在《数字图像处理》一书中介绍了用于降低图像噪声的均值滤波器，分别有算数均值滤波器、几何均值滤波器、谐波均值滤波器、逆谐波均值滤波器。除了降噪，均值滤波器也可以模糊图像，滤波器大小为3、5、7...2n+1，滤波器越大计算量越大，产生的图像越模糊。这里采用纯java对几种算法进行实现，代码如下：
实现类：
import java.awt.image.BufferedImage;
/**
* 几种均值滤波算法实现 1、算术均值滤波 2、几何均值滤波 3、谐波均值滤波 4、逆谐波均值滤波
*
* @author admin
*/
public class AverageFilter {
// singleton
private static AverageFilter averageFilter = new AverageFilter();
public static AverageFilter getInstance() {
return averageFilter;
}
private AverageFilter() {
}
/**
* 算术均值滤波 滤波器大小：param取值3、5、7、9...(2n+1 ) 产生一幅模糊图像
*
* @param image
*/
public void arithmeticAverageFilter(BufferedImage image, int param) {
// 创建一个临时图像，为了保证原图边缘参与计算，临时图像比原图大param-1
BufferedImage tempImage = new BufferedImage(image.getWidth() + param - 1, image.getHeight() + param - 1,
image.getType());
// 对图像进行填充，边缘像素采用最近像素填充
nearFillEdge(tempImage, image, param);
// 进行卷积运算
for (int i = (param - 1) / 2; i < tempImage.getWidth() - (param - 1) / 2; i++) {
for (int j = (param - 1) / 2; j < tempImage.getHeight() - (param - 1) / 2; j++) {
int r = 0, g = 0, b = 0;
// 计算滤波器内所有像素，R、G、B各个分量总和
for (int x = -(param - 1) / 2; x <= (param - 1) / 2; x++) {
for (int y = -(param - 1) / 2; y <= (param - 1) / 2; y++) {
int tempRGB = tempImage.getRGB(i + x, j + y);
int tempR = (tempRGB >> 16) & 0xff;
int tempG = (tempRGB >> 8) & 0xff;
int tempB = tempRGB & 0xff;
r += tempR;
g += tempG;
b += tempB;
}
}
// 采用总和除以滤波器内总像素数量得到均值
r = (int) (r / Math.pow(param, 2));
g = (int) (g / Math.pow(param, 2));
b = (int) (b / Math.pow(param, 2));
int rgb = (255 & 0xff) << 24 | (clamp(r) & 0xff) << 16 | (clamp(g) & 0xff) << 8 | (clamp(b) & 0xff);
image.setRGB(i - (param - 1) / 2, j - (param - 1) / 2, rgb);
}
}
}
/**
* 几何均值滤波
*
* @param image
* @param param
*/
public void geometryAverageFilter(BufferedImage image, int param) {
// 创建临时图像
BufferedImage tempImage = new BufferedImage(image.getWidth() + param - 1, image.getHeight() + param - 1,
image.getType());
// 填充边缘
nearFillEdge(tempImage, image, param);
// 进行卷积运算
for (int i = (param - 1) / 2; i < tempImage.getWidth() - (param - 1) / 2; i++) {
for (int j = (param - 1) / 2; j < tempImage.getHeight() - (param - 1) / 2; j++) {
double r = 1.0, g = 1.0, b = 1.0;
for (int x = -(param - 1) / 2; x <= (param - 1) / 2; x++) {
for (int y = -(param - 1) / 2; y <= (param - 1) / 2; y++) {
int tempRGB = tempImage.getRGB(i + x, j + y);
double tempR = (tempRGB >> 16) & 0xff;
double tempG = (tempRGB >> 8) & 0xff;
double tempB = tempRGB & 0xff;
r *= Math.pow(tempR + 1, 1.0 / (param * param));
g *= Math.pow(tempG + 1, 1.0 / (param * param));
b *= Math.pow(tempB + 1, 1.0 / (param * param));
}
}
int rgb = (255 & 0xff) << 24 | (clamp((int) r) & 0xff) << 16 | (clamp((int) g) & 0xff) << 8
| (clamp((int) b) & 0xff);
image.setRGB(i - (param - 1) / 2, j - (param - 1) / 2, rgb);
}
}
}
/**
* 谐波均值滤波
*
* @param image
* @param param
*/
public void harmonicFilter(BufferedImage image, int param) {
// 创建temp图像
BufferedImage tempImage = new BufferedImage(image.getWidth() + param - 1, image.getHeight() + param - 1,
image.getType());
// 填充边缘
nearFillEdge(tempImage, image, param);
// 进行卷积运算
for (int i = (param - 1) / 2; i < tempImage.getWidth() - (param - 1) / 2; i++) {
for (int j = (param - 1) / 2; j < tempImage.getHeight() - (param - 1) / 2; j++) {
double r = 0, g = 0, b = 0;
for (int x = -(param - 1) / 2; x <= (param - 1) / 2; x++) {
for (int y = -(param - 1) / 2; y <= (param - 1) / 2; y++) {
int tempRGB = tempImage.getRGB(i + x, j + y);
double tempR = (tempRGB >> 16) & 0xff;
double tempG = (tempRGB >> 8) & 0xff;
double tempB = tempRGB & 0xff;
r += 1 / tempR;
g += 1 / tempG;
b += 1 / tempB;
}
}
r = param * param / r;
g = param * param / g;
b = param * param / b;
int rgb = (255 & 0xff) << 24 | (clamp((int) r) & 0xff) << 16 | (clamp((int) g) & 0xff) << 8
| (clamp((int) b) & 0xff);
image.setRGB(i - (param - 1) / 2, j - (param - 1) / 2, rgb);
}
}
}
/**
* 逆谐波均值滤波
*
* @param image
* @param param
* @param Q
* 当Q=0为算术均值滤波，Q=-1为谐波均值滤波；Q为正，消除胡椒噪声，Q为负，消除盐粒噪声
*/
public void reverseHarmonicFilter(BufferedImage image, int param, int Q) {
// 创建temp图像
BufferedImage tempImage = new BufferedImage(image.getWidth() + param - 1, image.getHeight() + param - 1,
image.getType());
// 填充边缘
nearFillEdge(tempImage, image, param);
// 进行卷积运算
for (int i = (param - 1) / 2; i < tempImage.getWidth() - (param - 1) / 2; i++) {
for (int j = (param - 1) / 2; j < tempImage.getHeight() - (param - 1) / 2; j++) {
double r = 0, g = 0, b = 0, r1 = 0, g1 = 0, b1 = 0;
for (int x = -(param - 1) / 2; x <= (param - 1) / 2; x++) {
for (int y = -(param - 1) / 2; y <= (param - 1) / 2; y++) {
int tempRGB = tempImage.getRGB(i + x, j + y);
double tempR = (tempRGB >> 16) & 0xff;
double tempG = (tempRGB >> 8) & 0xff;
double tempB = tempRGB & 0xff;
r += Math.pow(tempR, Q + 1);
g += Math.pow(tempG, Q + 1);
b += Math.pow(tempB, Q + 1);
r1 += Math.pow(tempR, Q);
g1 += Math.pow(tempG, Q);
b1 += Math.pow(tempB, Q);
}
}
r = r / r1;
g = g / g1;
b = b / b1;
int rgb = (255 & 0xff) << 24 | (clamp((int) r) & 0xff) << 16 | (clamp((int) g) & 0xff) << 8
| (clamp((int) b) & 0xff);
image.setRGB(i - (param - 1) / 2, j - (param - 1) / 2, rgb);
}
}
}
// 判断r,g,b值，大于256返回256，小于0则返回0,0到256之间则直接返回原始值
private int clamp(int rgb) {
if (rgb > 255)
return 255;
if (rgb < 0)
return 0;
return rgb;
}
// 填充图像边缘空白像素,使用最近像素填充
private void nearFillEdge(BufferedImage tempImage, BufferedImage image, int param) {
for (int i = 0; i < tempImage.getWidth(); i++) {
for (int j = 0; j < tempImage.getHeight(); j++) {
// 临时图像位置没超过原图第一个位置,左下角
if (i <= (param - 1) / 2 & j <= (param - 1) / 2) {
int rgb = image.getRGB(0, 0);
tempImage.setRGB(i, j, rgb);
}
// 临时图像位置超过横坐标最大，小于纵坐标最小，右下角
if (i >= tempImage.getWidth() - (param - 1) / 2 - 1 & j <= (param - 1) / 2) {
int rgb = image.getRGB(image.getWidth() - 1, 0);
tempImage.setRGB(i, j, rgb);
}
// 临时图像位置超过纵坐标最大，小于横坐标最小，左上角
if (j >= tempImage.getHeight() - (param - 1) / 2 - 1 & i <= (param - 1) / 2) {
int rgb = image.getRGB(0, image.getHeight() - 1);
tempImage.setRGB(i, j, rgb);
}
// 临时图像位置横纵坐标都超过原图最大位置，右上角
if (i >= tempImage.getWidth() - (param - 1) / 2 - 1
& j >= tempImage.getHeight() - (param - 1) / 2 - 1) {
int rgb = image.getRGB(image.getWidth() - 1, image.getHeight() - 1);
tempImage.setRGB(i, j, rgb);
}
// 临时图像位置横坐标大于最小，小于最大，纵坐标小于最小，正下方
if (i > (param - 1) / 2 & i < tempImage.getWidth() - (param - 1) / 2 - 1 & j <= (param - 1) / 2) {
int rgb = image.getRGB(i - (param - 1) / 2, 0);
tempImage.setRGB(i, j, rgb);
}
// 临时图像位置横坐标小于最小，纵坐标大于最小，小于最大，左边
if (j > (param - 1) / 2 & j < tempImage.getHeight() - (param - 1) / 2 - 1 & i <= (param - 1) / 2) {
int rgb = image.getRGB(0, j - (param - 1) / 2);
tempImage.setRGB(i, j, rgb);
}
// 右边
if (j > (param - 1) / 2 & j < tempImage.getHeight() - (param - 1) / 2 - 1
& i >= tempImage.getWidth() - (param - 1) / 2 - 1) {
int rgb = image.getRGB(image.getWidth() - 1, j - (param - 1) / 2);
tempImage.setRGB(i, j, rgb);
}
// 上方
if (i > (param - 1) / 2 & i < tempImage.getWidth() - (param - 1) / 2 - 1
& j >= tempImage.getHeight() - (param - 1) / 2 - 1) {
int rgb = image.getRGB(i - (param - 1) / 2, image.getHeight() - 1);
tempImage.setRGB(i, j, rgb);
}
// 中间
if (i > (param - 1) / 2 & i < tempImage.getWidth() - (param - 1) / 2 - 1 & j > (param - 1) / 2
& j < tempImage.getHeight() - (param - 1) / 2 - 1) {
int rgb = image.getRGB(i - (param - 1) / 2, j - (param - 1) / 2);
tempImage.setRGB(i, j, rgb);
}
}
}
}
}
测试类：
import java.awt.image.BufferedImage;
import java.io.File;
import javax.imageio.ImageIO;
public class test {
public static void main(String[] args) throws Exception{
File input = new File("C:/桌面/AverageFilter/1.jpg");
File output = new File("C:/桌面/AverageFilter/2.jpg");
BufferedImage image = ImageIO.read(input);
AverageFilter.getInstance().reverseHarmonicFilter(image, 3, 2);
ImageIO.write(image, "jpg", output);
}
}
507
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
