Resizing from 4K to 2K with maximum PQ

[Tenaka]

Hi !


I'm asking for advice concerning my settings.

I would like to resize the source file  from 4K to 2K and passthrough everything, while ensuring maximum picture quality and encoding speed.

Source file :
Resolution : 3840 x 2160
Codec : HVEC
Profile : Main 10@L5.1@High
Color primaries : BT.2020

Target file :
Resolution : 1920x1080
Codec : HVEC

I don't mind a file size around 30Go to 70Go for 2 hours of video content.
My video player can downscale 4K to 2K using SSIM in live, so it should be quick once the streams are extracted no ? (i5 3570K 4 thread @ 4.41Ghz)


Here are my settings :

Base tab

• Video : x265
  
• Audio : passthrough all
  

x265 -> Base  tab

• Constant quantizer (1 pass), strengh at 0, lossless box ticked
  
• Profile/Level/Tier : Main 10@L5.1@High
  

x265 -> Signaling tab

• Primaries box ticked
  
• Matrix box ticked
  

Crop/Risize tab

• Picture resize : bicubic-spline
  
• Auto adjust to mod 2
  
• 1920x1080
  

Everything else by default.

x265 --input --output-depth 10 --y4m --qg-size 8 --limit-modes --no-open-gop --keyint 24 --min-keyint 1 --lookahead-slices 0 --lossless --aud --no-deblock --repeat-headers --colorprim bt709 --transfer bt709 --colormatrix bt2020nc

Am I doing something wrong ?

Do you have any recommendation concerning my settings ?

I would be happy to provide more informations.

[Selur]

4k HEVC encoding always will take quite some time with that CPU.
Since you are using lossless encoding picture quality will always be as high as it gets, but file size will be huge.

Using slower settings for motion search etc. will decrease the file size but slow down the encoding.
Using faster settings for motion search etc. will speedup the encoding but increase the file size.
-> Choose your poison.

Cu Selur

Ps.: Personally I wouldn't use lossless encoding and probably go with a crf of 16 or lower, but that depends on what you want to use that encoding for. As intermediate lossless should be fine, but be aware that no hardware decoder I know of currently can decode lossless HEVC. So playback will be cpu intensive.

PPs.: depending on the content you can try different resizes and see there differences (which should be minimal for most)

[Tenaka]

While using the lossless options I was anticipated a file around the same size as the source one.

I think I'll go with a CRF of 10 and leave the rest at the previous settings described in my message.

About your "PPS" are you talking of resize algorithms ?

I'm going for a sharp image and I'm divided between :

• 9 - lanczos (sharp image but heavy ringing)
  
• 10 - bicubic spline (limited ringing but soft image)
  
• 6 - luma bicubic, chroma bilinear (faster than Lanczos and B-spline)
  

When I watch 4K content with madvr as renderer, I'm using the Mitchell-Netravali (cubic(1/3,1/3)) algorithm for downscaling to 2K.
My CPU is rocketing at 100% on all threads though.

[Selur]

While using the lossless options I was anticipated a file around the same size as the source one.

Lol, really unlikely unless your source already used a lossless compression.

About the resizer, in the Vapoursynth-tab you can also set Vapoursynth specific resizers.

Cu Selur

[JackReacher]

While using the lossless options I was anticipated a file around the same size as the source one.

Lol, really unlikely unless your source already used a lossless compression.

About the resizer, in the Vapoursynth-tab you can also set Vapoursynth specific resizers.

Cu Selur

Hi when selecting Vapoursynth specific resizers, does it override the one selected in the crop/resize tab?

[Selur]

Hi when selecting Vapoursynth specific resizers, does it override the one selected in the crop/resize tab?

Yes, unless you select a resizer which is only for upscaling and you are downscaling.

Cu Selur