HDR10 to SDR with Hable tone-mapping

[age]

Hi!
This is a basic script for hdr to sdr conversion, I'm not sure i've used the hable tonemapping in the right way.
it's a tone mapping operator used for videogames but it seems good even for videos.

import vapoursynth as vs


core = vs.get_core()
c = core.ffms2.Source(source = 'path/to/file')

source_peak=1200 #set manually


c=core.resize.Bilinear(clip=c, format=vs.YUV444PS, range_in_s="limited", range_s="full",chromaloc_in_s="center",dither_type="none")

c=core.resize.Bilinear(clip=c, format=vs.RGBS, matrix_in_s="2020ncl", range_in_s="full",dither_type="none")

c=core.resize.Bilinear(clip=c, format=vs.RGBS, transfer_in_s="st2084", transfer_s="linear",dither_type="none", nominal_luminance=source_peak)


exposure_bias=source_peak/100

#hable/uncharted tone mapping
tm = core.std.Expr(c, expr="x  {exposure_bias} * 0.15 x  {exposure_bias} * * 0.05 + * 0.004 + x  {exposure_bias} * 0.15 x  {exposure_bias} * * 0.50 + * 0.06 + / 0.02 0.30 / -  ".format(exposure_bias=exposure_bias),format=vs.RGBS)#12=1200 nits / 100 nits

w = core.std.Expr(c, expr="{exposure_bias} 0.15 {exposure_bias} * 0.05 + * 0.004 + {exposure_bias} 0.15 {exposure_bias} * 0.50 + * 0.06 + / 0.02 0.30 / -  ".format(exposure_bias=exposure_bias),format=vs.RGBS)#  


c = core.std.Expr(clips=[tm,w], expr=" x  1 y  / *  ",format=vs.RGBS)

#c=core.fmtc.primaries(clip=c, prims="2020", primd="709")
c=core.resize.Bilinear(clip=c, format=vs.RGBS, primaries_in_s="2020", primaries_s="709",dither_type="none")
c=core.resize.Bilinear(clip=c, format=vs.RGBS, transfer_in_s="linear", transfer_s="709",dither_type="none")

c=core.resize.Bilinear(clip=c, format=vs.YUV420P8, matrix_s="709", range_in_s="full",range_s="limited",chromaloc_in_s="center")


c.set_output()

HDR10 is generally mastered with with a peak brightness of 1000,1200,1500 nits and it corrisponds with the parameter in the script "source_peak".

I only have some hdr video sample downloaded so I can't test exhaustively this script, but it produces reasonably good output.

There's at least one issue:
I use the resizer in the core for conversion from bt.2020 to bt.709 and from the github
"Note that "z" is not a color management system and should not be used to perform drastic contrast or gamut reduction, such as BT.2020 to BT.709."


Screenshot before/after:
http://screenshotcomparison.com/comparison/203274

http://screenshotcomparison.com/comparison/203275

http://screenshotcomparison.com/comparison/203276

[Cary Knoop]

Super!

One quick question, going from full to limited range is designed to capture out of range values (WTW and BTB) and bring them into legal or is the script designed for full range video to be converted to limited range?
Most video obviously uses limited range with the exception of 4:4:4 sources.


It would be great to have an "up" conversion from SDR to HDR as well.

[sneaker_ger]

Neither. Both input of first filter as well as output of last filter are limited range.

I wonder if and how to handle e.g. color coordinates. The script only asks for peak luminance while HDR(10) metadata looks something like this:

|   + Video colour information
|    + Colour matrix: 9
|    + Colour range: 1
|    + Colour transfer: 16
|    + Colour primaries: 9
|    + Video colour mastering metadata
|     + Max luminance: 1000
|     + Min luminance: 0.001
|     + Red colour coordinate x: 0.68
|     + Red colour coordinate y: 0.31996
|     + Green colour coordinate x: 0.26494
|     + Green colour coordinate y: 0.68996
|     + Blue colour coordinate x: 0.15
|     + Blue colour coordinate y: 0.05998
|     + White colour coordinate x: 0.3127
|     + White colour coordinate y: 0.32896

Do only the player and/or display need these? In fmtc.primaries you can supply such coordinates.

[age]

Originally Posted by sneaker_ger

Neither. Both input of first filter as well as output of last filter are limited range.

I wonder if and how to handle e.g. color coordinates. The script only asks for peak luminance while HDR(10) metadata looks something like this:

|   + Video colour information
|    + Colour matrix: 9
|    + Colour range: 1
|    + Colour transfer: 16
|    + Colour primaries: 9
|    + Video colour mastering metadata
|     + Max luminance: 1000
|     + Min luminance: 0.001
|     + Red colour coordinate x: 0.68
|     + Red colour coordinate y: 0.31996
|     + Green colour coordinate x: 0.26494
|     + Green colour coordinate y: 0.68996
|     + Blue colour coordinate x: 0.15
|     + Blue colour coordinate y: 0.05998
|     + White colour coordinate x: 0.3127
|     + White colour coordinate y: 0.32896

Do only the player and/or display need these? In fmtc.primaries you can supply such coordinates.

Max luminance should be the right value for "source_peak"

Bt2020 is the standard for hdr10

Maybe i can try to pass all the info from the metadata

[Cary Knoop]

Originally Posted by sneaker_ger

Neither. Both input of first filter as well as output of last filter are limited range.

I got it, RGB processing is full while in and out is limited.

[age]

Right :-)


I think tone mapping is needed for hdr to sdr conversion,
for sdr to hdr up conversion is needed an inverse tone mapping algo but i can't find such algorithm on google.

[zub35]

Originally Posted by age

I only have some hdr video sample downloaded so I can't test exhaustively this script, but it produces reasonably good output.

more samples SDR and HDR :
https://cloud.mail.ru/public/2iEk/CiyUARGZy

[kolak]

Originally Posted by age

Hi!
This is a basic script for hdr to sdr conversion, I'm not sure i've used the hable tonemapping in the right way.
it's a tone mapping operator used for videogames but it seems good even for videos.

HDR10 is generally mastered with with a peak brightness of 1000,1200,1500 nits and it corrisponds with the parameter in the script "source_peak".

I only have some hdr video sample downloaded so I can't test exhaustively this script, but it produces reasonably good output.

There's at least one issue:
I use the resizer in the core for conversion from bt.2020 to bt.709 and from the github
"Note that "z" is not a color management system and should not be used to perform drastic contrast or gamut reduction, such as BT.2020 to BT.709."


Screenshot before/after:
http://screenshotcomparison.com/comparison/203274

http://screenshotcomparison.com/comparison/203275

http://screenshotcomparison.com/comparison/203276

Have a look here:
http://www.mysterybox.us/blog/2016/1...delivering-hdr

and than use free Resolve and compare your result to Resolve.

[kolak]

Originally Posted by age

Right :-)


I think tone mapping is needed for hdr to sdr conversion,
for sdr to hdr up conversion is needed an inverse tone mapping algo but i can't find such algorithm on google.

Dolby and Technicolor are working on such a algorithms with adaptive per scene analysis.

MadVR does it quite well, can you still some math from there?

[kolak]

Originally Posted by age

Max luminance should be the right value for "source_peak"

Bt2020 is the standard for hdr10

Maybe i can try to pass all the info from the metadata

I assume fact that these videos are graded to P3 gamut doesn't change anything here as it's within Bt2020.

[CruNcher]

Looks nice at least not like such a over saturated mess

https://www.youtube.com/watch?v=vI91heLv_v4

sorry i know it gets old but i find this a master example of how your HDR->SDR tone mapping shouldn't endup

You should also try

https://knarkowicz.wordpress.com/201...mapping-curve/

[kolak]

This sample is terrible.

[age]

Thank you all for your suggestion and the hdr samples.
Well I've discovered some interesting things....
First, a better version of the hable tonemapping that gives better contrast,brightness and saturation.
http://screenshotcomparison.com/comparison/203550


I've tried different algorithms, the aces tonemapping too but I don't like it very much.
Here a good link for a lot of open source tm operators :

https://github.com/tizian/tonemapper.../src/operators

And a different version of aces:

https://github.com/keijiro/PostProce...emapping.cginc


new script with correct parameters

import vapoursynth as vs


core = vs.get_core()
c = core.ffms2.Source(source  = 'C:/bla')

source_peak=1000 #set manually
LDR_nits=100     #set 150 or 200 for lowering the brightness

c=core.resize.Bilinear(clip=c, format=vs.YUV444PS, range_in_s="limited", range_s="full",chromaloc_in_s="center",dither_type="none")

c=core.resize.Bicubic(clip=c, format=vs.RGBS, matrix_in_s="2020ncl", range_in_s="full",dither_type="none")

c=core.resize.Bilinear(clip=c, format=vs.RGBS, transfer_in_s="st2084", transfer_s="linear",dither_type="none", nominal_luminance=source_peak)

exposure_bias=(source_peak/LDR_nits)

#hable tone mapping  correct parameters
#["A"] = 0.22
#["B"] = 0.3
#["C"] = 0.1
#["D"] = 0.2
#["E"] = 0.01
#["F"] = 0.3
#["W"] = 11.2
#((x * (A*x + C*B) + D*E) / (x * (A*x+B) + D*F)) - E/F"   
tm = core.std.Expr(c, expr="x    {exposure_bias} * 0.22 x    {exposure_bias} * * 0.03 + * 0.002 +    x   {exposure_bias} * 0.22 x   {exposure_bias} * * 0.3 + * 0.06 + / 0.01 0.30 / -  ".format(exposure_bias=exposure_bias),format=vs.RGBS)#12=1200 nits / 100 nits



w = core.std.Expr(c, expr="{exposure_bias} 0.22 {exposure_bias} * 0.03 + * 0.002 + {exposure_bias} 0.22 {exposure_bias} * 0.3 + * 0.06 + / 0.01 0.30 / -  ".format(exposure_bias=exposure_bias),format=vs.RGBS)

c = core.std.Expr(clips=[tm,w], expr=" x  1 y  / *  ",format=vs.RGBS)

#c=core.fmtc.primaries(clip=c, prims="2020", primd="709")
c=core.resize.Bilinear(clip=c, format=vs.RGBS, primaries_in_s="2020", primaries_s="709",dither_type="none")
c=core.resize.Bilinear(clip=c, format=vs.RGBS, transfer_in_s="linear", transfer_s="709",dither_type="none")

c=core.resize.Bilinear(clip=c, format=vs.YUV420P8, matrix_s="709", range_in_s="full",range_s="limited",chromaloc_in_s="center")


c.set_output()

[qtwigg]

Greyed

Saturated


Normal Bluray

UHD BluRay


Ran that script, thank you so so very much for it. I just need a few tweaks, any ideas? Thanks

[george84]

@age

Screenshot comparison does not work since 1.8.2017

[Selur]

for easier use, I wrapped the corrected version into a function:
(UPDATED VERSION in later post)

'''
HDR-10 to SDR mapping, from 'age' (https://forum.doom9.org/showthread.php?p=1800675)
'''
import vapoursynth as vs

def hablehdr10tosdr(clip, source_peak=1200, ldr_nits=100, tFormat=vs.YUV420P8, tMatrix="709", tRange="limited"):
  core = vs.get_core()
  clip=core.resize.Bilinear(clip=clip, format=vs.YUV444PS, range_in_s="limited", range_s="full",chromaloc_in_s="center",dither_type="none")
  clip=core.resize.Bilinear(clip=clip, format=vs.RGBS, matrix_in_s="2020ncl", range_in_s="full",dither_type="none")
  clip=core.resize.Bilinear(clip=clip, format=vs.RGBS, transfer_in_s="st2084", transfer_s="linear",dither_type="none", nominal_luminance=source_peak)
  exposure_bias=(source_peak/ldr_nits) #set 150 or 200 for lowering the brightness

  #hable tone mapping
  #["A"] = 0.22
  #["B"] = 0.3
  #["C"] = 0.1
  #["D"] = 0.2
  #["E"] = 0.01
  #["F"] = 0.3
  #["W"] = 11.2
  #((x * (A*x + C*B) + D*E) / (x * (A*x+B) + D*F)) - E/F"   
  tm = core.std.Expr(clip, expr="x    {exposure_bias} * 0.22 x    {exposure_bias} * * 0.03 + * 0.002 +    x   {exposure_bias} * 0.22 x   {exposure_bias} * * 0.3 + * 0.06 + / 0.01 0.30 / -  ".format(exposure_bias=exposure_bias),format=vs.RGBS)#12=1200 nits / 100 nits
  
  w = core.std.Expr(clip, expr="{exposure_bias} 0.15 {exposure_bias} * 0.05 + * 0.004 + {exposure_bias} 0.15 {exposure_bias} * 0.50 + * 0.06 + / 0.02 0.30 / -  ".format(exposure_bias=exposure_bias),format=vs.RGBS)#  

  clip = core.std.Expr(clips=[tm,w], expr=" x  1 y  / *  ",format=vs.RGBS)
  clip=core.resize.Bilinear(clip=clip, format=vs.RGBS, primaries_in_s="2020", primaries_s=tMatrix,dither_type="none")
  clip=core.resize.Bilinear(clip=clip, format=vs.RGBS, transfer_in_s="linear", transfer_s=tMatrix,dither_type="none")

  if format != vs.RGBS:
    clip=core.resize.Bilinear(clip=clip, format=tFormat, matrix_s=tMatrix, range_in_s="full",range_s=tRange,chromaloc_in_s="center")
  return clip;

Cu Selur

[Magik Mark]

Is there a way to convert this to avisynth+?

[jkilez]

Originally Posted by Magik Mark

Is there a way to convert this to avisynth+?

I would really like to see that as well. The thread on HDR to SDR using FFmpeg is certainly helpful, but it would great to do it all within Avisynth+ scripting.

[Selur]

small adjustment to the above code,.. (made color location available)

'''
HDR-10 to SDR mapping, from 'age' (https://forum.doom9.org/showthread.php?p=1800675)
'''
def hablehdr10tosdr(clip, source_peak=1200, ldr_nits=100, tFormat=vs.YUV420P8, tMatrix="709", tRange="limited", color_loc="center"):
  core = vs.get_core()
  clip=core.resize.Bilinear(clip=clip, format=vs.YUV444PS, range_in_s="limited", range_s="full",chromaloc_in_s=color_loc,dither_type="none")
  clip=core.resize.Bilinear(clip=clip, format=vs.RGBS, matrix_in_s="2020ncl", range_in_s="full",dither_type="none")
  clip=core.resize.Bilinear(clip=clip, format=vs.RGBS, transfer_in_s="st2084", transfer_s="linear",dither_type="none", nominal_luminance=source_peak)
  exposure_bias=(source_peak/ldr_nits) #set 150 or 200 for lowering the brightness

  #hable tone mapping
  #["A"] = 0.22
  #["B"] = 0.3
  #["C"] = 0.1
  #["D"] = 0.2
  #["E"] = 0.01
  #["F"] = 0.3
  #["W"] = 11.2
  #((x * (A*x + C*B) + D*E) / (x * (A*x+B) + D*F)) - E/F"   
  tm = core.std.Expr(clip, expr="x    {exposure_bias} * 0.22 x    {exposure_bias} * * 0.03 + * 0.002 +    x   {exposure_bias} * 0.22 x   {exposure_bias} * * 0.3 + * 0.06 + / 0.01 0.30 / -  ".format(exposure_bias=exposure_bias),format=vs.RGBS)#12=1200 nits / 100 nits
  
  w = core.std.Expr(clip, expr="{exposure_bias} 0.15 {exposure_bias} * 0.05 + * 0.004 + {exposure_bias} 0.15 {exposure_bias} * 0.50 + * 0.06 + / 0.02 0.30 / -  ".format(exposure_bias=exposure_bias),format=vs.RGBS)#  

  clip = core.std.Expr(clips=[tm,w], expr=" x  1 y  / *  ",format=vs.RGBS)
  clip=core.resize.Bilinear(clip=clip, format=vs.RGBS, primaries_in_s="2020", primaries_s=tMatrix,dither_type="none")
  clip=core.resize.Bilinear(clip=clip, format=vs.RGBS, transfer_in_s="linear", transfer_s=tMatrix,dither_type="none")

  if format != vs.RGBS:
    clip=core.resize.Bilinear(clip=clip, format=tFormat, matrix_s=tMatrix, range_in_s="full",range_s=tRange,chromaloc_in_s=color_loc)
  return clip;

I agree that it would be nice to have a full port of ffmpegs tone map in Vapoursynth (and Avisynth).

[ifb]

I have a vapoursynth port available. I've only tested it with HLG content. YMMV.

https://github.com/ifb/vapoursynth-tonemap