Gluas

lylejk · **Joined:** Sep 24, 2010 **Posts:** 12531

It's just an un-written rule that if someone share's their photo for play purposes, that that said person owns all rights to and results create from said photo and you need permission from that person to share results on other forums. Anyway, decided to use a photo from the free photo index to show why I like Kuwahara. Used G'MIC Paint, Kuwahara twice and added some texture for the most part.

ref: http://www.fontplay.com/freephotos/seve ... 107-03.jpg

MareroQ · **Posted:** Mon May 30, 2011 12:29 pm **(#12)**

LYLE - thanks - for another interesting link.
Already difficult to count - how many times you've found interesting things ?
For the convenience of others in the annex (GLUAS SCRIPTS.ZIP) selected scripts to Gluas
(http://www.thebest3d.com/dogwaffle/lua/samples.html:
. . for your enjoyment. Grab'em and experiment, play with different parameters, etc....).
It's strange, but some work only with version 0.1.19

mahvin · **Joined:** Oct 06, 2010 **Posts:** 4050

None of the dogwaffle or GUI server scripts worked in gimp-gluas. At least not on my machine (Ubuntu).

lylejk · **Joined:** Sep 24, 2010 **Posts:** 12531

One of the best downsizers (not uprezzing) that I've run across is Decimator (bet someone can get somebody with G'MIC programming capability to do this one right). You have to crop and sharpen afterwards, but still, I'm pretty impressed witht the quality of this. Anyway, you are right Mahvin. Looks like any script that calls a dog(or something) subroutine will not work. A few do though.

mahvin · **Joined:** Oct 06, 2010 **Posts:** 4050

I only found one, that did work. Contrast sharpen. It had no references in the code to GUI or dogwaffle, though. Please share any that do work for you. I might be missing them inadvertently.

lylejk · **Joined:** Sep 24, 2010 **Posts:** 12531

Can't remember which ones I got to work from dog-lua Mahvin, but I want to say sharpen by luminosity was one of them. Anyway, I attached a screenshot of all the ones that I have thus far that work (that I tested anyway).

MareroQ · **Posted:** Mon May 30, 2011 2:44 pm **(#17)**

For Windows (XP 32-bit) with Gluas 0.1.19 [http://www.geocities.jp/gimproject/plug-ins/gluas.html] - all (of GLUAS SCRIPTS.ZIP) work well

lylejk · **Joined:** Sep 24, 2010 **Posts:** 12531

I didn't realize how to get the version of Gluas til I saw your post MareroQ; since samj's is 0.1.20, decided to try it and for a few of the presets I tested, it still works fine.

mahvin · **Joined:** Oct 06, 2010 **Posts:** 4050

My list is the same as yours, minus jitter.

mahvin · **Joined:** Oct 06, 2010 **Posts:** 4050

MareroQ:

Could you provide copies of your Windows version gluas, so I can test them with Ubuntu?

MareroQ · **Posted:** Mon May 30, 2011 10:41 pm **(#21)**

Maybe for Ubuntu, well worth a try version 0.1.19 ?
http://pippin.gimp.org/plug-ins/gluas/files/

mahvin · **Joined:** Oct 06, 2010 **Posts:** 4050

I have those. I noticed the Japanese site had a couple that gimp-gluas doesn't have.

Rod · **Joined:** May 16, 2010 **Posts:** 14709 **Location:** USA

Night Vision

Rod · **Joined:** May 16, 2010 **Posts:** 14709 **Location:** USA

Here is the Kuwahara filter with curves after ward set to a cone line.
Pretty cool grunge effect Lyle.

PhotoComix · **Joined:** Apr 12, 2010 **Posts:** 5870

Fix the dogwaflle scripts should not be hard , replace the "dog" calls with a equivalent "gimp" call should fix, but i wonder if is worth the effort

may be more interesting port script as Kuwahara to gmic or mathmap (the language seems a bit more similar to MM ) that for sure may support it.

just in case somebody interested here the code

SHOW SPOILER

Kuwahara filter

Code: Select all

-- origine : http://pippin.gimp.org/image_processing/chap_area.html#example_kuwahara


-- Kuwahara filter
--
-- 
-- Performs the Kuwahara Filter. This filter is an edge-preserving filter.
--
--
-- ( a  a  ab   b  b)
-- ( a  a  ab   b  b)
-- (ac ac abcd bd bd)
-- ( c  c  cd   d  d)
-- ( c  c  cd   d  d)
--
-- In each of the four regions (a, b, c, d), the mean brightness and the variance are calculated. The
-- output value of the center pixel (abcd) in the window is the mean value of that region that has the
-- smallest variance.
--
-- description copied from http://www.incx.nec.co.jp/imap-vision/library/wouter/kuwahara.html
--
-- implemented by Øyvind Kolås <oeyvindk@hig.no> 2004


-- the sampling window is:  width=size*2+1  height=size*2+1

size = 4

edge_duplicate = 1;

-- local function to get the mean value, and
-- variance from the rectangular area specified
function mean_and_variance (x0,y0,x1,y1)
 local variance
 local mean

 local min         = 1.0
 local max         = 0.0
 local accumulated = 0
 local count       = 0
   
 local x, y

 for y=y0,y1 do
   for x=x0,x1 do
     local v = get_value(x,y)

     accumulated   = accumulated + v
     count         = count + 1
      
     if v<min then min = v end
     if v>max then max = v end

    end
  end

  variance = max-min
  mean     = accumulated /count
   
  return mean, variance
end

-- local function to get the mean value, and
-- variance from the rectangular area specified
function rgb_mean_and_variance (x0,y0,x1,y1)
 local variance
 local mean
 local r_mean
 local g_mean
 local b_mean 
 
 local min           = 1.0
 local max           = 0.0
 local accumulated_r = 0
 local accumulated_g = 0
 local accumulated_b = 0
 local count         = 0
   
 local x, y

 for y=y0,y1 do
   for x=x0,x1 do
     local v     = get_value(x,y)
     local r,g,b = get_rgb (x,y)

     accumulated_r   = accumulated_r + r
     accumulated_g   = accumulated_g + g
     accumulated_b   = accumulated_b + b
     count         = count + 1
      
     if v<min then min = v end
     if v>max then max = v end

    end
  end

  variance = max-min
  mean_r   = accumulated_r /count
  mean_g   = accumulated_g /count
  mean_b   = accumulated_b /count
    
  return mean_r, mean_g, mean_b, variance
end

-- return the kuwahara computed value
function kuwahara(x, y, size)
  local best_mean     = 1.0
  local best_variance = 1.0

  local mean, variance
   
  mean, variance = mean_and_variance (x-size, y-size, x, y)   

  if variance < best_variance then
    best_mean     = mean
    best_variance = variance
  end
   
  mean, variance = mean_and_variance (x, y-size, x+size,y)

  if variance < best_variance then
    best_mean     = mean
    best_variance = variance
  end   
   
  mean, variance = mean_and_variance (x, y, x+size, y+size)

  if variance < best_variance then
    best_mean     = mean
    best_variance = variance
  end   
   
  mean, variance = mean_and_variance (x-size, y, x,y+size)

  if variance < best_variance then
    best_mean     = mean
    best_variance = variance
  end   

  return best_mean
end


-- return the kuwahara computed value
function rgb_kuwahara(x, y, size)
  local best_r, best_g, best_b
  local best_variance = 1.0

  local r,g,b, variance
   
  r,g,b, variance = rgb_mean_and_variance (x-size, y-size, x, y)   

  if variance < best_variance then
    best_r, best_g, best_b = r, g, b
    
    best_variance = variance
  end
   
  r,g,b, variance = rgb_mean_and_variance (x, y-size, x+size,y)

  if variance < best_variance then
    best_r, best_g, best_b = r, g, b
    best_variance = variance
  end   
   
  r,g,b, variance = rgb_mean_and_variance (x, y, x+size, y+size)

  if variance < best_variance then
    best_r, best_g, best_b = r, g, b
    best_variance = variance
  end   
   
  r,g,b, variance = rgb_mean_and_variance (x-size, y, x,y+size)

  if variance < best_variance then
    best_r, best_g, best_b = r, g, b
    best_variance = variance
  end   

  return best_r, best_g, best_b
end

function kuwahara(radius)
  for y=0, height-1 do
    for x=0, width-1 do
      r,g,b = rgb_kuwahara (x,y, radius)
      set_rgb (x,y, r,g,b)
    end
    progress (y/height)
  end
end

kuwahara(2)

Ronounours · **Joined:** Jan 03, 2011 **Posts:** 1656

Is there any code visible for the Kuwahara effect ? I think this could be done in G'MIC

mahvin · **Joined:** Oct 06, 2010 **Posts:** 4050

It's hiding behind PC's spoiler

PhotoComix · **Joined:** Apr 12, 2010 **Posts:** 5870

David maybe the original C script (well look as C to me , but i may be wrong) is easier to port.
The Lua filter was derieved by this

here #include <stddef.h>
#include <imapio.h>
#include <stdimap.h>

/*-----------------------------------------------------------------
Kuwahara Filter

12/01/99
filename: kuwahara.lc
author : W.J.Bokhove
----------------------------------------------------------------*/

/*
M_DOCDEF kuwahara

NAME kuwahara

SYNOPSIS
void kuwahara(src, dst, lines)
sep uchar src[]; /* source image */ /*
sep uchar dst[]; /* destination image */ /*
uint lines; /* number of lines to work on */ /*

MEMORY CONSUMPTION
This function consumes about 22 lines of IMAP memory

DESCRIPTION
Performs the Kuwahara Filter. This filter is an edge-preserving filter.
( a a ab b b)
( a a ab b b)
(ac ac abcd bd bd)
( c c cd d d)
( c c cd d d)
In each of the four regions (a, b, c, d), the mean brightness and the
variance are calculated. The output value of the center pixel (abcd) in
the window is the mean value of that region that has the smallest variance.

M_DOCEND
*/

void kuwahara(sep uchar src[], sep uchar dst[], uint lines)
{
#define square(a) (a*a)
int i;
sep ulong dif;
sep uint dtmp1, dtmp2, dtmp3, dtmp4, acc;
sep uchar var[3], mean[3], vartmp, meantmp;

dst[0] = 0;
dst[1] = 0;
/* Calculate Variance and Mean of first 2 lines */
dtmp2 = square((src[0] - :>src[1]));
dtmp3 = square((src[0] - src[1]));
dtmp4 = square((src[0] - :<src[1]));
dif = :<dtmp2 + dtmp3 + :>dtmp4;
dtmp1 = square((src[1] - :>src[1]));
dif += dtmp1;
dif += :<dtmp1;
dtmp2 = square((src[1] - :>src[2]));
dif += dtmp2;
dtmp3 = square((src[1] - src[2]));
dif += dtmp3;
dtmp4 = square((src[1] - :<src[2]));
dif += dtmp4;
var[1] = dif;
dif = :<dtmp2 + dtmp3 + :>dtmp4;
dtmp1 = square((src[2] - :>src[2]));
dif += dtmp1;
dif += :<dtmp1;
dtmp2 = square((src[2] - :>src[3]));
dif += dtmp2;
dtmp3 = square((src[2] - src[3]));
dif += dtmp3;
dtmp4 = square((src[2] - :<src[3]));
dif += dtmp4;
var[2] = dif;
acc = src[0] + src[1] + src[2];
acc += (:<acc + :>acc);
mean[1] = acc / 9;
acc = src[2] + src[2] + src[3];
acc += (:<acc + :>acc);
mean[2] = acc / 9;

for (i=3 ; i<lines-1; i++ ) {
/* Calculate Variance */
dif = :<dtmp2 + dtmp3 + :>dtmp4;
dtmp1 = square((src[i] - :>src[i]));
dif += dtmp1;
dif += :<dtmp1;
dtmp2 = square((src[i] - :>src[i+1]));
dif += dtmp2;
dtmp3 = square((src[i] - src[i+1]));
dif += dtmp3;
dtmp4 = square((src[i] - :<src[i+1]));
dif += dtmp4;
var[3]= sqrtsepi(dif>>3);
/* Calculate Mean */
acc = src[i-1] + src[i] + src[i+1];
acc += (:<acc + :>acc);
mean[3] = acc / 9;
/* Find minimum variance and its mean */
vartmp = :>var[1];
meantmp = :>mean[1];
mif (:<var[1] < vartmp) {meantmp = :<mean[1]; vartmp = :<var[1];}
mif (:>var[3] < vartmp) {meantmp = :>mean[3]; vartmp = :>var[3];}
mif (:<var[3] < vartmp) meantmp = :<mean[3];
dst[i-1] = meantmp;
var[1] = var[2]; mean[1] = mean[2];
var[2] = var[3]; mean[2] = mean[3];
}
dst[lines-2] = 0;
dst[lines-1] = 0;

From http://web.archive.org/web/200803130937 ... uwahara.lc

Ronounours · **Joined:** Jan 03, 2011 **Posts:** 1656

Ok, Thanks for the pointers. I've recoded it from scratch in G'MIC, it is quite short to perform in this beautiful language

#@gmic kuwahara : size>0
#@gmic : Apply Kuwahara filter of specified size on selected images.
#@gmic : $ image.jpg --kuwahara 5
kuwahara : -check $1>0
-e[^-1] "Apply Kuwahara filter of size $1 on image$?."
-v - -to_rgb -repeat @#
--dilate[0] $1 --erode[0] $1 -compose_channels[-1] min -compose_channels[-2] max --[-2,-1]
$1,1,1,1,{1/$1} -convolve[0] [-1] -transpose[-1] -convolve[0] [-1] -rm[-1]
-a[-2,-1] c
-f "v1=i(x-1,y-1,0,3,1); \
v2=i(x+1,y+1,0,3,1); \
v3=i(x-1,y+1,0,3,1); \
v4=i(x+1,y+1,0,3,1); \
vm=min(v1,v2,v3,v4); \
if(c>=3,i, \
if(vm==v1,i(x-1,y-1,0,c,1),
if(vm==v2,i(x+1,y-1,0,c,1),
if(vm==v3,i(x-1,y+1,0,c,1),
i(x+1,y+1,0,c,1)))))"
-channels[-1] 0,2
-mv[-1] 0 -done -v +

#@gimp Kuwahara filtering : gimp_kuwahara, gimp_kuwahara_preview(0)
#@gimp : Radius = int(3,1,30)
#@gimp : Channel(s) = choice("All","RGBA","RGB","Luminance","Blue/red chrominances","Blue chrominance","Red chrominance","Lightness","ab-components","a-component","b-component","Hue","Saturation","Value","Key","Alpha","ch-components","c-component","h-component")
#@gimp : sep = separator(), Preview type = choice("Full","Forward horizontal","Forward vertical","Backward horizontal","Backward vertical")
#@gimp : sep = separator(), note = note("Author : David Tschumperlé. Latest update : 2011/05/31.")
gimp_kuwahara :
-apply_channels "-kuwahara $1",$2,0

gimp_kuwahara_preview :
-gimp_split_preview "-gimp_kuwahara ${1--2}",$-1

I got this :

Should be already available after filter update.

PhotoComix · **Joined:** Apr 12, 2010 **Posts:** 5870

thank David

I am wrong or the original filter was basically equivalent to a Convolution Matrix preset ?
Something as the many presets (erode ,dilate,highpass,etc) that may be created by
http://docs.gimp.org/2.6/en/plug-in-convmatrix.html ?