Package helma.image

Source Code of helma.image.ColorQuantizer$Node

/*
* Helma License Notice
*
* The contents of this file are subject to the Helma License
* Version 2.0 (the "License"). You may not use this file except in
* compliance with the License. A copy of the License is available at
* http://adele.helma.org/download/helma/license.txt
*
* Copyright 1998-2003 Helma Software. All Rights Reserved.
*
* $RCSfile$
* $Author: lehni $
* $Revision: 9967 $
* $Date: 2009-09-25 12:57:50 +0200 (Fre, 25. Sep 2009) $
*/

package helma.image;

import java.awt.AlphaComposite;
import java.awt.Graphics2D;
import java.awt.image.BufferedImage;
import java.awt.image.DataBufferByte;
import java.awt.image.DataBufferInt;
import java.awt.image.IndexColorModel;

/*
* Modifications by Juerg Lehni:
*
* - Ported to Java from C
* - Support for alpha-channels.
* - Returns a BufferedImage of TYPE_BYTE_INDEXED with a IndexColorModel.
* - Dithering of images through helma.image.DiffusionFilterOp by setting
*   the dither parameter to true.
* - Support for a transparent color, which is correctly rendered by GIFEncoder.
*   All pixels with alpha < 0x80 are converted to this color when the parameter
*   alphaToBitmask is set to true.
*/
/*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%                                                                             %
%                                                                             %
%                                                                             %
%           QQQ   U   U   AAA   N   N  TTTTT  IIIII   ZZZZZ  EEEEE            %
%          Q   Q  U   U  A   A  NN  N    T      I        ZZ  E                %
%          Q   Q  U   U  AAAAA  N N N    T      I      ZZZ   EEEEE            %
%          Q  QQ  U   U  A   A  N  NN    T      I     ZZ     E                %
%           QQQQ   UUU   A   A  N   N    T    IIIII   ZZZZZ  EEEEE            %
%                                                                             %
%                                                                             %
%         Methods to Reduce the Number of Unique Colors in an Image           %
%                                                                             %
%                                                                             %
%                           Software Design                                   %
%                             John Cristy                                     %
%                              July 1992                                      %
%                                                                             %
%                                                                             %
%  Copyright (C) 2003 ImageMagick Studio, a non-profit organization dedicated %
%  to making software imaging solutions freely available.                     %
%                                                                             %
%  Permission is hereby granted, free of charge, to any person obtaining a    %
%  copy of this software and associated documentation files ("ImageMagick"),  %
%  to deal in ImageMagick without restriction, including without limitation   %
%  the rights to use, copy, modify, merge, publish, distribute, sublicense,   %
%  and/or sell copies of ImageMagick, and to permit persons to whom the       %
%  ImageMagick is furnished to do so, subject to the following conditions:    %
%                                                                             %
%  The above copyright notice and this permission notice shall be included in %
%  all copies or substantial portions of ImageMagick.                         %
%                                                                             %
%  The software is provided "as is", without warranty of any kind, express or %
%  implied, including but not limited to the warranties of merchantability,   %
%  fitness for a particular purpose and noninfringement.  In no event shall   %
%  ImageMagick Studio be liable for any claim, damages or other liability,    %
%  whether in an action of contract, tort or otherwise, arising from, out of  %
%  or in connection with ImageMagick or the use or other dealings in          %
%  ImageMagick.                                                               %
%                                                                             %
%  Except as contained in this notice, the name of the ImageMagick Studio     %
%  shall not be used in advertising or otherwise to promote the sale, use or  %
%  other dealings in ImageMagick without prior written authorization from the %
%  ImageMagick Studio.                                                        %
%                                                                             %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
%  Realism in computer graphics typically requires using 24 bits/pixel to
%  generate an image.  Yet many graphic display devices do not contain the
%  amount of memory necessary to match the spatial and color resolution of
%  the human eye.  The Quantize methods takes a 24 bit image and reduces
%  the number of colors so it can be displayed on raster device with less
%  bits per pixel.  In most instances, the quantized image closely
%  resembles the original reference image.
%
%  A reduction of colors in an image is also desirable for image
%  transmission and real-time animation.
%
%  QuantizeImage() takes a standard RGB or monochrome images and quantizes
%  them down to some fixed number of colors.
%
%  For purposes of color allocation, an image is a set of n pixels, where
%  each pixel is a point in RGB space.  RGB space is a 3-dimensional
%  vector space, and each pixel, Pi,  is defined by an ordered triple of
%  red, green, and blue coordinates, (Ri, Gi, Bi).
%
%  Each primary color component (red, green, or blue) represents an
%  intensity which varies linearly from 0 to a maximum value, Cmax, which
%  corresponds to full saturation of that color.  Color allocation is
%  defined over a domain consisting of the cube in RGB space with opposite
%  vertices at (0,0,0) and (Cmax, Cmax, Cmax).  QUANTIZE requires Cmax =
%  255.
%
%  The algorithm maps this domain onto a tree in which each node
%  represents a cube within that domain.  In the following discussion
%  these cubes are defined by the coordinate of two opposite vertices:
%  The vertex nearest the origin in RGB space and the vertex farthest from
%  the origin.
%
%  The tree's root node represents the the entire domain, (0,0,0) through
%  (Cmax,Cmax,Cmax).  Each lower level in the tree is generated by
%  subdividing one node's cube into eight smaller cubes of equal size.
%  This corresponds to bisecting the parent cube with planes passing
%  through the midpoints of each edge.
%
%  The basic algorithm operates in three phases: Classification,
%  Reduction, and Assignment.  Classification builds a color description
%  tree for the image.  Reduction collapses the tree until the number it
%  represents, at most, the number of colors desired in the output image.
%  Assignment defines the output image's color map and sets each pixel's
%  color by restorage_class in the reduced tree.  Our goal is to minimize
%  the numerical discrepancies between the original colors and quantized
%  colors (quantization error).
%
%  Classification begins by initializing a color description tree of
%  sufficient depth to represent each possible input color in a leaf.
%  However, it is impractical to generate a fully-formed color description
%  tree in the storage_class phase for realistic values of Cmax.  If
%  colors components in the input image are quantized to k-bit precision,
%  so that Cmax= 2k-1, the tree would need k levels below the root node to
%  allow representing each possible input color in a leaf.  This becomes
%  prohibitive because the tree's total number of nodes is 1 +
%  sum(i=1, k, 8k).
%
%  A complete tree would require 19,173,961 nodes for k = 8, Cmax = 255.
%  Therefore, to avoid building a fully populated tree, QUANTIZE: (1)
%  Initializes data structures for nodes only as they are needed;  (2)
%  Chooses a maximum depth for the tree as a function of the desired
%  number of colors in the output image (currently log2(colormap size)).
%
%  For each pixel in the input image, storage_class scans downward from
%  the root of the color description tree.  At each level of the tree it
%  identifies the single node which represents a cube in RGB space
%  containing the pixel's color.  It updates the following data for each
%  such node:
%
%    n1: Number of pixels whose color is contained in the RGB cube which
%    this node represents;
%
%    n2: Number of pixels whose color is not represented in a node at
%    lower depth in the tree;  initially,  n2 = 0 for all nodes except
%    leaves of the tree.
%
%    Sr, Sg, Sb: Sums of the red, green, and blue component values for all
%    pixels not classified at a lower depth. The combination of these sums
%    and n2  will ultimately characterize the mean color of a set of
%    pixels represented by this node.
%
%    E: The distance squared in RGB space between each pixel contained
%    within a node and the nodes' center.  This represents the
%    quantization error for a node.
%
%  Reduction repeatedly prunes the tree until the number of nodes with n2
%  > 0 is less than or equal to the maximum number of colors allowed in
%  the output image.  On any given iteration over the tree, it selects
%  those nodes whose E count is minimal for pruning and merges their color
%  statistics upward. It uses a pruning threshold, Ep, to govern node
%  selection as follows:
%
%    Ep = 0
%    while number of nodes with (n2 > 0) > required maximum number of colors
%      prune all nodes such that E <= Ep
%      Set Ep to minimum E in remaining nodes
%
%  This has the effect of minimizing any quantization error when merging
%  two nodes together.
%
%  When a node to be pruned has offspring, the pruning procedure invokes
%  itself recursively in order to prune the tree from the leaves upward.
%  n2,  Sr, Sg,  and  Sb in a node being pruned are always added to the
%  corresponding data in that node's parent.  This retains the pruned
%  node's color characteristics for later averaging.
%
%  For each node, n2 pixels exist for which that node represents the
%  smallest volume in RGB space containing those pixel's colors.  When n2
%  > 0 the node will uniquely define a color in the output image. At the
%  beginning of reduction,  n2 = 0  for all nodes except a the leaves of
%  the tree which represent colors present in the input image.
%
%  The other pixel count, n1, indicates the total number of colors within
%  the cubic volume which the node represents.  This includes n1 - n2
%  pixels whose colors should be defined by nodes at a lower level in the
%  tree.
%
%  Assignment generates the output image from the pruned tree.  The output
%  image consists of two parts: (1)  A color map, which is an array of
%  color descriptions (RGB triples) for each color present in the output
%  image;  (2)  A pixel array, which represents each pixel as an index
%  into the color map array.
%
%  First, the assignment phase makes one pass over the pruned color
%  description tree to establish the image's color map.  For each node
%  with n2  > 0, it divides Sr, Sg, and Sb by n2 .  This produces the mean
%  color of all pixels that classify no lower than this node.  Each of
%  these colors becomes an entry in the color map.
%
%  Finally,  the assignment phase reclassifies each pixel in the pruned
%  tree to identify the deepest node containing the pixel's color.  The
%  pixel's value in the pixel array becomes the index of this node's mean
%  color in the color map.
%
%  This method is based on a similar algorithm written by Paul Raveling.
%
%
*/

public class ColorQuantizer {
   public static final int MAX_NODES = 266817;
   public static final int MAX_TREE_DEPTH = 8;
   public static final int MAX_CHILDREN = 16;
   public static final int MAX_RGB = 255;

   static class ClosestColor {
       int distance;
       int colorIndex;
   }

   static class Node {
       Cube cube;
       Node parent;
       Node children[];
       int numChildren;

       int id;
       int level;

       int uniqueCount;

       int totalRed;
       int totalGreen;
       int totalBlue;
       int totalAlpha;
       long quantizeError;

       int colorIndex;

       Node(Cube cube) {
           this(cube, 0, 0, null);
           this.parent = this;
       }

       Node(Cube cube, int id, int level, Node parent) {
           this.cube = cube;
           this.parent = parent;
           this.id = id;
           this.level = level;
           this.children = new Node[MAX_CHILDREN];
           this.numChildren = 0;
           if (parent != null) {
               parent.children[id] = this;
               parent.numChildren++;
           }
           cube.numNodes++;
       }

       void pruneLevel() {
           // Traverse any children.
           if (numChildren > 0)
               for (int id = 0; id < MAX_CHILDREN; id++)
                   if (children[id] != null)
                       children[id].pruneLevel();
           if (level == cube.depth)
               prune();
       }

       void pruneToCubeDepth() {
           // Traverse any children.
           if (numChildren > 0)
               for (int id = 0; id < MAX_CHILDREN; id++)
                   if (children[id] != null)
                       children[id].pruneToCubeDepth();
           if (level > cube.depth)
               prune();
       }

       void prune() {
           // Traverse any children.
           if (numChildren > 0)
               for (int id = 0; id < MAX_CHILDREN; id++)
                   if (children[id] != null)
                       children[id].prune();
           // Merge color statistics into parent.
           parent.uniqueCount += uniqueCount;
           parent.totalRed += totalRed;
           parent.totalGreen += totalGreen;
           parent.totalBlue += totalBlue;
           parent.totalAlpha += totalAlpha;
           parent.children[id] = null;
           parent.numChildren--;
           cube.numNodes--;
       }

       void reduce(long pruningThreshold) {
           // Traverse any children.
           if (numChildren > 0)
               for (int id = 0; id < MAX_CHILDREN; id++)
                   if (children[id] != null)
                       children[id].reduce(pruningThreshold);
           if (quantizeError <= pruningThreshold)
               prune();
           else {
               // Find minimum pruning threshold.
               if (uniqueCount > 0)
                   cube.numColors++;
               if (quantizeError < cube.nextThreshold)
                   cube.nextThreshold = quantizeError;
           }
       }

       void findClosestColor(int red, int green, int blue, int alpha, ClosestColor closest) {
           // Traverse any children.
           if (numChildren > 0)
               for (int id = 0; id < MAX_CHILDREN; id++)
                   if (children[id] != null)
                       children[id].findClosestColor(red, green, blue, alpha, closest);
           if (uniqueCount != 0) {
               // Determine if this color is "closest".
               int dr = (cube.colorMap[0][colorIndex] & 0xff) - red;
               int dg = (cube.colorMap[1][colorIndex] & 0xff) - green;
               int db = (cube.colorMap[2][colorIndex] & 0xff) - blue;
               int da = (cube.colorMap[3][colorIndex] & 0xff) - alpha;
               int distance = da * da + dr * dr + dg * dg + db * db;
               if (distance < closest.distance) {
                   closest.distance = distance;
                   closest.colorIndex = colorIndex;
               }
            }
       }

       int fillColorMap(byte colorMap[][], int index) {
           // Traverse any children.
           if (numChildren > 0)
               for (int id = 0; id < MAX_CHILDREN; id++)
                   if (children[id] != null)
                       index = children[id].fillColorMap(colorMap, index);
           if (uniqueCount != 0) {
               // Colormap entry is defined by the mean color in this cube.
               colorMap[0][index] = (byte) (totalRed / uniqueCount + 0.5);
               colorMap[1][index] = (byte) (totalGreen / uniqueCount + 0.5);
               colorMap[2][index] = (byte) (totalBlue / uniqueCount + 0.5);
               colorMap[3][index] = (byte) (totalAlpha / uniqueCount + 0.5);
               colorIndex = index++;
           }
           return index;
       }
   }

   static class Cube {
       Node root;

       int numColors;
       boolean addTransparency;
       // firstColor is set to 1 when when addTransparency is true!
       int firstColor;
       byte colorMap[][];

       long nextThreshold;

       int numNodes;
       int depth;

       Cube(int maxColors) {
           this.depth = getDepth(maxColors);
           this.numColors = 0;
           root = new Node(this);
       }
      
       int getDepth(int numColors) {
           // Depth of color tree is: Log4(colormap size)+2.
           int depth;
           for (depth = 1; numColors != 0; depth++)
               numColors >>= 2;
           if (depth > MAX_TREE_DEPTH)
               depth = MAX_TREE_DEPTH;
           if (depth < 2)
               depth = 2;
           return depth;
       }

       void classifyImageColors(BufferedImage image, boolean alphaToBitmask) {
           addTransparency = false;
           firstColor = 0;
          
           Node node, child;
           int x, px, y, index, level, id, count;
           int pixel, red, green, blue, alpha;
           int bisect, midRed, midGreen, midBlue, midAlpha;

           int width = image.getWidth();
           int height = image.getHeight();

           // Classify the first 256 colors to a tree depth of MAX_TREE_DEPTH.
           int levelThreshold = MAX_TREE_DEPTH;
           // create a BufferedImage of only 1 pixel height for fetching the rows
           // of the image in the correct format (ARGB)
           // This speeds up things by more than factor 2, compared to the standard
           // BufferedImage.getRGB solution
           BufferedImage row = new BufferedImage(width, 1, BufferedImage.TYPE_INT_ARGB);
           Graphics2D g2d = row.createGraphics();
           int pixels[] = ((DataBufferInt) row.getRaster().getDataBuffer()).getData();
           // make sure alpha values do not add up for each row:
           g2d.setComposite(AlphaComposite.Src);
           // calculate scanline by scanline in order to safe memory.
           // It also seems to run faster like that
           for (y = 0; y < height; y++) {
               g2d.drawImage(image, null, 0, -y);
               // now pixels contains the rgb values of the row y!
               if (numNodes > MAX_NODES) {
                   // Prune one level if the color tree is too large.
                   root.pruneLevel();
                   depth--;
               }
               for (x = 0; x < width;) {
                   pixel = pixels[x];
                   red = (pixel >> 16) & 0xff;
                   green = (pixel >> 8) & 0xff;
                   blue = (pixel >> 0) & 0xff;
                   alpha = (pixel >> 24) & 0xff;
                   if (alphaToBitmask)
                       alpha = alpha < 0x80 ? 0 : 0xff;

                   // skip same pixels, but count them
                   px = x;
                   for (++x; x < width; x++)
                       if (pixels[x] != pixel)
                           break;
                   count = x - px;

                   // Start at the root and descend the color cube tree.
                   if (alpha > 0) {
                       index = MAX_TREE_DEPTH - 1;
                       bisect = (MAX_RGB + 1) >> 1;
                       midRed = bisect;
                       midGreen = bisect;
                       midBlue = bisect;
                       midAlpha = bisect;
                       node = root;
                       for (level = 1; level <= levelThreshold; level++) {
                           id = (((red >> index) & 0x01) << 3 |
                                 ((green >> index) & 0x01) << 2 |
                                 ((blue >> index) & 0x01) << 1 |
                                 ((alpha >> index) & 0x01));
                           bisect >>= 1;
                           midRed += (id & 8) != 0 ? bisect : -bisect;
                           midGreen += (id & 4) != 0 ? bisect : -bisect;
                           midBlue += (id & 2) != 0 ? bisect : -bisect;
                           midAlpha += (id & 1) != 0 ? bisect : -bisect;
                           child = node.children[id];
                           if (child == null) {
                               // Set colors of new node to contain pixel.
                               child = new Node(this, id, level, node);
                               if (level == levelThreshold) {
                                   numColors++;
                                   if (numColors == 256) {
                                       // More than 256 colors; classify to the
                                       // cube_info.depth tree depth.
                                       levelThreshold = depth;
                                       root.pruneToCubeDepth();
                                   }
                               }
                           }
                           // Approximate the quantization error represented by
                           // this node.
                           node = child;
                           int r = red - midRed;
                           int g = green - midGreen;
                           int b = blue - midBlue;
                           int a = alpha - midAlpha;
                           node.quantizeError += count * (r * r + g * g + b * b + a * a);
                           root.quantizeError += node.quantizeError;
                           index--;
                       }
                       // Sum RGB for this leaf for later derivation of the mean
                       // cube color.
                       node.uniqueCount += count;
                       node.totalRed += count * red;
                       node.totalGreen += count * green;
                       node.totalBlue += count * blue;
                       node.totalAlpha += count * alpha;
                   } else if (!addTransparency) {
                       addTransparency = true;
                       numColors++;
                       firstColor = 1; // start at 1 as 0 will be the transparent color
                   }
               }
           }
       }

       void reduceImageColors(int maxColors) {
           nextThreshold = 0;
           while (numColors > maxColors) {
               long pruningThreshold = nextThreshold;
               nextThreshold = root.quantizeError - 1;
               numColors = firstColor;
               root.reduce(pruningThreshold);
           }
       }

       BufferedImage assignImageColors(BufferedImage image, boolean dither, boolean alphaToBitmask) {
           // Allocate image colormap.
           colorMap = new byte[4][numColors];
           root.fillColorMap(colorMap, firstColor);
           // create the right color model, depending on transparency settings:
           IndexColorModel icm;
          
           int width = image.getWidth();
           int height = image.getHeight();
          
           if (alphaToBitmask) {
               if (addTransparency) {
                   icm = new IndexColorModel(depth, numColors, colorMap[0], colorMap[1], colorMap[2], 0);
               } else {
                   icm = new IndexColorModel(depth, numColors, colorMap[0], colorMap[1], colorMap[2]);
               }
           } else {
               icm = new IndexColorModel(depth, numColors, colorMap[0], colorMap[1], colorMap[2], colorMap[3]);
           }

           // create the indexed BufferedImage:
           BufferedImage dest = new BufferedImage(width, height, BufferedImage.TYPE_BYTE_INDEXED, icm);

           if (dither)
               new DiffusionFilterOp().filter(image, dest);
           else {
               ClosestColor closest = new ClosestColor();
               // convert to indexed color
               byte[] dst = ((DataBufferByte) dest.getRaster().getDataBuffer()).getData();

               // create a BufferedImage of only 1 pixel height for fetching
               // the rows of the image in the correct format (ARGB)
               // This speeds up things by more than factor 2, compared to the
               // standard BufferedImage.getRGB solution
               BufferedImage row = new BufferedImage(width, 1, BufferedImage.TYPE_INT_ARGB);
               Graphics2D g2d = row.createGraphics();
               int pixels[] = ((DataBufferInt) row.getRaster().getDataBuffer()).getData();
               // make sure alpha values do not add up for each row:
               g2d.setComposite(AlphaComposite.Src);
               // calculate scanline by scanline in order to safe memory.
               // It also seems to run faster like that
               Node node;
               int x, y, i, id;
               int pixel, red, green, blue, alpha;
               int pos = 0;
               for (y = 0; y < height; y++) {
                   g2d.drawImage(image, null, 0, -y);
                   // now pixels contains the rgb values of the row y!
                   // filter this row now:
                   for (x = 0; x < width;) {
                       pixel = pixels[x];
                       red = (pixel >> 16) & 0xff;
                       green = (pixel >> 8) & 0xff;
                       blue = (pixel >> 0) & 0xff;
                       alpha = (pixel >> 24) & 0xff;

                       if (alphaToBitmask)
                           alpha = alpha < 128 ? 0 : 0xff;

                       byte col;
                       if (alpha == 0 && addTransparency) {
                           col = 0; // transparency color is at position 0 of color map
                       } else {
                           // walk the tree to find the cube containing that
                           // color
                           node = root;
                           for (i = MAX_TREE_DEPTH - 1; i > 0; i--) {
                               id = (((red >> i) & 0x01) << 3 |
                                     ((green >> i) & 0x01) << 2 |
                                     ((blue >> i) & 0x01) << 1 |
                                     ((alpha >> i) & 0x01));
                               if (node.children[id] == null)
                                   break;
                               node = node.children[id];
                           }

                           // Find the closest color.
                           closest.distance = Integer.MAX_VALUE;
                           node.parent.findClosestColor(red, green, blue, alpha, closest);
                           col = (byte) closest.colorIndex;
                       }

                       // first color
                       dst[pos++] = col;
                      
                       // next colors the same?
                       for (++x; x < width; x++) {
                           if (pixels[x] != pixel)
                               break;
                           dst[pos++] = col;
                       }
                   }
               }
           }
           return dest;
       }
   }

   static BufferedImage quantizeImage(BufferedImage image, int maxColors, boolean dither, boolean alphaToBitmask) {
       Cube cube = new Cube(maxColors);
       cube.classifyImageColors(image, alphaToBitmask);
       cube.reduceImageColors(maxColors);
       return cube.assignImageColors(image, dither, alphaToBitmask);
   }

}
TOP

Related Classes of helma.image.ColorQuantizer$Node

TOP
Copyright © 2015 www.massapi.com. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.