look at the sample but the x264 build r951 included in megui was screwed due to a wrong patch (my fault...well, not properly but however it's somewhat my fault...). try with the new build (r956 in the auto-update) and see if it's getting better.
just a suggestion: merange 64 is an overkill...

Thanks for the reply but it was done using r956 i should have stated that my fault :)

[Sharktooth]

what build?

[gav1577]

what build?

bob0rs Build x264.956

[Sharktooth]

weird... i did a couple of test encodes and they didnt have any artifacts. i better look at your sample...

[gav1577]

weird... i did a couple of test encodes and they didnt have any artifacts. i better look at your sample...

Thanks :) BTW i did not use psy rdo because the movie is anime

[Sharktooth]

ah i see... try with a vanilla git build (remove --psy-rd, --aud and --nal-hrd from the commandline) and see if it behaves the same.

[gav1577]

ah i see... try with a vanilla git build (remove --psy-rd, --aud and --nal-hrd from the commandline) and see if it behaves the same.

Dont mean to sound stupid but where can i get the vanilla build ?

[Sharktooth]

http://x264.nl. get r956 (the same version as the modified build you currently have)

[gav1577]

Is it ok if i use the first pass stats file from bob0rs Build as it was a whole first pass of the movie it would save me time if i could ?

[Sharktooth]

nope. do it all from scratch since the problem may also come from the stats...

[gav1577]

nope. do it all from scratch since the problem may also come from the stats...

Ok i will report back thanks for the help. :)

[Dark Shikari]

1. Psy-RD is (in my experience) very good on CGI animation, but, that's not really related...

2. About the artifacting: Its VBV, folks! Dontcha love it? :devil:

Can you upload the smallest possible sample of the source that generates the problem (a few seconds is best), along with the exact settings that produced the problem with an unpatched build? I will forward this to Gabriel for VBV analysis :)

[Sagekilla]

On the upside we are improving VBV support, no? ;)

[gav1577]

1. Psy-RD is (in my experience) very good on CGI animation, but, that's not really related...

2. About the artifacting: Its VBV, folks! Dontcha love it? :devil:

Can you upload the smallest possible sample of the source that generates the problem (a few seconds is best), along with the exact settings that produced the problem with an unpatched build? I will forward this to Gabriel for VBV analysis :)

Ok no probs :) EDIT: Ok i have cut a sample from the source but cant seem to reproduce the problem. Thing is i still have the whole movies first pass stats if i use that i can reproduce the issue and have found the problem occurs with both patched and unpatched builds but through testing have found by using a patched build if i enable --psy-rd 1.0:0.0 the problem is solved. If disable --psy-rd 0.0:0.0 the problem reappears Weird I can still upload the source sample although i can not reproduce the problem with the source sample only the whole movie

[qyqgpower]

It seems to be the same problem I reported here (http://forum.doom9.org/showthread.php?p=1177647#post1177647)
although I can't reproduce it so fortunately I can fix it by reencoding and replacing the broken GOP:p


So until the issue is fixed, I still need to check every file that encoded even after r949.:devil:

[Dark Shikari]

Ok no probs :) do you want a few seconds more or less than the sample i already uploaded ? EDIT: Ok i have cut a sample from the source but cant seem to reproduce the problem. Thing is i still have the whole movies first pass stats if i use that i can reproduce the issue and have found the problem occurs with both patched and unpatched builds but through testing have found by using a patched build if i enable --psy-rd 1.0:0.0 the problem is solved if disable --psy-rd 0.0:0.0 the problem reappears WeirdFind the easiest and smallest test case you can contrive where it occurs. Doesn't matter what settings, if you can make it happen again on a small input sample (try, say, 30 seconds), upload that and the settings needed to replicate it.

[gav1577]

ds i cant reproduce the problem with a sample i tried 30 second 1 mim 2 min still no luck only seems to be with the whole movie i will try some more. Ok i tried again cant get the problem to happen from a sample i can keep the 1st pass stats file and upload it if someone who owns the movie wants to try and reproduce the issue using the stats file and exact same settings

[bob0r]

@gav1577

What blu-ray is it?
I could help try make a good sample also...

[gav1577]

Hi its Ratatouille Bluray if you have it i can upload the stats if you want to give it a go :)

[Dark Shikari]

ds i cant reproduce the problem with a sample i tried 30 second 1 mim 2 min still no luck only seems to be with the whole movie i will try some more. Ok i tried again cant get the problem to happen from a sample i can keep the 1st pass stats file and upload it if someone who owns the movie wants to try and reproduce the issue using the stats file and exact same settingsCould post just the section of the statsfile applicable to the small section of frames containing the problem, then, so it can be inspected?

[gav1577]

Could post just the section of the statsfile applicable to the small section of frames containing the problem, then, so it can be inspected?

To be honest i wouldn't know where to begin i could upload the entire stats file and and give the times/frames where the issue begins and ends if thats any help maybe someone could trawl through it and take a look :)

[F J Walter]

To be honest i wouldn't know where to begin i could upload the entire stats file and and give the times/frames where the issue begins and ends if thats any help maybe someone could trawl through it and take a look :)

Yes, that would be useful - post the entire stats file, and tell us approximately where in the resulting movie the problems occurred, in terms of seconds or frame number or whatever.

I have found VBV problems very infuriating to reproduce because naturally, any tiny shift in where you begin and end your small sample can make the problems go away or come sharply in other places. Still, it's worth trying to track down the problem.

It would also be useful if you could try reproducing the problem without AQ and/or without multithreaded, to get an idea of whether they affect things.

[gav1577]

Yes, that would be useful - post the entire stats file, and tell us approximately where in the resulting movie the problems occurred, in terms of seconds or frame number or whatever.

I have found VBV problems very infuriating to reproduce because naturally, any tiny shift in where you begin and end your small sample can make the problems go away or come sharply in other places. Still, it's worth trying to track down the problem.

It would also be useful if you could try reproducing the problem without AQ and/or without multithreaded, to get an idea of whether they affect things.

Well if someone thinks they may be able to track down the issue through the stats and wants to give it a shot i will upload the stats :)

[Sharktooth]

upload the stats... and tell us what is the time or frame where the problem occurs...

[gav1577]

upload the stats... and tell us what is the time or frame where the problem occurs...

Ok here is the stats and times/frames

Stats http://www.mediafire.com/?sharekey=460e480d51406f9ad2db6fb9a8902bda

Start Time
00:06:44 Frame [9691]

End Time
00:06:46 Frame [9736]

[leoenc]

Here is a small sample (SD) that results in ugly artefacts every time:
http://www.mediafire.com/?zzgmcq8ppqn

x264_957.exe --bitrate 2000 --level 3 --keyint 40 --ref 3 --mixed-refs --bframes 3 --b-rdo --bime --weightb --subme 7 --trellis 2 --analyse p8x8,b8x8,i4x4 --vbv-bufsize 3000 --vbv-maxrate 2000 --vbv-init 0.3 --ratetol 0.1 --thread-input --threads 3 --progress --no-psnr --no-ssim --output "PN_sample.264" "PN_sample.avs"

[Sharktooth]

why keyint 40, maxrate 2000/bitrate 2000, vbv-init 0.3 and ratetol 0.1?
also 1 pass and VBV will misbehave.. expecially if you specify a bitrate = maxbitrate...
fix your settings and try again...

[leoenc]

I can't really change most of these settings, as I'm testing a VOD setup with a hardware decoder.
specifying bitrate lower than maxrate doesn't help (I tried 1800).

[Sharktooth]

try it again with threads 1...

[Dark Shikari]

Here is a small sample (SD) that results in ugly artefacts every time:
http://www.mediafire.com/?zzgmcq8ppqn

x264_957.exe --bitrate 2000 --level 3 --keyint 40 --ref 3 --mixed-refs --bframes 3 --b-rdo --bime --weightb --subme 7 --trellis 2 --analyse p8x8,b8x8,i4x4 --vbv-bufsize 3000 --vbv-maxrate 2000 --vbv-init 0.3 --ratetol 0.1 --thread-input --threads 3 --progress --no-psnr --no-ssim --output "PN_sample.264" "PN_sample.avs"I'll analyse this and report the results in a bit.

Edit: Eh, I wouldn't say this is a failure of VBV; there's no unnecessary QP51 or a failure of row-based ratecontrol. What's going on is the ordinary problem of VBV; a case where its just hard to get good quality in a given sequence with the given VBV restrictions! Its made worse by the fact that x264 doesn't have lookahead, and you're using 1pass instead of 2pass mode, but I don't see any catastrophic failures of the algorithm here.

[Sharktooth]

i cant see those ugly artifacts... but i cant see i frames are much better than p and b frames and fades are not properly the best... that's due to your weird settings AND 1 pass ABR...

[leoenc]

I'll analyse this and report the results in a bit.

Edit: Eh, I wouldn't say this is a failure of VBV; there's no unnecessary QP51 or a failure of row-based ratecontrol. What's going on is the ordinary problem of VBV; a case where its just hard to get good quality in a given sequence with the given VBV restrictions! Its made worse by the fact that x264 doesn't have lookahead, and you're using 1pass instead of 2pass mode, but I don't see any catastrophic failures of the algorithm here.

So the low VBV maxrate is the main problem?

[Sharktooth]

try doing a 2 pass encode first. if it doesnt get better then it can be fixed only by rising the VBV maxrate

[lithiumus]

Here is a clip I have been testing that I am able to reproduce some artifacts and blocking. I tried deblocking -1 -1 and trellis 2 and there was little to no difference in improvement. Tried VBV at 0 and 40000.

The blocking is around the letters "Pole to pole". This is a clip from Planet Earth disc 1.

program --pass 2 --bitrate 10500 --stats ".stats" --level 4.1 --keyint 24 --min-keyint 2 --ref 3 --mixed-refs --no-fast-pskip --bframes 3 --b-rdo --bime --weightb --direct auto --subme 7 --trellis 1 --partitions p8x8,b8x8,i4x4,i8x8 --8x8dct --vbv-bufsize 40000 --vbv-maxrate 40000 --me umh --threads auto --thread-input --progress --no-ssim --output "output" "input" --mvrange 511 --aud --nal-hrd --sar 1:1

Source
http://www.mediafire.com/?fjithvcqz0m

Test encode
http://www.mediafire.com/?ch5kvaviedc

[engene]

I tried r969 with megui's x264-device-psp(2pass) profile,the artifact is seem fixed