我有这个.java数据文件。数据文件是imagej插件的一部分。 整个数据结构在这里： enter link description here
package mosaic.plugins;
import ij.IJ;
import ij.ImagePlus;
import ij.macro.Interpreter;
import ij.measure.ResultsTable;
import ij.process.ByteProcessor;
import java.awt.Color;
import java.awt.event.ActionEvent;
import java.awt.event.ActionListener;
import java.util.Map;
import java.util.Map.Entry;
import java.util.TreeMap;
import javax.swing.BorderFactory;
import javax.swing.JButton;
import javax.swing.JDialog;
import javax.swing.JPanel;
import javax.swing.JTextPane;
import javax.swing.WindowConstants;
import mosaic.plugins.utils.PlugIn8bitBase;
import net.imglib2.Cursor;
import net.imglib2.IterableInterval;
import net.imglib2.RandomAccess;
import net.imglib2.img.ImagePlusAdapter;
import net.imglib2.img.Img;
import net.imglib2.img.ImgFactory;
import net.imglib2.img.array.ArrayImgFactory;
import net.imglib2.img.display.imagej.ImageJFunctions;
import net.imglib2.type.NativeType;
import net.imglib2.type.numeric.NumericType;
import net.imglib2.type.numeric.RealType;
import net.imglib2.type.numeric.integer.UnsignedByteType;
import net.imglib2.type.numeric.real.FloatType;
import net.imglib2.view.IntervalView;
import net.imglib2.view.Views;
public class Naturalization extends PlugIn8bitBase
{
// Precision in finding your best T
private static final float EPS = 0.0001f;
// Prior parameter for first oder
// In this case is for all channels
// Fixed parameter
private static final float T1_pr = 0.3754f;
// Number of bins for the Laplacian Histogram
// In general is 4 * N_Grad
// max of laplacian value is 4 * 255
private static final int N_Lap = 2041;
// Offset shift in the histogram bins
// Has to be N_Lap / 2;
private static final int Lap_Offset = 1020;
// Number of bins for the Gradient
private static final int N_Grad = 512;
// Offset for the gradient histogram shift
private static final int Grad_Offset = 256;
// Prior parameter for second order (Parameters learned from trained data set)
// For different color R G B
// For one channel image use an average of them
private final float T2_pr[] = {0.2421f ,0.2550f, 0.2474f, 0.24816666f};
// Keeps values of PSNR for all images and channels in case of RGB. Maps: imageNumber -> map (channel, PSNR value)
private final Map<Integer, Map<Integer, Float>> iPsnrOutput = new TreeMap<Integer, Map<Integer, Float>>();
private synchronized void addPsnr(int aSlice, int aChannel, float aValue) {
Map<Integer, Float> map = iPsnrOutput.get(aSlice);
boolean isNewMap = false;
if (map == null) {
map = new TreeMap<Integer, Float>();
isNewMap = true;
}
map.put(aChannel, aValue);
if (isNewMap) {
iPsnrOutput.put(aSlice, map);
}
}
@Override
protected void processImg(ByteProcessor aOutputImg, ByteProcessor aOrigImg, int aChannelNumber) {
// perform naturalization
final ImagePlus naturalizedImg = naturalize8bitImage(aOrigImg, aChannelNumber);
// set processed pixels to output image
aOutputImg.setPixels(naturalizedImg.getProcessor().getPixels());
}
@Override
protected void postprocessBeforeShow() {
// Create result table with all stored PSNRs.
final ResultsTable rs = new ResultsTable();
for (final Entry<Integer, Map<Integer, Float>> e : iPsnrOutput.entrySet()) {
rs.incrementCounter();
for (final Entry<Integer, Float> m : e.getValue().entrySet()) {
switch(m.getKey()) {
case CHANNEL_R: rs.addValue("Naturalization R", m.getValue()); rs.addValue("Estimated R PSNR", calculate_PSNR(m.getValue())); break;
case CHANNEL_G: rs.addValue("Naturalization G", m.getValue()); rs.addValue("Estimated G PSNR", calculate_PSNR(m.getValue())); break;
case CHANNEL_B: rs.addValue("Naturalization B", m.getValue()); rs.addValue("Estimated B PSNR", calculate_PSNR(m.getValue())); break;
case CHANNEL_8G: rs.addValue("Naturalization", m.getValue()); rs.addValue("Estimated PSNR", calculate_PSNR(m.getValue())); break;
default: break;
}
}
}
if (!Interpreter.isBatchMode()) {
rs.show("Naturalization and PSNR");
showMessage();
}
}
private ImagePlus naturalize8bitImage(ByteProcessor imp, int aChannelNumber) {
Img<UnsignedByteType> TChannel = ImagePlusAdapter.wrap(new ImagePlus("", imp));
final float T2_prior = T2_pr[(aChannelNumber <= CHANNEL_B) ? 2-aChannelNumber : CHANNEL_8G];
final float[] result = {0.0f}; // ugly but one of ways to get result back via parameters;
// Perform naturalization and store PSNR result. Finally return image in ImageJ format.
TChannel = performNaturalization(TChannel, T2_prior, result);
addPsnr(imp.getSliceNumber(), aChannelNumber, result[0]);
return ImageJFunctions.wrap(TChannel,"temporaryName");
}
/**
* Naturalize the image
* @param Img original image
* @param Theta parameter
* @param Class<T> Original image
* @param Class<S> Calculation Type
* @param T2_prior Prior to use
* @param result One element array to store nautralization factor
*/
private <T extends NumericType<T> & NativeType<T> & RealType<T>, S extends RealType<S>> Img<T> doNaturalization(Img<T> image_orig, S Theta,Class<T> cls_t, float T2_prior, float[] result) throws InstantiationException, IllegalAccessException
{
if (image_orig == null) {return null;}
// Check that the image data set is 8 bit
// Otherwise return an error or hint to scale down
final T image_check = cls_t.newInstance();
final Object obj = image_check;
if (!(obj instanceof UnsignedByteType)) {
IJ.error("Error it work only with 8-bit type");
return null;
}
final float Nf = findNaturalizationFactor(image_orig, Theta, T2_prior);
result[0] = Nf;
final Img<T> image_result = naturalizeImage(image_orig, Nf, cls_t);
return image_result;
}
private <S extends RealType<S>, T extends NumericType<T> & NativeType<T> & RealType<T>>
Img<T> naturalizeImage(Img<T> image_orig, float Nf, Class<T> cls_t)
throws InstantiationException, IllegalAccessException
{
// Mean of the original image
// S mean_original = cls_s.newInstance();
// Mean<T,S> m = new Mean<T,S>();
// m.compute(image_orig.cursor(), mean_original);
// TODO: quick fix for deprecated code above. Is new 'mean' utility introduced in imglib2?
float mean_original = 0.0f;
final Cursor<T> c2 = image_orig.cursor();
float count = 0.0f;
while (c2.hasNext()) {
c2.next();
mean_original += c2.get().getRealFloat();
count += 1.0f;
}
mean_original /= count;
// Create result image
final long[] origImgDimensions = new long[2];
image_orig.dimensions(origImgDimensions);
final Img<T> image_result = image_orig.factory().create(origImgDimensions, cls_t.newInstance());
// for each pixel naturalize
final Cursor<T> cur_orig = image_orig.cursor();
final Cursor<T> cur_ir = image_result.cursor();
while (cur_orig.hasNext()) {
cur_orig.next();
cur_ir.next();
final float tmp = cur_orig.get().getRealFloat();
// Naturalize
float Nat = (int) ((tmp - mean_original)*Nf + mean_original + 0.5);
if (Nat < 0)
{Nat = 0;}
else if (Nat > 255)
{Nat = 255;}
cur_ir.get().setReal(Nat);
}
return image_result;
}
private <S extends RealType<S>, T extends NumericType<T> & NativeType<T> & RealType<T>> float findNaturalizationFactor(Img<T> image_orig, S Theta, float T2prior) {
final ImgFactory<FloatType> imgFactoryF = new ArrayImgFactory<FloatType>();
// Create one dimensional image (Histogram)
final Img<FloatType> LapCDF = imgFactoryF.create(new long[] {N_Lap}, new FloatType());
// Two dimensional image for Gradient
final Img<FloatType> GradCDF = imgFactoryF.create(new long[] {N_Grad, 2}, new FloatType());
// GradientCDF = Integral of the histogram of the of the Gradient field
// LaplacianCDF = Integral of the Histogram of the Laplacian field
final Img<FloatType> GradD = create2DGradientField();
calculateLaplaceFieldAndGradient(image_orig, LapCDF, GradD);
convertGrad2dToCDF(GradD);
calculateGradCDF(GradCDF, GradD);
calculateLapCDF(LapCDF);
// For each channel find the best T1
// EPS=precision
// for X component
float T_tmp = (float)FindT(Views.iterable(Views.hyperSlice(GradCDF, GradCDF.numDimensions()-1 , 0)), N_Grad, Grad_Offset, EPS);
// for Y component
T_tmp += FindT(Views.iterable(Views.hyperSlice(GradCDF, GradCDF.numDimensions()-1 , 1)), N_Grad, Grad_Offset, EPS);
// Average them and divide by the prior parameter
final float T1 = T_tmp/(2*T1_pr);
// Find the best parameter and divide by the T2 prior
final float T2 = (float)FindT(LapCDF, N_Lap, Lap_Offset, EPS)/T2prior;
// Calculate naturalization factor!
final float Nf = (float) ((1.0-Theta.getRealDouble())*T1 + Theta.getRealDouble()*T2);
return Nf;
}
/**
* Calculate the peak SNR from the Naturalization factor
*
* @param Nf naturalization factor
* @return the PSNR
*/
String calculate_PSNR(double x)
{
if (x >= 0 && x <= 0.934)
{
return String.format("%.2f", new Float(23.65 * Math.exp(0.6 * x) - 20.0 * Math.exp(-7.508 * x)));
}
else if (x > 0.934 && x < 1.07)
{
return new String("> 40");
}
else if (x >= 1.07 && x < 1.9)
{
return String.format("%.2f", new Float(-11.566 * x + 52.776));
}
else
{
return String.format("%.2f",new Float(13.06*x*x*x*x - 121.4 * x*x*x + 408.5 * x*x -595.5*x + 349));
}
}
private Img<UnsignedByteType> performNaturalization(Img<UnsignedByteType> channel, float T2_prior, float[] result) {
// Parameters balance between first order and second order
final FloatType Theta = new FloatType(0.5f);
try {
channel = doNaturalization(channel, Theta, UnsignedByteType.class, T2_prior, result);
} catch (final InstantiationException e) {
e.printStackTrace();
} catch (final IllegalAccessException e) {
e.printStackTrace();
}
return channel;
}
// Original data
// N = nuber of bins
// offset of the histogram
// T current
private double FindT_Evalue(float[] p_d, int N, int offset, float T)
{
double error = 0;
for (int i=-offset; i<N-offset; ++i) {
final double tmp = Math.atan(T*(i)) - p_d[i+offset];
error += (tmp*tmp);
}
return error;
}
// Find the T
// data CDF Histogram
// N number of bins
// Offset of the histogram
// eps precision
private double FindT(IterableInterval<FloatType> data, int N, int OffSet, float eps)
{
//find the best parameter between data and model atan(Tx)/pi+0.5
// Search between 0 and 1.0
float left = 0;
float right = 1.0f;
float m1 = 0.0f;
float m2 = 0.0f;
// Crate p_t to save computation (shift and rescale the original CDF)
final float p_t[] = new float[N];
// Copy the data
final Cursor<FloatType> cur_data = data.cursor();
for (int i = 0; i < N; ++i)
{
cur_data.next();
p_t[i] = (float) ((cur_data.get().getRealFloat() - 0.5)*Math.PI);
}
// While the precision is bigger than eps
while (right-left>=eps)
{
// move left and right of 1/3 (m1 and m2)
m1=left+(right-left)/3;
m2=right-(right-left)/3;
// Evaluate on m1 and m2, ane move the extreme point
if (FindT_Evalue(p_t, N, OffSet, m1) <=FindT_Evalue(p_t, N, OffSet, m2)) {
right=m2;
}
else {
left=m1;
}
}
// return the average
return (m1+m2)/2;
}
private Img<FloatType> create2DGradientField() {
final long dims[] = new long[2];
dims[0] = N_Grad;
dims[1] = N_Grad;
final Img<FloatType> GradD = new ArrayImgFactory<FloatType>().create(dims, new FloatType());
return GradD;
}
private void calculateLapCDF(Img<FloatType> LapCDF) {
final RandomAccess<FloatType> Lap_hist2 = LapCDF.randomAccess();
//convert Lap to CDF
for (int i = 1; i < N_Lap; ++i)
{
Lap_hist2.setPosition(i-1,0);
final float prec = Lap_hist2.get().getRealFloat();
Lap_hist2.move(1,0);
Lap_hist2.get().set(Lap_hist2.get().getRealFloat() + prec);
}
}
private void calculateGradCDF(Img<FloatType> GradCDF, Img<FloatType> GradD) {
final RandomAccess<FloatType> Grad_dist = GradD.randomAccess();
// Gradient on x pointer
final IntervalView<FloatType> Gradx = Views.hyperSlice(GradCDF, GradCDF.numDimensions()-1 , 0);
// Gradient on y pointer
final IntervalView<FloatType> Grady = Views.hyperSlice(GradCDF, GradCDF.numDimensions()-1 , 1);
integrateOverRowAndCol(Grad_dist, Gradx, Grady);
scaleGradiens(Gradx, Grady);
}
private void scaleGradiens(IntervalView<FloatType> Gradx, IntervalView<FloatType> Grady) {
final RandomAccess<FloatType> Gradx_r2 = Gradx.randomAccess();
final RandomAccess<FloatType> Grady_r2 = Grady.randomAccess();
//scale, divide the number of integrated bins
for (int i = 0; i < N_Grad; ++i)
{
Gradx_r2.setPosition(i,0);
Grady_r2.setPosition(i,0);
Gradx_r2.get().set((float) (Gradx_r2.get().getRealFloat() / 255.0));
Grady_r2.get().set((float) (Grady_r2.get().getRealFloat() / 255.0));
}
}
private void integrateOverRowAndCol(RandomAccess<FloatType> Grad_dist, IntervalView<FloatType> Gradx, IntervalView<FloatType> Grady) {
final int[] loc = new int[2];
// pGrad2D has 2D CDF
final RandomAccess<FloatType> Gradx_r = Gradx.randomAccess();
// Integrate over the row
for (int i = 0; i < N_Grad; ++i)
{
loc[1] = i;
Gradx_r.setPosition(i,0);
// get the row
for (int j = 0; j < N_Grad; ++j)
{
loc[0] = j;
// Set the position
Grad_dist.setPosition(loc);
// integrate over the row to get 1D vector
Gradx_r.get().set(Gradx_r.get().getRealFloat() + Grad_dist.get().getRealFloat());
}
}
final RandomAccess<FloatType> Grady_r = Grady.randomAccess();
// Integrate over the column
for (int i = 0; i < N_Grad; ++i)
{
loc[1] = i;
Grady_r.setPosition(0,0);
for (int j = 0; j < N_Grad; ++j)
{
loc[0] = j;
Grad_dist.setPosition(loc);
Grady_r.get().set(Grady_r.get().getRealFloat() + Grad_dist.get().getRealFloat());
Grady_r.move(1,0);
}
}
}
private <T extends RealType<T>> void calculateLaplaceFieldAndGradient(Img<T> image, Img<FloatType> LapCDF, Img<FloatType> GradD) {
final RandomAccess<FloatType> Grad_dist = GradD.randomAccess();
final long[] origImgDimensions = new long[2];
image.dimensions(origImgDimensions);
final Img<FloatType> laplaceField = new ArrayImgFactory<FloatType>().create(origImgDimensions, new FloatType());
// Cursor localization
final int[] indexD = new int[2];
final int[] loc_p = new int[2];
final RandomAccess<T> img_cur = image.randomAccess();
final RandomAccess<FloatType> Lap_f = laplaceField.randomAccess();
final RandomAccess<FloatType> Lap_hist = LapCDF.randomAccess();
// Normalization 1/(Number of pixel of the original image)
long n_pixel = 1;
for (int i = 0 ; i < laplaceField.numDimensions() ; i++)
{n_pixel *= laplaceField.dimension(i)-2;}
// unit to sum
final double f = 1.0/(n_pixel);
// Inside the image for Y
final Cursor<FloatType> cur = laplaceField.cursor();
// For each point of the Laplacian field
while (cur.hasNext())
{
cur.next();
// Localize cursors
cur.localize(loc_p);
// Exclude the border
boolean border = false;
for (int i = 0 ; i < image.numDimensions() ; i++)
{
if (loc_p[i] == 0)
{border = true;}
else if (loc_p[i] == image.dimension(i)-1)
{border = true;}
}
if (border == true) {
continue;
}
// get the stencil value;
img_cur.setPosition(loc_p);
float L = -4*img_cur.get().getRealFloat();
// Laplacian
for (int i = 0 ; i < 2 ; i++)
{
img_cur.move(1, i);
final float G_p = img_cur.get().getRealFloat();
img_cur.move(-1,i);
final float G_m = img_cur.get().getRealFloat();
img_cur.move(-1, i);
final float L_m = img_cur.get().getRealFloat();
img_cur.setPosition(loc_p);
L += G_p + L_m;
// Calculate the gradient + convert into bin
indexD[1-i] = (int) (Grad_Offset + G_p - G_m);
}
Lap_f.setPosition(loc_p);
// Set the Laplacian field
Lap_f.get().setReal(L);
// Histogram bin conversion
L += Lap_Offset;
Lap_hist.setPosition((int)(L),0);
Lap_hist.get().setReal(Lap_hist.get().getRealFloat() + f);
Grad_dist.setPosition(indexD);
Grad_dist.get().setReal(Grad_dist.get().getRealFloat() + f);
}
}
private void convertGrad2dToCDF(Img<FloatType> GradD) {
final RandomAccess<FloatType> Grad_dist = GradD.randomAccess();
final int[] loc = new int[GradD.numDimensions()];
// for each row
for (int j = 0; j < GradD.dimension(1); ++j)
{
loc[1] = j;
for (int i = 1; i < GradD.dimension(0) ; ++i)
{
loc[0] = i-1;
Grad_dist.setPosition(loc);
// Precedent float
final float prec = Grad_dist.get().getRealFloat();
// Move to the actual position
Grad_dist.move(1, 0);
// integration up to the current position
Grad_dist.get().set(Grad_dist.get().getRealFloat() + prec);
}
}
//col integration
for (int j = 1; j < GradD.dimension(1); ++j)
{
// Move to the actual position
loc[1] = j-1;
for (int i = 0; i < GradD.dimension(0); ++i)
{
loc[0] = i;
Grad_dist.setPosition(loc);
// Precedent float
final float prec = Grad_dist.get().getRealFloat();
// Move to the actual position
Grad_dist.move(1, 1);
Grad_dist.get().set(Grad_dist.get().getRealFloat() + prec);
}
}
}
/**
* Show information about authors and paper.
*/
private void showMessage()
{
// Create main window with panel to store gui components
final JDialog win = new JDialog((JDialog)null, "Naturalization", true);
final JPanel msg = new JPanel();
msg.setBorder(BorderFactory.createEmptyBorder(10, 10, 10, 10));
// Create message not editable but still focusable for copying
final JTextPane text = new JTextPane();
text.setContentType("text/html");
text.setText("<html>Y. Gong and I. F. Sbalzarini. Image enhancement by gradient distribution specification. In Proc. ACCV, <br>"
+ "12th Asian Conference on Computer Vision, Workshop on Emerging Topics in Image Enhancement and Restoration,<br>"
+ "pages w7–p3, Singapore, November 2014.<br><br>"
+ "Y. Gong and I. F. Sbalzarini, Gradient Distributions Priors for Biomedical Image Processing, 2014<br><a href=\"http://arxiv.org/abs/1408.3300\">http://arxiv.org/abs/1408.3300</a><br><br>"
+ "Y. Gong and I. F. Sbalzarini. A Natural-Scene Gradient Distribution Prior and its Application in Light-Microscopy Image Processing.<br>"
+ "IEEE Journal of Selected Topics in Signal Processing, Vol.10, No.1, February 2016, pages 99-114<br>"
+ "ISSN: 1932-4553, DOI: 10.1109/JSTSP.2015.2506122<br><br>"
+ "</html>");
text.setBorder(BorderFactory.createLineBorder(Color.BLACK, 2));
text.setEditable(false);
msg.add(text);
// Add button "Close" for closing window easily
final JButton button = new JButton("Close");
button.addActionListener(new ActionListener() {
@Override
public void actionPerformed(ActionEvent e) {
win.dispose();
}
});
msg.add(button);
// Finally show window with message
win.add(msg);
win.pack();
win.setDefaultCloseOperation(WindowConstants.DISPOSE_ON_CLOSE);
win.setVisible(true);
}
@Override
protected boolean showDialog() {
return true;
}
@Override
protected boolean setup(String aArgs) {
setFilePrefix("naturalized_");
return true;
}
}
我希望它再次编译并获得此插件的.class文件或整个.jar文件。
获得.class数据需要哪些结构和数据？ 什么是导入文件，在哪里可以获取ij，java，javax和net文件？
我是Java的新手，只知道编译后的命令是javac。
答案 0 :(得分：0)
在linux上有一个命令是javac
just：javac HelloWorld.java
在Windows上可能是同一件事，但我不确定（如果没有其他方法，请安装虚拟linux盒）
如果出现问题，谷歌错误
答案 1 :(得分：0)
如果要从命令行编译Java程序，则应使用javac命令并调用它，只需编写Java，然后编写程序名称即可。 编译文件后，您将拥有要查找的.class文件。