Reptorian's G'MIC Code Workshop

[Reptillian]

Audacity style glitch art filter possible? Seem that it is.

 

Edited December 17, 2020 by Reptillian

[Reptillian]

Last night, I added negative overload count support for Thorn Fractal. I know negative overload doesn't make sense as count numbers are not negative, but it activates a special feature within Thorn Fractal.

[Reptillian]

Not a big deal, but I had optimized Grouped Pixel Axis-Based Shift. The last filter to optimize is the Chirokov-Taylor filter.

[Reptillian]

Axis Streak, Perspective Streak, Grouped Pixels Shift By Rectangular Polar Transformation has been optimized.

Edited December 27, 2020 by Reptillian

[Reptillian]

I had changed Reverse Engineer Gradient Map. Thanks to garagecoder at discuss.pixls.us, he was able to optimize the part of reverse engineering gradient map. Now, you can use large image, and you get result quickly. No need to wait 3 s or more.

Edited December 27, 2020 by Reptillian

[Reptillian]

Chirikov-Taylor has a new feature where you can enable/disable multi-threading.

[Reptillian]

Roadmap for 2021:

1. Refactor Naeddyar mapmaking script to more suitable version. He wrote scripts with little knowledge of gmic scripting. So, hence the need of refactoring. <- I’m here now.

2. Refactor few of my script and adding new feature.

3. New filters only after the refactor. I won’t be refactoring my script for optimization, etc. I'll be making filters that are so much harder to create.

Edited January 2, 2021 by Reptillian

[Reptillian]

New filter Color Curve Reverse Engineering. This can be used to create a gradient map using essentially what the calculated color curve on a black and white gradient. So, you can find approximate Instagram filters color curves, etc.

Edited January 3, 2021 by Reptillian

[Vagabondi]

On 1/2/2021 at 2:12 AM, Reptillian said:

Roadmap for 2021

 

On 1/2/2021 at 2:12 AM, Reptillian said:

I'll be making filters that are so much harder to create

 

Good luck on the road!

 

Speaking of the filters that already exist: I discovered your Non-Convolution Edge Extraction the other day and was really impressed by its accuracy and settings (Blending Opacity is the best). Added to Faves; how come its status is still "testing"?

 

And playing with Offset / Angle can make surprising things...

 

[Reptillian]

7 minutes ago, Vagabondi said:

Added to Faves; how come its status is still "testing"?

I am not the one to decide whether it should be moved out of testing. That would go to @G'MIC

[Reptillian]

Looks like upgrading Perspective Streak is going to take a while. That being said, I added 3 new formulas to Thorn Fractal:

• Chaotic Hooks Unearthing
• Sinusoidal Liquid
• Cosinusoidal Liquid

Should be here after 2:22 EST.

[lynxster4]

On 1/6/2021 at 1:22 PM, Reptillian said:

That being said, I added 3 new formulas to Thorn Fractal:

 

Have you had a chance to make this upgrade?  🙂  Thanks!

[Reptillian]

18 minutes ago, lynxster4 said:

 

Have you had a chance to make this upgrade?  🙂  Thanks!

 

Yes. If it haven't arrived, then that means that gmic version is no longer supported and a new version should arrive at some point later which would have this update after refresh.

[lynxster4]

Yes, it's updated now @Reptillian.  Thank you!

[Reptillian]

Note: The new rep_recpoltrans has not been implemented yet. I still have to edit rep_axis_streak_color to finally fix a bug before release.

 

Wow! Now I think I solved recpoltrans! The new changes is by far more faster. From 1.5 s to .194 s. It is also more accurate. I only worked on if -3, and 2 part of rep_recpoltrans since they matter a lot more than the others as they’re both designed for precision. The others if in rep_recpoltrans don’t really matter too much for me to fix them. In addition to these changes, I’m closer to fixing a bug in perspective streak!

 

See here.

 

 

Left: Original Middle: recpoltrans -3 to recpoltrans 2 Right: Xor Analysis

 

[gmic]-2./ Elapsed time: 0.19 s.

 

Also, saving code here:

[Reptillian]

I'm gonna put the rep_recpoltrans on hold for now.

 

I'm working on mode-related image processing. Speaking of which, does anyone know that thread where a person asked to fill image based on mode of colors? I really would like to mention that person even if it fruitless now.

 

In the spoiler, there are 3 cases, and how the mode are calculated are different.

 

Case IMGS>2 : Find the mode of all images within the same pixel coordinates. Outputs nan if multiple modes are found or no modes are found.

Case IMGS==2: Checks if pixels are the same. If they are, output the current pixel, otherwise nan.

Case IMGS==1: Find all the mode within a image. If there are no mode, do nothing. If there are multiple modes, then you can choose which modes to choose or output a nan if no input is detected.

 

Spoiler

rep_mode_color:
if $!>2
 +a z colormap. 0
 a[^-1] c
 {w#0},{h#0},{d#0},{s#1},"begin(
  const vs=w#1;
  const ss=s#1;
  const css=s#0/ss;
  endnan=vectors(nan);
 );
 nmaxfreq=0;
 vfreq=vectorvs(-1);
 vp=I#0;
 for(pos=0,pos<css,pos++,
  curcol=vp[pos*ss,ss];
  for(posfreq=0,posfreq<vs,posfreq++,
   coltestpos=I(#1,posfreq,0,0);
   if(curcol==coltestpos,vfreq[posfreq]++;);
  );
 );
 maxfreqcount=max(vfreq);
 if(maxfreqcount+1,
  posmaxfreqcol=find(vfreq,maxfreqcount,0,1);
  for(fs=0,fs<vs,fs++,if(vfreq[fs]==maxfreqcount,nmaxfreq++;););
  if(nmaxfreq==1,I(#1,posmaxfreqcol,0,0);,endnan);
 ,endnan;
 );"
 k.
elif $!==2
 f. "begin(endnan=vectors(nan););I==I#0?I:endnan;" rm..
else
 minv={im}
 - $minv
 
 repeat s
  sh. $>
  bsize$>={iM#-1}
  rm.
 done 
 
 if s==1
  $bsize0,1,1,1 eval.. ++i(#-1,i,0,0,0)
 elif s==2
  $bsize0,$bsize1,1,1 eval.. ++i(#-1,I,0)
 elif s==3
  $bsize0,$bsize1,$bsize2,1 eval.. ++i(#-1,I)
 else
  $bsize0,$bsize1,$bsize2,$bsize3 eval.. ++i(#-1,I)
 fi
fi

 

 

EDIT: Found the thread -

That was the thread that inspired me to do this.

 

Also, tagging @FanofSMBX
even if pointless.

Edited February 11, 2021 by Reptillian

[Reptillian]

I finished with the mode filter. I tested this code

 

$ sp cat rep_dupsdaxy 12,50,20,1,1,0 rep_mode_color ,, {w},{h},{d},1,!isnan(i#0)?255 replace_nan.. 0 a c drgba

 

Potential for a new artistic filter. It's pretty slow though because you are calculating the mode as a blending mode.

[Reptillian]

I haven't really managed to get new filters though I was experimenting for a bit. Sorry about that.

 

Can anyone take a guess at what this is?

 

[midora]

Looks like a special kind of radial gradient bur maybe I'm totally wrong...

[Reptillian]

5 hours ago, midora said:

Looks like a special kind of radial gradient bur maybe I'm totally wrong...

It's a Henon Phase Diagram - http://paulbourke.net/fractals/henonphase/

 

I thought these shapes are interesting to add as a filter. Even more useful than my Chirikov-Taylor filter.

[Reptillian]

New Filter! Henon Phase Diagram

 

EDIT: Addressed all issues with Henon Phase Diagram filter. Check in 1 hour or so, and then you'll see a good version of Henon Phase Diagram.

Edited March 15, 2021 by Reptillian

[Reptillian]

Bad news! Henon Phase Diagram has been disabled. Bluescreen crashes, so I was forced to disable it until it get looked at.

[Reptillian]

I restored it with a measure to hopefully prevent BSOD crash.

 

If it happens to you with Henon Phase Diagram, don't hesitate to report it as it's going to need to be blocked again.

 

Details from github:

 

I'm going with these measures to avoid BSOD crash:
1. Added if inf,-inf break line. I theoretize that this is the main reason for the crashes.
2. Limited X0 Increments.
3. And test some more to see if it needs to be blocked again.

 

Clarification on 1: I believe that it was using inf, and -inf, nan, -nan weirdly on multiple threads at once running as fast as possible, and that's why I get BSOD error, so I added the break line in hope to resolve that.

Edited March 16, 2021 by Reptillian

[Reptillian]

It seems that Henon Phase Diagram has been working. Had not caught a BSOD problem.

 

That being said, not much update for now. I will point out that I had split Overload Count in Thorn Fractal into 2 different variables. So, you only have positive number on overload count, and a checkmark option to switch to negative overload mode. This makes it easier to switch from positive to negative, and vice versa.

[Reptillian]

Added new filter! Skew!

 

 

Thanks to @BoltBait for inspiration, and @mszlazak for the idea of using angle.

 

It should be here in a hour or two. Probably here by the time you're reading this comment. So check by.

 

Commit: https://github.com/dtschump/gmic-community/commit/9ae1f91f611840cafc893b8c85e08aada62d39b5

Edited April 16, 2021 by Reptillian