Filter recommendation for Eyes Wide Shut

[stax76]

I want to encode Eyes Wide Shut, the problem is the image quality is a desaster. I wonder which filters would do well? Here is a sample:

http://www.planetdvb.net//non_drupal/files/Eyes_Wide_Shut.m2v

Any recommendations? Usually I encode anamorphic using the HQ Slow profile with CRF 22, would this work well on the source? Maybe the quality is so bad that it's not worth to keep the full size.

[Guest]

Is that your encoded result? What does the source look like?

[jeffy]

I'm afraid this is a source...

[Nikos]

This is my script with good results in the "Eyes Wide Shut" movie:

SetMTMode(5,4)

SetMemoryMax(512)
DirectShowSource("EYES.GRF", audio=false, seek=true, fps=23.976023976023976023976023976024, framecount=228767)

Spline36Resize(1280, 720, 8, 4, 1904 ,1068)

source=last

pre = source.FFT3Dgpu(sigma=4, sigma2=0.1, sigma3=3, sigma4=3, bt=3, plane=4, wintype=1, precision=2, bw=48, bh=48, ow=16, oh=16)

SetMTMode(2)
b2_vec=pre.MVAnalyse(isb=true, delta=2, blksize=8, overlap=4, pel=2, chroma=true, search=2, truemotion=true, sharp=2, dct=0, divide=0, idx=1)
b1_vec=pre.MVAnalyse(isb=true, delta=1, blksize=8, overlap=4, pel=2, chroma=true, search=2, truemotion=true, sharp=2, dct=0, divide=0, idx=1)
f1_vec=pre.MVAnalyse(isb=false, delta=1, blksize=8, overlap=4, pel=2, chroma=true, search=2, truemotion=true, sharp=2, dct=0, divide=0, idx=1)
f2_vec=pre.MVAnalyse(isb=false, delta=2, blksize=8, overlap=4, pel=2, chroma=true, search=2, truemotion=true, sharp=2, dct=0, divide=0, idx=1)

den1=source.MVDegrain2(b1_vec, f1_vec, b2_vec, f2_vec, thSAD=400, limit=10, idx=2)
den2=den1.MVDegrain1(b1_vec, f1_vec, thSAD=300, limit=5, idx=3)

sharp=contra(den2, source)

return sharp

contra= Didie's contra sharpenig function

x264 log:

x264 [info]: slice I:1408 Avg QP:15.42 size:100670
x264 [info]: slice P:112516 Avg QP:16.44 size: 27621
x264 [info]: slice B:114843 Avg QP:19.19 size: 5767

encoded 228767 frames, 11.41 fps, 3279.94 kb/s

[stax76]

@neuron2

Unfortunately it's the source.

@Nikos

Thanks for the suggestion, I'll try it, can I also use DGDecode?

[jeffy]

@stax: I think Nikos used a Blu-ray source.

[Blue_MiSfit]

And surely suffered dearly in encoding time, with that double whammy MVDegrain script for 1080p :D

~MiSfit

[Nikos]

My source was HDDVD 1080p. The 11.41 fps is acceptable. My CPU is Intel Quad, Q6600 (2.4GHz).

[Blue_MiSfit]

Wow...

MVDegrain at 11fps for 720p?

I didnt think it was possible...

~MiSfit

[Blue_MiSfit]

Are you sure??? I'm managing like 1-2fps for 1080p, how did you get 720p to run at 11.41 fps?!?
~MiSfit

[Nikos]

This is the full x254 log.

x264.exe --pass 2 --bitrate 3280 --stats "K:\EYES\EYES.LOG" --level4.1 --keyint 200 --min-keyint 20 --ref 4 --mixed-refs --no-fast-pskip --bframes3 --b-pyramid --b-rdo --bime --weightb --direct auto --filter -2,-1 --subme 6 --trellis 1 --partitions p8x8,b8x8,i4x4,i8x8 --8x8dct --me umh --threads auto --thread-input --zones 221026,228766,b=0.25 --progress --no-psnr --output "eyes.MKV" "EYES.avs"

avis [info]: 1280x720 @ 23.98 fps (228767 frames)
x264 [info]: using cpu capabilities: MMX MMXEXT SSE SSE2 SSE3 SSSE3 Cache64

x264 [info]: slice I:1408 Avg QP:15.42 size:100670
x264 [info]: slice P:112516 Avg QP:16.44 size: 27621
x264 [info]: slice B:114843 Avg QP:19.19 size: 5767

x264 [info]: mb I I16..4: 23.3% 61.9% 14.8%
x264 [info]: mb P I16..4: 5.4% 6.6% 1.6% P16..4: 44.3% 22.7% 11.4% 0.0% 0.0% skip: 8.1%
x264 [info]: mb B I16..4: 0.2% 0.3% 0.1% B16..8: 37.6% 1.3% 3.3% direct: 1.0% skip:56.2%

x264 [info]: 8x8 transform intra:49.9% inter:61.8%
x264 [info]: direct mvs spatial:98.8% temporal:1.2%
x264 [info]: ref P 78.8% 12.3% 5.4% 3.5%
x264 [info]: ref B 89.3% 9.0% 1.7%
x264 [info]: SSIM Mean Y:0.9860813
x264 [info]: kb/s:3279.9

encoded 228767 frames, 11.41 fps, 3279.94 kb/s

My CPU (Q6600) was oc to 3GHz. Maybe i encode to Lagarith first. I don't remember :D

[Blue_MiSfit]

Well then - that's the answer :)

[stax76]

Any recommendations for the SD source sample (which I assume is different from the Blu-ray source)? Some hints would already be a great help, I'm not lazy, I just need starting points/directions. What type of issues has the sample and what filtes are popular to fix these issues?