/*
* $RCSfile: NeuQuantOpImage.java,v $
*
* Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
*
* Use is subject to license terms.
*
* $Revision: 1.2 $
* $Date: 2005/05/10 01:03:23 $
* $State: Exp $
*/
package com.lightcrafts.media.jai.opimage;
import java.awt.Rectangle;
import java.awt.image.RenderedImage;
import java.util.Map;
import com.lightcrafts.mediax.jai.ImageLayout;
import com.lightcrafts.mediax.jai.LookupTableJAI;
import com.lightcrafts.mediax.jai.PlanarImage;
import com.lightcrafts.mediax.jai.iterator.RandomIter;
import com.lightcrafts.mediax.jai.iterator.RandomIterFactory;
import com.lightcrafts.mediax.jai.ROI;
/**
* An <code>OpImage</code> implementing the "ColorQuantizer" operation as
* described in <code>com.lightcrafts.mediax.jai.operator.ExtremaDescriptor</code>
* based on the median-cut algorithm.
*
* This is based on a java-version of Anthony Dekker's implementation of
* NeuQuant Neural-Net Quantization Algorithm
*
* NEUQUANT Neural-Net quantization algorithm by Anthony Dekker, 1994.
* See "Kohonen neural networks for optimal colour quantization"
* in "Network: Computation in Neural Systems" Vol. 5 (1994) pp 351-367.
* for a discussion of the algorithm.
*
* Any party obtaining a copy of these files from the author, directly or
* indirectly, is granted, free of charge, a full and unrestricted irrevocable,
* world-wide, paid up, royalty-free, nonexclusive right and license to deal
* in this software and documentation files (the "Software"), including without
* limitation the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons who receive
* copies from any such party to do so, with the only requirement being
* that this copyright notice remain intact.
*
* @see com.lightcrafts.mediax.jai.operator.ExtremaDescriptor
* @see ExtremaCRIF
*/
public class NeuQuantOpImage extends ColorQuantizerOpImage {
/** four primes near 500 - assume no image has a length so large
* that it is divisible by all four primes
*/
protected static final int prime1 = 499;
protected static final int prime2 = 491;
protected static final int prime3 = 487;
protected static final int prime4 = 503;
/* minimum size for input image */
protected static final int minpicturebytes = (3 * prime4);
/** The size of the histogram. */
private int ncycles;
/* Program Skeleton
----------------
[select samplefac in range 1..30]
[read image from input file]
pic = (unsigned char*) malloc(3*width*height);
initnet(pic,3*width*height,samplefac);
learn();
unbiasnet();
[write output image header, using writecolourmap(f)]
inxbuild();
write output image using inxsearch(b,g,r) */
/* Network Definitions
------------------- */
private final int maxnetpos = maxColorNum - 1;
private final int netbiasshift = 4; /* bias for colour values */
/* defs for freq and bias */
private final int intbiasshift = 16; /* bias for fractions */
private final int intbias = 1 << intbiasshift;
private final int gammashift = 10; /* gamma = 1024 */
private final int gamma = 1 << gammashift;
private final int betashift = 10;
private final int beta = intbias >> betashift; /* beta = 1/1024 */
private final int betagamma = intbias << (gammashift - betashift);
/* defs for decreasing radius factor */
private final int initrad = maxColorNum >> 3;
private final int radiusbiasshift = 6; /* at 32.0 biased by 6 bits */
private final int radiusbias = 1 << radiusbiasshift;
private final int initradius = initrad * radiusbias; /* and decreases by a */
private final int radiusdec = 30; /* factor of 1/30 each cycle */
/* defs for decreasing alpha factor */
private final int alphabiasshift = 10; /* alpha starts at 1.0 */
private final int initalpha = 1 << alphabiasshift;
private int alphadec; /* biased by 10 bits */
/* radbias and alpharadbias used for radpower calculation */
private final int radbiasshift = 8;
private final int radbias = 1 << radbiasshift;
private final int alpharadbshift = alphabiasshift + radbiasshift;
private final int alpharadbias = 1 << alpharadbshift;
// typedef int pixel[4]; /* BGRc */
private int[][] network; /* the network itself - [maxColorNum][4] */
private int[] netindex = new int[256]; /* for network lookup - really 256 */
private int[] bias = new int[maxColorNum]; /* bias and freq arrays for learning */
private int[] freq = new int[maxColorNum];
private int[] radpower = new int[initrad]; /* radpower for precomputation */
/**
* Constructs an <code>NeuQuantOpImage</code>.
*
* @param source The source image.
*/
public NeuQuantOpImage(RenderedImage source,
Map config,
ImageLayout layout,
int maxColorNum,
int upperBound,
ROI roi,
int xPeriod,
int yPeriod) {
super(source, config, layout, maxColorNum, roi, xPeriod, yPeriod);
colorMap = null;
this.ncycles = upperBound;
}
protected synchronized void train() {
// intialize the network
network = new int[maxColorNum][];
for (int i = 0; i < maxColorNum; i++) {
network[i] = new int[4];
int[] p = network[i];
p[0] = p[1] = p[2] = (i << (netbiasshift + 8)) / maxColorNum;
freq[i] = intbias / maxColorNum; /* 1/maxColorNum */
bias[i] = 0;
}
PlanarImage source = getSourceImage(0);
Rectangle rect = source.getBounds();
if (roi != null)
rect = roi.getBounds();
RandomIter iterator = RandomIterFactory.create(source, rect);
int samplefac = xPeriod * yPeriod;
int startX = rect.x / xPeriod;
int startY = rect.y / yPeriod;
int offsetX = rect.x % xPeriod;
int offsetY = rect.y % yPeriod;
int pixelsPerLine = (rect.width - 1) / xPeriod + 1;
int numSamples =
pixelsPerLine * ((rect.height - 1) / yPeriod + 1);
if (numSamples < minpicturebytes)
samplefac = 1;
alphadec = 30 + ((samplefac - 1) / 3);
int pix = 0;
int delta = numSamples / ncycles;
int alpha = initalpha;
int radius = initradius;
int rad = radius >> radiusbiasshift;
if (rad <= 1)
rad = 0;
for (int i = 0; i < rad; i++)
radpower[i] = alpha * (((rad * rad - i * i) * radbias) / (rad * rad));
int step;
if (numSamples < minpicturebytes)
step = 3;
else if ((numSamples % prime1) != 0)
step = 3 * prime1;
else {
if ((numSamples % prime2) != 0)
step = 3 * prime2;
else {
if ((numSamples % prime3) != 0)
step = 3 * prime3;
else
step = 3 * prime4;
}
}
int[] pixel = new int[3];
for (int i = 0; i < numSamples;) {
int y = (pix / pixelsPerLine + startY) * yPeriod + offsetY;
int x = (pix % pixelsPerLine + startX) * xPeriod + offsetX;
try {
iterator.getPixel(x, y, pixel);
} catch (Exception e) {
continue;
}
int b = pixel[2] << netbiasshift;
int g = pixel[1] << netbiasshift;
int r = pixel[0] << netbiasshift;
int j = contest(b , g, r);
altersingle(alpha, j, b , g, r);
if (rad != 0)
alterneigh(rad, j, b , g, r); /* alter neighbours */
pix += step;
if (pix >= numSamples)
pix -= numSamples;
i++;
if (i % delta == 0) {
alpha -= alpha / alphadec;
radius -= radius / radiusdec;
rad = radius >> radiusbiasshift;
if (rad <= 1)
rad = 0;
for (j = 0; j < rad; j++)
radpower[j] = alpha * (((rad * rad - j * j) * radbias) / (rad * rad));
}
}
unbiasnet();
inxbuild();
createLUT();
setProperty("LUT", colorMap);
setProperty("JAI.LookupTable", colorMap);
}
private void createLUT() {
colorMap = new LookupTableJAI(new byte[3][maxColorNum]);
byte[][] map = colorMap.getByteData();
int[] index = new int[maxColorNum];
for (int i = 0; i < maxColorNum; i++)
index[network[i][3]] = i;
for (int i = 0; i < maxColorNum; i++) {
int j = index[i];
map[2][i] = (byte) (network[j][0]);
map[1][i] = (byte) (network[j][1]);
map[0][i] = (byte) (network[j][2]);
}
}
/** Insertion sort of network and building of netindex[0..255]
* (to do after unbias)
*/
private void inxbuild() {
int previouscol = 0;
int startpos = 0;
for (int i = 0; i < maxColorNum; i++) {
int[] p = network[i];
int smallpos = i;
int smallval = p[1]; /* index on g */
/* find smallest in i..maxColorNum-1 */
int j;
for (j = i + 1; j < maxColorNum; j++) {
int[] q = network[j];
if (q[1] < smallval) { /* index on g */
smallpos = j;
smallval = q[1]; /* index on g */
}
}
int[] q = network[smallpos];
/* swap p (i) and q (smallpos) entries */
if (i != smallpos) {
j = q[0]; q[0] = p[0]; p[0] = j;
j = q[1]; q[1] = p[1]; p[1] = j;
j = q[2]; q[2] = p[2]; p[2] = j;
j = q[3]; q[3] = p[3]; p[3] = j;
}
/* smallval entry is now in position i */
if (smallval != previouscol) {
netindex[previouscol] = (startpos + i) >> 1;
for (j = previouscol + 1; j < smallval; j++)
netindex[j] = i;
previouscol = smallval;
startpos = i;
}
}
netindex[previouscol] = (startpos + maxnetpos) >> 1;
for (int j = previouscol + 1; j < 256; j++)
netindex[j] = maxnetpos; /* really 256 */
}
/** Search for BGR values 0..255 (after net is unbiased) and
* return colour index
*/
protected byte findNearestEntry(int r, int g, int b) {
int bestd = 1000; /* biggest possible dist is 256*3 */
int best = -1;
int i = netindex[g]; /* index on g */
int j = i - 1; /* start at netindex[g] and work outwards */
while (i < maxColorNum || j >= 0) {
if (i < maxColorNum) {
int[] p = network[i];
int dist = p[1] - g; /* inx key */
if (dist >= bestd)
i = maxColorNum; /* stop iter */
else {
i++;
if (dist < 0)
dist = -dist;
int a = p[0] - b;
if (a < 0)
a = -a;
dist += a;
if (dist < bestd) {
a = p[2] - r;
if (a < 0)
a = -a;
dist += a;
if (dist < bestd) {
bestd = dist;
best = p[3];
}
}
}
}
if (j >= 0) {
int[] p = network[j];
int dist = g - p[1]; /* inx key - reverse dif */
if (dist >= bestd)
j = -1; /* stop iter */
else {
j--;
if (dist < 0)
dist = -dist;
int a = p[0] - b;
if (a < 0)
a = -a;
dist += a;
if (dist < bestd) {
a = p[2] - r;
if (a < 0)
a = -a;
dist += a;
if (dist < bestd) {
bestd = dist;
best = p[3];
}
}
}
}
}
return (byte)best;
}
/** Unbias network to give byte values 0..255 and record
* position i to prepare for sort.
*/
private void unbiasnet() {
for (int i = 0; i < maxColorNum; i++) {
network[i][0] >>= netbiasshift;
network[i][1] >>= netbiasshift;
network[i][2] >>= netbiasshift;
network[i][3] = i; /* record colour no */
}
}
/** Move adjacent neurons by precomputed
* alpha*(1-((i-j)^2/[r]^2)) in radpower[|i-j|]
*/
private void alterneigh(int rad, int i, int b, int g, int r) {
int lo = i - rad;
if (lo < -1)
lo = -1;
int hi = i + rad;
if (hi > maxColorNum)
hi = maxColorNum;
int j = i + 1;
int k = i - 1;
int m = 1;
while ((j < hi) || (k > lo)) {
int a = radpower[m++];
if (j < hi) {
int[] p = network[j++];
// try {
p[0] -= (a * (p[0] - b)) / alpharadbias;
p[1] -= (a * (p[1] - g)) / alpharadbias;
p[2] -= (a * (p[2] - r)) / alpharadbias;
// } catch (Exception e) {} // prevents 1.3 miscompilation
}
if (k > lo) {
int[] p = network[k--];
// try {
p[0] -= (a * (p[0] - b)) / alpharadbias;
p[1] -= (a * (p[1] - g)) / alpharadbias;
p[2] -= (a * (p[2] - r)) / alpharadbias;
// } catch (Exception e) {}
}
}
}
/** Move neuron i towards biased (b,g,r) by factor alpha. */
private void altersingle(int alpha, int i, int b, int g, int r) {
/* alter hit neuron */
int[] n = network[i];
n[0] -= (alpha * (n[0] - b)) / initalpha;
n[1] -= (alpha * (n[1] - g)) / initalpha;
n[2] -= (alpha * (n[2] - r)) / initalpha;
}
/** Search for biased BGR values. */
private int contest(int b, int g, int r) {
/* finds closest neuron (min dist) and updates freq */
/* finds best neuron (min dist-bias) and returns position */
/* for frequently chosen neurons, freq[i] is high and bias[i] is negative */
/* bias[i] = gamma*((1/maxColorNum)-freq[i]) */
int bestd = ~(((int) 1) << 31);
int bestbiasd = bestd;
int bestpos = -1;
int bestbiaspos = bestpos;
for (int i = 0; i < maxColorNum; i++) {
int[] n = network[i];
int dist = n[0] - b;
if (dist < 0)
dist = -dist;
int a = n[1] - g;
if (a < 0)
a = -a;
dist += a;
a = n[2] - r;
if (a < 0)
a = -a;
dist += a;
if (dist < bestd) {
bestd = dist;
bestpos = i;
}
int biasdist = dist - ((bias[i]) >> (intbiasshift - netbiasshift));
if (biasdist < bestbiasd) {
bestbiasd = biasdist;
bestbiaspos = i;
}
int betafreq = (freq[i] >> betashift);
freq[i] -= betafreq;
bias[i] += (betafreq << gammashift);
}
freq[bestpos] += beta;
bias[bestpos] -= betagamma;
return (bestbiaspos);
}
}