package train;

import java.awt.Color;

import java.awt.image.BufferedImage;

import java.io.File;

import java.io.FilenameFilter;

import java.io.IOException;

import javax.imageio.ImageIO;

public class Preprocess {

public Preprocess(){

}

private void run(){

File dir = new File("download");

//只列出jpg

File[] files = dir.listFiles(new FilenameFilter() {

public boolean isJpg(String file){

if (file.toLowerCase().endsWith(".jpg")){

return true;

}else{

return false;

}

}

@Override

public boolean accept(File dir, String name) {

// TODO Auto-generated method stub

return isJpg(name);

}

});

for (File file : files) {

try {

BufferedImage img = ImageIO.read(file);

BufferedImage binaryImg = getBinaryImage(img);

ImageIO.write(binaryImg, "JPG", new File("1_gray/" + file.getName()));

} catch (IOException e) {

e.printStackTrace();

}

}

}

/**

* 二值化

* @param sourceImage

* @return 二值化之后的图像

*/

public BufferedImage getBinaryImage(BufferedImage sourceImage){

double Wr = 0.299;

double Wg = 0.587;

double Wb = 0.114;

int width = sourceImage.getWidth();

int height = sourceImage.getHeight();

int[][] gray = new int[width][height];

//灰度化

for (int x = 0; x < width; x++) {

for (int y = 0; y < height; y++) {

Color color = new Color(sourceImage.getRGB(x, y));

int rgb = (int) ((color.getRed()*Wr + color.getGreen()*Wg + color.getBlue()*Wb) / 3);

gray[x][y] = rgb;

}

}

BufferedImage binaryBufferedImage = new BufferedImage(width, height, BufferedImage.TYPE_BYTE_BINARY);

//二值化

int threshold = getOstu(gray, width, height);

for (int x = 0; x < width; x++) {

for (int y = 0; y < height; y++) {

if (gray[x][y] > threshold) {

int max = new Color(255, 255, 255).getRGB();

gray[x][y] = max;

}else{

int min = new Color(0, 0, 0).getRGB();

gray[x][y] = min;

}

binaryBufferedImage.setRGB(x, y, gray[x][y]);

}

}

return binaryBufferedImage;

}

/**

* 获得二值化图像

* 最大类间方差法

* @param gray

* @param width

* @param height

*/

private int getOstu(int[][] gray, int width, int height){

int grayLevel = 256;

int[] pixelNum = new int[grayLevel];

//计算所有色阶的直方图

for (int x = 0; x < width; x++) {

for (int y = 0; y < height; y++) {

int color = gray[x][y];

pixelNum[color] ++;

}

}

double sum = 0;

int total = 0;

for (int i = 0; i < grayLevel; i++) {

sum += i*pixelNum[i]; //x*f(x)质量矩，也就是每个灰度的值乘以其点数（归一化后为概率），sum为其总和

total += pixelNum[i]; //n为图象总的点数，归一化后就是累积概率

}

double sumB = 0;//前景色质量矩总和

int threshold = 0;

double wF = 0;//前景色权重

double wB = 0;//背景色权重

double maxFreq = -1.0;//最大类间方差

for (int i = 0; i < grayLevel; i++) {

wB += pixelNum[i]; //wB为在当前阈值背景图象的点数

if (wB == 0) { //没有分出前景后景

continue;

}

wF = total - wB; //wB为在当前阈值前景图象的点数

if (wF == 0) {//全是前景图像，则可以直接break

break;

}

sumB += (double)(i*pixelNum[i]);

double meanB = sumB / wB;

double meanF = (sum - sumB) / wF;

//freq为类间方差

double freq = (double)(wF)*(double)(wB)*(meanB - meanF)*(meanB - meanF);

if (freq > maxFreq) {

maxFreq = freq;

threshold = i;

}

}

return threshold;

}

public static void main(String[] args){

System.out.println("---begin---");

long start = System.currentTimeMillis();

Preprocess model = new Preprocess();

model.run();

long end = System.currentTimeMillis();

System.out.println("耗时：" + (end - start));

System.out.println("---end----");

}

}