图像边缘检测(Canny 算法)的Java实现

来源:互联网 发布:软件防火墙 编辑:程序博客网 时间:2024/06/05 22:30

边缘检测算法的基本步骤 
(1)滤波。边缘检测主要基于导数计算,但受噪声影响。但滤波器在降低噪声的同时也导致边缘强度的损失。 (2)增强。增强算法将邻域中灰度有显著变化的点突出显示。一般通过计算梯度幅值完成。 
(3)检测。但在有些图象中梯度幅值较大的并不是边缘点。最简单的边缘检测是梯度幅值阈值判定。 
(4)定位。精确确定边缘的位置。

Canny边缘检测算法 
step1:用高斯滤波器平滑图象; 
step2:用一阶偏导的有限差分来计算梯度的幅值和方向; 
step3:对梯度幅值进行非极大值抑制; 
step4:用双阈值算法检测和连接边缘。

效果图如下:

代码如下:

package tools;
 
import java.awt.*;
import java.awt.image.*;
 
publicclass EdgeDetector extends Component {
 
        public EdgeDetector() {
               threshold1 = 50;
               threshold2 = 230;
               setThreshold(128);
               setWidGaussianKernel(15);
        }
 
        publicvoid process() throws EdgeDetectorException {
               if (threshold < 0 || threshold > 255)
                       thrownew EdgeDetectorException("The value of the threshold is out of its valid range.");
               if (widGaussianKernel < 3 || widGaussianKernel > 40)
                       thrownew EdgeDetectorException("The value of the widGaussianKernel is out of its valid range.");
               width = sourceImage.getWidth(this);
               height = sourceImage.getHeight(this);
               picsize = width * height;
               data = newint[picsize];
               magnitude = newint[picsize];
               orientation = newint[picsize];
               float f = 1.0F;
               canny_core(f, widGaussianKernel);
               thresholding_tracker(threshold1, threshold2);
               for (int i = 0; i < picsize; i++)
                       if (data[i] > threshold)
                               data[i] = 0xff000000;
                       else
                               data[i] = -1;
 
               edgeImage = pixels2image(data);
               data = null;
               magnitude = null;
               orientation = null;
        }
 
        privatevoid canny_core(float f, int i) {
               boolean flag = false;
               boolean flag1 = false;
               derivative_mag = newint[picsize];
               float af4[] = newfloat[i];
               float af5[] = newfloat[i];
               float af6[] = newfloat[i];
               data = image2pixels(sourceImage);
               int k4 = 0;
               do{
                       if (k4 >= i)
                               break;
                       float f1 = gaussian(k4, f);
                       if (f1 <= 0.005F && k4 >= 2)
                               break;
                       float f2 = gaussian((float) k4 - 0.5F, f);
                       float f3 = gaussian((float) k4 + 0.5F, f);
                       float f4 = gaussian(k4, f * 0.5F);
                       af4[k4] = (f1 + f2 + f3) / 3F / (6.283185F * f * f);
                       af5[k4] = f3 - f2;
                       af6[k4] = 1.6F * f4 - f1;
                       k4++;
               }while (true);
               int j = k4;
               float af[] = newfloat[picsize];
               float af1[] = newfloat[picsize];
               int j1 = width - (j - 1);
               int l = width * (j - 1);
               int i1 = width * (height - (j - 1));
               for (int l4 = j - 1; l4 < j1; l4++) {
                       for (int l5 = l; l5 < i1; l5 += width) {
                               int k1 = l4 + l5;
                               float f8 = (float) data[k1] * af4[0];
                               float f10 = f8;
                               int l6 = 1;
                               int k7 = k1 - width;
                               for (int i8 = k1 + width; l6 < j; i8 += width) {
                                      f8 += af4[l6] * (float) (data[k7] + data[i8]);
                                      f10 += af4[l6] * (float) (data[k1 - l6] + data[k1 + l6]);
                                      l6++;
                                      k7 -= width;
                               }
 
                               af[k1] = f8;
                               af1[k1] = f10;
                       }
 
               }
 
               float af2[] = newfloat[picsize];
               for (int i5 = j - 1; i5 < j1; i5++) {
                       for (int i6 = l; i6 < i1; i6 += width) {
                               float f9 = 0.0F;
                               int l1 = i5 + i6;
                               for (int i7 = 1; i7 < j; i7++)
                                      f9 += af5[i7] * (af[l1 - i7] - af[l1 + i7]);
 
                               af2[l1] = f9;
                       }
 
               }
 
               af = null;
               float af3[] = newfloat[picsize];
               for (int j5 = k4; j5 < width - k4; j5++) {
                       for (int j6 = l; j6 < i1; j6 += width) {
                               float f11 = 0.0F;
                               int i2 = j5 + j6;
                               int j7 = 1;
                               for (int l7 = width; j7 < j; l7 += width) {
                                      f11 += af5[j7] * (af1[i2 - l7] - af1[i2 + l7]);
                                      j7++;
                               }
 
                               af3[i2] = f11;
                       }
 
               }
 
               af1 = null;
               j1 = width - j;
               l = width * j;
               i1 = width * (height - j);
               for (int k5 = j; k5 < j1; k5++) {
                       for (int k6 = l; k6 < i1; k6 += width) {
                               int j2 = k5 + k6;
                               int k2 = j2 - width;
                               int l2 = j2 + width;
                               int i3 = j2 - 1;
                               int j3 = j2 + 1;
                               int k3 = k2 - 1;
                               int l3 = k2 + 1;
                               int i4 = l2 - 1;
                               int j4 = l2 + 1;
                               float f6 = af2[j2];
                               float f7 = af3[j2];
                               float f12 = hypotenuse(f6, f7);
                               int k = (int) ((double) f12 * 20D);
                               derivative_mag[j2] = k >= 256 ? 255 : k;
                               float f13 = hypotenuse(af2[k2], af3[k2]);
                               float f14 = hypotenuse(af2[l2], af3[l2]);
                               float f15 = hypotenuse(af2[i3], af3[i3]);
                               float f16 = hypotenuse(af2[j3], af3[j3]);
                               float f18 = hypotenuse(af2[l3], af3[l3]);
                               float f20 = hypotenuse(af2[j4], af3[j4]);
                               float f19 = hypotenuse(af2[i4], af3[i4]);
                               float f17 = hypotenuse(af2[k3], af3[k3]);
                               float f5;
                               if (f6 * f7 <= (float) 0
                                      ? Math.abs(f6) >= Math.abs(f7)
                                      ? (f5 = Math.abs(f6 * f12))
                                              >= Math.abs(f7 * f18 - (f6 + f7) * f16)
                                      && f5
                                              > Math.abs(f7 * f19 - (f6 + f7) * f15) : (
                                                     f5 = Math.abs(f7 * f12))
                                              >= Math.abs(f6 * f18 - (f7 + f6) * f13)
                                      && f5
                                              > Math.abs(f6 * f19 - (f7 + f6) * f14) : Math.abs(f6)
                                              >= Math.abs(f7)
                                              ? (f5 = Math.abs(f6 * f12))
                                                     >= Math.abs(f7 * f20 + (f6 - f7) * f16)
                                      && f5
                                              > Math.abs(f7 * f17 + (f6 - f7) * f15) : (
                                                     f5 = Math.abs(f7 * f12))
                                              >= Math.abs(f6 * f20 + (f7 - f6) * f14)
                                      && f5 > Math.abs(f6 * f17 + (f7 - f6) * f13)) {
                                      magnitude[j2] = derivative_mag[j2];
                                      orientation[j2] = (int) (Math.atan2(f7, f6) * (double) 40F);
                               }
                       }
 
               }
 
               derivative_mag = null;
               af2 = null;
               af3 = null;
        }
 
        privatefloat hypotenuse(float f, float f1) {
               if (f == 0.0F && f1 == 0.0F)
                       return 0.0F;
               else
                       return (float) Math.sqrt(f * f + f1 * f1);
        }
 
        privatefloat gaussian(float f, float f1) {
               return (float) Math.exp((-f * f) / ((float) 2 * f1 * f1));
        }
 
        privatevoid thresholding_tracker(int i, int j) {
               for (int k = 0; k < picsize; k++)
                       data[k] = 0;
 
               for (int l = 0; l < width; l++) {
                       for (int i1 = 0; i1 < height; i1++)
                               if (magnitude[l + width * i1] >= i)
                                      follow(l, i1, j);
 
               }
 
        }
 
        privateboolean follow(int i, int j, int k) {
               int j1 = i + 1;
               int k1 = i - 1;
               int l1 = j + 1;
               int i2 = j - 1;
               int j2 = i + j * width;
               if (l1 >= height)
                       l1 = height - 1;
               if (i2 < 0)
                       i2 = 0;
               if (j1 >= width)
                       j1 = width - 1;
               if (k1 < 0)
                       k1 = 0;
               if (data[j2] == 0) {
                       data[j2] = magnitude[j2];
                       boolean flag = false;
                       int l = k1;
                       do{
                               if (l > j1)
                                      break;
                               int i1 = i2;
                               do{
                                      if (i1 > l1)
                                              break;
                                      int k2 = l + i1 * width;
                                      if ((i1 != j || l != i)
                                              && magnitude[k2] >= k
                                              && follow(l, i1, k)) {
                                              flag = true;
                                              break;
                                      }
                                      i1++;
                               }while (true);
                               if (!flag)
                                      break;
                               l++;
                       }
                       while (true);
                       returntrue;
               }else{
                       returnfalse;
               }
        }
 
        private Image pixels2image(int ai[]) {
               MemoryImageSource memoryimagesource =
                       new MemoryImageSource(
                               width,
                               height,
                               ColorModel.getRGBdefault(),
                               ai,
                               0,
                               width);
               return Toolkit.getDefaultToolkit().createImage(memoryimagesource);
        }
 
        privateint[] image2pixels(Image image) {
               int ai[] = newint[picsize];
               PixelGrabber pixelgrabber =
                       new PixelGrabber(image, 0, 0, width, height, ai, 0, width);
               try{
                       pixelgrabber.grabPixels();
               }catch (InterruptedException interruptedexception) {
                       interruptedexception.printStackTrace();
               }
               boolean flag = false;
               int k1 = 0;
               do{
                       if (k1 >= 16)
                               break;
                       int i = (ai[k1] & 0xff0000) >> 16;
                       int k = (ai[k1] & 0xff00) >> 8;
                       int i1 = ai[k1] & 0xff;
                       if (i != k || k != i1) {
                               flag = true;
                               break;
                       }
                       k1++;
               }while (true);
               if (flag) {
                       for (int l1 = 0; l1 < picsize; l1++) {
                               int j = (ai[l1] & 0xff0000) >> 16;
                               int l = (ai[l1] & 0xff00) >> 8;
                               int j1 = ai[l1] & 0xff;
                               ai[l1] =
                                      (int) (0.29799999999999999D * (double) j
                                              + 0.58599999999999997D * (double) l
                                              + 0.113D * (double) j1);
                       }
 
               }else{
                       for (int i2 = 0; i2 < picsize; i2++)
                               ai[i2] = ai[i2] & 0xff;
 
               }
               return ai;
        }
 
        publicvoid setSourceImage(Image image) {
               sourceImage = image;
        }
 
        public Image getEdgeImage() {
               return edgeImage;
        }
 
        publicvoid setThreshold(int i) {
               threshold = i;
        }
 
        publicvoid setWidGaussianKernel(int i) {
               widGaussianKernel = i;
        }
 
        finalfloat ORIENT_SCALE = 40F;
        privateint height;
        privateint width;
        privateint picsize;
        privateint data[];
        privateint derivative_mag[];
        privateint magnitude[];
        privateint orientation[];
        private Image sourceImage;
        private Image edgeImage;
        privateint threshold1;
        privateint threshold2;
        privateint threshold;
        privateint widGaussianKernel;
}
 
//second file
 
package tools;
 
publicclass EdgeDetectorException extends Exception
{
  public EdgeDetectorException()
  {
    //do something?
  }
  public EdgeDetectorException(String s)
  {
    super(s);
  }
}

 

 

//使用示例

EdgeDetector edgeDetector=new EdgeDetector();
        edgeDetector.setSourceImage(sourceImage);
        edgeDetector.setThreshold(128);
        edgeDetector.setWidGaussianKernel(5);
        try{
            edgeDetector.process();
        }
        catch(EdgeDetectorException e) {
            System.out.println(e.getMessage());
        }
        Image edgeImage=edgeDetector.getEdgeImage();
        yourPanel.show(edgeImage);
转自:http://blog.csdn.net/Haohappy2004/article/details/476820
0 0
原创粉丝点击