AutoAdjust [v2.60 - Update 2015/11/15]

[LaTo]

  +---------------------------------------------------------------------------------+
  |                                                                                 |
  |                    Automatic adjustement plugin: AutoAdjust()                   |
  |                                                                                 |
  |                               v2.60 by "LaTo INV."                              |
  |                                                                                 |
  |                                 15 November 2015                                |
  |                                                                                 |
  +---------------------------------------------------------------------------------+



      I.    DESCRIPTION
      II.   DISCLAIMER
      III.  INSTALLATION
      IV.   REQUIREMENTS
      V.    MULTITHREADING
      VI.   USAGE
      VII.  GENERAL SETTINGS
      VIII. AUTOGAIN SETTINGS
      IX.   AUTOBALANCE SETTINGS
      X.    CHANGELOG



 I. DESCRIPTION :
 ================

  AutoAdjust is an automatic adjustement filter. It calculates statistics of clip,
  stabilizes them temporally and uses them to adjust luminance gain & color balance.

  AutoAdjust have a smoothing & dithering algorithm to avoid banding issue.
  Calculations are made in 32bits float to avoid rounding errors and
  are internally multithreaded and SSE2/AVX2 optimized.

  For manual adjustments, see all plugins in "SmoothAdjust" package.



 II. DISCLAIMER :
 ================

  This program is free for non-commercial use only and must be distributed
  in its original form without any modification, including this documentation file.
  This program is provided as-is in the hope it will be useful but 
  without any warranty, including any hypothetical damages or issues.

  The official home page is "http://forum.doom9.org/showthread.php?t=167573",
  it always contains the latest version of this program.
  This page acts as the official support center, also the author is reachable
  through his member page: "http://forum.doom9.org/member.php?u=131032".

  This plugin has required a lot of work and much coding hours, so if you like
  it and feel the need to reward or inspire the author then please consider donating.
  Donations are a fast, easy and secure operation through Paypal: simply follow the
  link into the "Donate.url" shortcut (you do not need a Paypal account to donate).
  All donations will be very useful and I thank in advance all the generous people,
  also if you are on the doom9's forum please specify your nickname when donating.



 III. INSTALLATION :
 ===================

  There are different versions of AutoAdjust to suit all installations of Avisynth.
   - x86 builds are for Avisynth 32bits
   - x64 builds are for Avisynth 64bits



 IV. REQUIREMENTS :
 ==================

  - Avisynth v2.6.x or Avisynth Plus
  - YUV Planar (Y8, YV411, YV12, YV16, YV24)
  - 8bits or 16bits (MSB/LSB)



 V. MULTITHREADING :
 ===================
  AutoAdjust is internally multithreaded with official Avisynth version. 
  If AvisynthPlus or AvisynthMT is detected, internal multithreading is automatically disabled
  even if external multithreading isn't used (this can be reactivated in parameters).

  AutoAdjust is designed to work with external multithreading, so fastest MT mode can be used.
  With AvisynthPlus, this plugin is a MT_NICE_FILTER (automatic register, no need to call SetFilterMtMode).
  With AvisynthMT, SetMtMode(1) must be called before AutoAdjust.



 VI. USAGE :
 ===========

    AutoAdjust ( [clip], external_clip,
                 auto_gain,
                 dark_limit, bright_limit, gamma_limit, 
                 dark_exclude, bright_exclude, 
                 gain_mode, chroma_process, avg_safety,
                 input_tv, output_tv,
                 auto_balance,
                 chroma_limit, balance_str,
                 scd_threshold, temporal_radius, change_status,
                 use_interp, use_dither,
                 high_quality, high_bitdepth, 
                 threads_count, asm_opt,
                 debug_view )

Last version (2015/11/15): AutoAdjust v2.60

Old version (for Avs2.5 or WinXP or Wine): AutoAdjust v1.02


If you really like this filter, please consider donating...
It will help me a lot to continue development and support.
Thanks for all! (link in the .7z package)


v2.60:
updated for Avs v2.6.0 final
fixed AVX2 crash with Win7 without SP1

v2.50:
fixed crash with external_clip
fixed chroma processing bug
removed temporal_radius=-1
reintroduced internal caching

v2.40:
optimized frame access & frame cache
fixed potential freeze with very slow source filter

v2.20:
optimized memory usage with very long clip
fixed crash with ffdshow_source() and external MT

v2.00:
dropped Avs v2.5.x support
added AVX2 code (40% faster)
modified plugin to be thread-safe (MT_NICE_FILTER)
removed internal caching
added frame cache hints
improved debug_view

v1.90:
changed name of plugin & parameters
added color balance
added C version
fixed regression from v1.75 (bug at scenechange)

v1.75:
added avg_safety parameter
fixed a minor bug in temporal averaging
tweaked multi-threading code

v1.60:
split 'tv_range' into 'input_tv' & 'output_tv'
fixed possible freeze in MT code
added explicit error messages

v1.45:
fixed bugs inside multi-threading code
removed user32 dependency
added AVX2 detection (for future opt)

v1.40:
used new C++11 features
rewritten multi-threading (native)
speed optimization (10% faster)
exception-safe code
added a high quality interpolation mode
improved core algorithm
fixed various little things

v1.02:
fixed minor things & bugs
dropped ICL builds support

v1.00:
first stable release


_

[LaTo]

Screenshot of AutoAdjust(debug_view=true):


* Click the image to enlarge *

[LaTo]

Early alpha development discussion is available on the SmoothAdjust's thread, development and support will now continue here...

[LaTo]

/* Post from 01 april 2013 on SmoothAdjust's thread */


New AutoGain update: AutoGain-beta1.

Big update, any inputs is appreciated

[LaTo]

Update: AutoGain-beta2

More robust scenechange detector, less fail on fade in/out

[LaTo]

New update: AutoGain-beta3.

The plugin seems pretty mature now... One thing to add to be complete is a debug view, but later

[Reel.Deel]

Hey LaTo, thanks for adding 16 bit support.
The overall results are indeed very pleasing.
I was testing it out and I noticed something odd. When I use 8 bit the histogram looks smooth but with 16 bit the histogram looks posterized (hope that's the right term).
Here's the comparison. I use all default settings.

Originally Posted by LaTo

One thing to add to be complete is a debug view, but later

Like AutoLevels, maybe you can also add a scene change indicator to the debug view? I had another idea but seems you got rid of the bright/dark_ignore parameters.

[LaTo]

Originally Posted by Reel.Deel

I was testing it out and I noticed something odd. When I use 8 bit the histogram looks smooth but with 16 bit the histogram looks posterized (hope that's the right term).
Here's the comparison. I use all default settings.

Yes I already fixed this in next version.

Originally Posted by Reel.Deel

Like AutoLevels, maybe you can also add a scene change indicator to the debug view?

Yes it's planed

Originally Posted by Reel.Deel

I had another idea but seems you got rid of the bright/dark_ignore parameters.

Yes I want to keep the plugin as simple as possible, so I removed some parameters... But if people prefers an hardcore filter with a lot of parameters, I can add at least 5 new one ^^

Maybe I should expose more parameters, and split them in "basic settings" and "advanced settings"... Don't know what is best.

Also I will have a look at AutoLevels plugin, maybe I can pick some ideas from this one.

Anyways, all ideas are welcome... So I want to hear your idea even if it's not realizable

[Reel.Deel]

Originally Posted by LaTo

Yes I already fixed this in next version.
...
Yes it's planed

Awesome.

Maybe I should expose more parameters, and split them in "basic settings" and "advanced settings".

Speaking for myself only, I generally stick to the basic settings, but I do appreciate the fact that the advance settings are there when I do need to use them. So yes, I would vote for basic and advanced settings.

Also I will have a look at AutoLevels plugin, maybe I can pick some ideas from this one.

Well, you're in luck, the AutoLevels readme does include a small to do list.
Also, frustum posted some interesting ideas regarding some of the shortcomings of AutoLevels. A little more information on that very topic can be found on his website.

Anyways, all ideas are welcome... So I want to hear your idea even if it's not realizable

Just a thought, and it's a debug view idea if anything...
For the bright/dark ignore parameters do something like ShowLevels plugin, where the ignored bright/dark pixels would get painted a certain color.
For instance, purple is for bright_ignore and green for dark_ignore. Example and exaggerated example. Just in case this is not a good idea, please take it easy on me.

[LaTo]

New version: AutoGain-beta5

Now with debug view

[LaTo]

Originally Posted by Reel.Deel

Well, you're in luck, the AutoLevels readme does include a small to do list.
Also, frustum posted some interesting ideas regarding some of the shortcomings of AutoLevels. A little more information on that very topic can be found on his website.

I've read all these, IMO only 2 things are interesting for AutoGain:
- chroma processing, but it's already on my todo list
- frameOverrides string in AutoLevels

Originally Posted by Reel.Deel

Just a thought, and it's a debug view idea if anything...
For the bright/dark ignore parameters do something like ShowLevels plugin, where the ignored bright/dark pixels would get painted a certain color.
For instance, purple is for bright_ignore and green for dark_ignore. Example and exaggerated example. Just in case this is not a good idea, please take it easy on me.

Yes, could be useful... Will see later which parameters need to be exposed or not


No report from Didée since 2 weeks... Maybe in vacations

[Forensic]

LaTo. This could be an extremely important asset for video forensic work. Over the next few days I will be testing it with several high-profile, low-resolution nighttime surveillance videos and report back my findings. Thank you for creating this tool.

[yup]

Hi Lato!
I try alpha version and see only autogain.
How replace internal ColorYUV call?
ColorYUV(autogain=true, autowhite=true)
may be
Autogain().ColorYUV(autowhite=true)?
Please advice.
yup.

[PDyksma]

Hello LaTo

Thanks for your efforts to create a REAL autolevels plugin.
All the ones I tried till now suffered from crunching the whites...
Also (f.i. the Avisynth one) they do not "obey" the requested outputlevels without clipping.

I'm a noob on scripting but is it not possible to do the following:

- get the brightest pixel in a frame and map it to, let's say 235
- get the darkest pixel and map it to 16
- interpolate the rest in between
- add some averaging to avoid "pumping"

The result should be as attached luminance waveform

[LaTo]

Originally Posted by yup

Hi Lato!
I try alpha version and see only autogain.
How replace internal ColorYUV call?
ColorYUV(autogain=true, autowhite=true)
may be
Autogain().ColorYUV(autowhite=true)?
Please advice.
yup.

IMO, I don't see how autowhite can produce good results... I should look at the source code to see how it works but I think this is an unsafe filter.

However, Autogain().ColorYUV(autowhite=true) will produce the same effect as before (I think).

Originally Posted by PDyksma

Hello LaTo

Thanks for your efforts to create a REAL autolevels plugin.
All the ones I tried till now suffered from crunching the whites...
Also (f.i. the Avisynth one) they do not "obey" the requested outputlevels without clipping.

I'm a noob on scripting but is it not possible to do the following:

- get the brightest pixel in a frame and map it to, let's say 235
- get the darkest pixel and map it to 16
- interpolate the rest in between
- add some averaging to avoid "pumping"

The result should be as attached luminance waveform

If I all understand correctly, you simply need this:
AutoGain(adjust_mode=1,dark_exclude=0.00,bright_exclude=0.00)

[PDyksma]

Originally Posted by LaTo

If I all understand correctly, you simply need this:
AutoGain(adjust_mode=1,dark_exclude=0.00,bright_exclude=0.00)

I do not know what these parameters would do...

Well, I had some old 8mm films laser-scanned. The results are very good but I had a camera in 1978 with a miserable auto-exposure.
So, in a lot of scenes the luminance is shifting back and forth and difficult to correct manually. That's the reason I tried some autolevel-filters. But all suffer from doing nothing in some (dark) scenes and/or blowing up the lights in other. Also, they do not "obey" my given output-levels (0-255, 16-235 or wathever). So I need something to produce a nice equal (given) lumi-level like the screenshot in my previous post and without clipping.
Your autogain filter already comes near to that but with mode=1 I have to put some "brake" on the video (adding a small 4px white border) to reduce the gain-effect on the whites. This gives me a reasonable result so far:

gained-gained.addborders(4,0,0,0,$FFFFFF)
gained=gained.autogain(adjust_mode=1,bright_limit=100.0,dark_limit=1.5)
gained=gained.crop(4,0,-0,-0)

[LaTo]

Originally Posted by PDyksma

I do not know what these parameters would do...

Hum... sorry my fault, these are not in AutoGain for now... It's in beta6

[LaTo]

New AutoGain: AutoGain-beta6

Any reports/tests are welcome

[Forensic]

I applied your filter on several low resolution, high noise, under lit videos. Your filter improved luminosity without amplifying the noise. Outstanding work LaTo!

[LaTo]

Originally Posted by Forensic

I applied your filter on several low resolution, high noise, under lit videos. Your filter improved luminosity without amplifying the noise. Outstanding work LaTo!

Thanks for testing!

Originally Posted by PDyksma

I do not know what these parameters would do...

[...]

Also, they do not "obey" my given output-levels (0-255, 16-235 or wathever). So I need something to produce a nice equal (given) lumi-level like the screenshot in my previous post and without clipping.
Your autogain filter already comes near to that but with mode=1 I have to put some "brake" on the video (adding a small 4px white border) to reduce the gain-effect on the whites

You can now try AutoGain(adjust_mode=1,dark_exclude=0.00,bright_exclude=0.00)