public void run(ImageProcessor image) { /** * ********************************************************************************* Initial * phase ********************************************************************************* */ int width = image.getWidth(); int heigh = image.getHeight(); ByteProcessor bp = Service.getByteProcessor(image); /** * ********************************************************************************* Convolve * with LoG kernel * ********************************************************************************* */ Convolver convolver = new Convolver(); convolver.setNormalize(false); convolver.convolve(bp, log, (int) Math.sqrt(log.length), (int) Math.sqrt(log.length)); ImagePlus outImg = new ImagePlus("Later Log kernel", bp); outImg.show(); /** * ********************************************************************************* threshold * ********************************************************************************* */ ByteProcessor bpThreshold = Service.getByteProcessor(bp); int lut[] = new int[256]; int i = 0; for (; i < threshold; i++) lut[i] = 0; for (int j = i; j < lut.length; j++) lut[j] = 255; bpThreshold.applyTable(lut); ImagePlus outImg1 = new ImagePlus("Later threshold phase", bpThreshold); outImg1.show(); /** * ********************************************************************************* Find Zero * crossing ********************************************************************************* */ ByteProcessor out = new ByteProcessor(width, heigh); for (i = 0; i < width; i++) for (int j = 0; j < heigh; j++) out.set(i, j, 255); for (int x = 0; x < width - 1; x++) { for (int y = 0; y < heigh - 1; y++) { if (bpThreshold.get(x, y) != bpThreshold.get(x, y + 1)) out.set(x, y, 0); if (bpThreshold.get(x, y) != bpThreshold.get(x + 1, y)) out.set(x, y, 0); if (bpThreshold.get(x, y) != bpThreshold.get(x + 1, y + 1)) out.set(x, y, 0); } } ImagePlus outImg2 = new ImagePlus("Edge Find", out); outImg2.show(); }
public void run(ImageProcessor ip) { int w = ip.getWidth(); // Get width of image int h = ip.getHeight(); // Get height of image /** * ------------------------------------------------------------------ BEGIN PRELIMINARY STEPS * ------------------------------------------------------------------* */ /** * ----------------------------------- BEGIN: CREATE IMAGES THAT WILL BE USED IN COMPUTATIONS * -----------------------------------* */ // Create the smoothed image to be used for the different edge images ImageProcessor Smoothed_Ip_xf = new FloatProcessor(w, h); ImageProcessor Smoothed_Ip_yf = new FloatProcessor(w, h); ImageProcessor Smoothed_Ip_45f = new FloatProcessor(w, h); ImageProcessor Smoothed_Ip_135f = new FloatProcessor(w, h); // Create the edge image detecting vertical edges ImageProcessor G_xf_Ip = new FloatProcessor(w, h); // Create the edge image detecting horizontal edges ImageProcessor G_yf_Ip = new FloatProcessor(w, h); // Create the edge image detecting 135 degree edges ImageProcessor G_45f_Ip = new FloatProcessor(w, h); // Create the edge image detecting 45 degree edges ImageProcessor G_135f_Ip = new FloatProcessor(w, h); // Create the gradient magnitude image ImageProcessor GMag_Ip = new ByteProcessor(w, h); // Byte version ImageProcessor GMagf_Ip = new FloatProcessor(w, h); // Floating point version // Create the gradient direction image ImageProcessor GDir_Ip = new ByteProcessor(w, h); // Byte version ImageProcessor GDirf_Ip = new FloatProcessor(w, h); // Floating point version // Create the edge image (output from non-maximal suppression) ImageProcessor Edge_Ip = new ByteProcessor(w, h); // Byte Version Edge_Ip.setValue(0); // 0 = Black Edge_Ip.fill(); // Fill Edge_Ip all black ImageProcessor Copy_Edge_Ip = new ByteProcessor(w, h); // Byte Version Copy_Edge_Ip.setValue(0); // 0 = Black Copy_Edge_Ip.fill(); // Fill Edge_Ip all black ImageProcessor Edgef_Ip = new FloatProcessor(w, h); // Floating Point version Edgef_Ip.setValue(0); // 0 = Black Edgef_Ip.fill(); // Fill Edge_Ip all black ImageProcessor Copy_Edgef_Ip = new FloatProcessor(w, h); // Floating Point version Copy_Edgef_Ip.setValue(0); // 0 = Black Copy_Edgef_Ip.fill(); // Fill Edge_Ip all black // Create the Threshold with Hysteresis image ImageProcessor Threshold_Ip = new ByteProcessor(w, h); Threshold_Ip.setValue(0); // 0 = Black Threshold_Ip.fill(); // Fill Edge_Ip all black /** * ----------------------------------- END: CREATE IMAGES THAT WILL BE USED IN COMPUTATIONS * -----------------------------------* */ /** * ----------------------------------------------------- BEGIN: USER INPUT * -----------------------------------------------------* */ double STDDev = 0, TLow = 0, THigh = 0; // Declare and initialize variables for the standard deviation, low threshold, and // high threshold int Size = 0; // Initialize size of Gaussian Filter boolean EdgeStrengthImage; // specifies whether edge strength image should be shown GenericDialog gd = new GenericDialog("User Inputs"); gd.addNumericField( "Size of Gaussian Filter (Odd Integer)", Size, 0); // Field for Size of Gaussian Filter gd.addNumericField("Standard Deviation", STDDev, 0); // Field for Standard Deviation gd.addNumericField("Low Threshold (1 - 255)", TLow, 0); // Field for Low Threshold gd.addNumericField("High Threshold (1 - 255)", THigh, 0); // Field for High Threshold gd.showDialog(); if (gd.wasCanceled()) { return; } else { Size = (int) gd .getNextNumber(); // Set Size variable from user input. This allows the user to // set the size of the Gaussian Filter STDDev = gd.getNextNumber(); // Set STDDev variable from user input TLow = gd.getNextNumber(); // Set TLow variable from user input THigh = gd.getNextNumber(); // Set THigh variable from user input } /** * ----------------------------------------------------- END: USER INPUT * -----------------------------------------------------* */ /** * ------------------------------------------------------------------ END PRELIMINARY STEPS * ------------------------------------------------------------------* */ /** * ------------------------------------------------------------------ BEGIN CANNY EDGE DETECTION * ------------------------------------------------------------------* */ /** * ----------------------------------------------------- BEGIN STEP 1: NOISE REDUCTION VIA * GAUSSIAN ----------------------------------------------------- * */ ImageProcessor ipf = ip.convertToFloat(); // Convert original image to floating point int SizeSquared = (int) Math.pow(Size, 2); // Square the size of Gaussian Kernel int HalfSize = (Size - 1) / 2; // Cut the size of the Gaussian Kernel almost in half float pix; // Temporary storage to be used throughout code to store pixel values float[] GaussianFilter = new float[SizeSquared]; // Initialize Gaussian Filter to be used in the convolution double[] GaussianFilterD = new double [SizeSquared]; // Initialize Gaussian Filter to be used...this filter will be composed // of numbers of type double double Constant = 1 / (2 * Math.PI * Math.pow(STDDev, 2)); // Compute 1/(2*pi*sigma^2) double ExponentDenom = 2 * Math.pow(STDDev, 2); // Compute 2*sigma^2 double Value; // Temporary storage to store the computations that will form the Gaussian Filter // FOR LOOP TO FORM GaussianFilterD for (int i = 0; i < Size; i++) { for (int j = 0; j < Size; j++) { Value = Math.exp( -1 * (Math.pow(j - HalfSize, 2) + Math.pow(i - HalfSize, 2)) / (ExponentDenom)); // Set Value = e^(-(i^2 + j^2)/(2*sigma^2)) GaussianFilterD[Size * i + j] = Constant * Value; // Place Value in GaussianFilterD } } // FOR LOOP TO FORM GaussianFilter using GaussianFilterD for (int i = 0; i < Math.pow(Size, 2); i++) { GaussianFilter[i] = (float) GaussianFilterD[i]; // Convert double values to float one by one } // CONVOLVE IMAGE WITH GAUSSIAN Convolver cv = new Convolver(); // Create the convolver cv.setNormalize(true); // Normalize the filter cv.convolve( ipf, GaussianFilter, Size, Size); // Apply the GaussianFilter using convolution on the image ipf // For loop to create 4 smoothed images to be used in detecting the 4 edges: 0 deg, 45 deg, 90 // deg, 135 deg for (int i = 0; i < w; i++) { for (int j = 0; j < h; j++) { pix = ipf.getf(i, j); Smoothed_Ip_xf.setf( i, j, pix); // Smoothed image to convolve with Sobel filter in x-direction (detects vertical // edges) Smoothed_Ip_yf.setf( i, j, pix); // Smoothed image to convolve with Sobel filter in y-direction (detects horizontal // edges) Smoothed_Ip_45f.setf( i, j, pix); // Smoothed image to convolve with Sobel filter in 45 degree direction (detects // 135 degree edges) Smoothed_Ip_135f.setf( i, j, pix); // Smoothed image to convolve with Sobel filter in 135 degree direction (detects // 45 degree edges) } } /** * ----------------------------------------------------- END STEP 1: NOISE REDUCTION VIA * GAUSSIAN ----------------------------------------------------- * */ /** * ----------------------------------------------------- BEGIN STEP 2: COMPUTE GRADIENT * MAGNITUDE AND DIRECTION IMAGES ----------------------------------------------------- * */ // Sobel Filters to detect edges float[] G_x = {-1, 0, 1, -2, 0, 2, -1, 0, 1}; // Sobel operator in x direction float[] G_y = {-1, -2, -1, 0, 0, 0, 1, 2, 1}; // Sobel operator in y direction float[] G_135 = {2, 1, 0, 1, 0, -1, 0, -1, -2}; // Sobel operator in 135 degree direction float[] G_45 = {0, 1, 2, -1, 0, 1, -2, -1, 0}; // Sobel operator in 45 degree direction // CONVOLVE IMAGE USING SOBEL FILTERS cv.convolve( Smoothed_Ip_xf, G_x, 3, 3); // Apply the Sobel filter in x-direction using convolution on the smoothed image cv.convolve( Smoothed_Ip_yf, G_y, 3, 3); // Apply the Sobel filter in y-direction using convolution on the smoothed image cv.convolve( Smoothed_Ip_45f, G_45, 3, 3); // Apply the Sobel filter in 45 degree direction using convolution on the smoothed image cv.convolve( Smoothed_Ip_135f, G_135, 3, 3); // Apply the Sobel filter in 135 degree direction using convolution on the smoothed // image // For loop to define the 4 floating point images G_xf_Ip, G_yf_Ip, G_45f_Ip, G_135f_Ip for (int i = 0; i < w; i++) { for (int j = 0; j < h; j++) { pix = Smoothed_Ip_xf.getf(i, j); G_xf_Ip.setf(i, j, pix); pix = Smoothed_Ip_yf.getf(i, j); G_yf_Ip.setf(i, j, pix); pix = Smoothed_Ip_45f.getf(i, j); G_45f_Ip.setf(i, j, pix); pix = Smoothed_Ip_135f.getf(i, j); G_135f_Ip.setf(i, j, pix); } } // COMPUTE THE GRADIENT MAGNITUDE IMAGE GMagf_Ip float pix1, pix2; // Variables to store pixel values of G_xf_Ip and G_yf_Ip for (int i = 0; i < w; i++) { for (int j = 0; j < h; j++) { pix1 = G_xf_Ip.getf(i, j); // Get pixel value pix2 = G_yf_Ip.getf(i, j); // Get Pixel Value pix = (float) Math.sqrt( pix1 * pix1 + pix2 * pix2); // Take the square root of the sum of the squares of both pixel // values GMagf_Ip.setf( i, j, 255 * pix); // Place in (i,j) position of GMagf_Ip and scale by 255 to view } } // COMPUTE THE GRADIENT DIRECTION IMAGE double v1, v2, v3, v4; // Temporary storage for the abs. value of pixel vlaues int pixi; // Temporary storage for pixel values double[][] GDir = new double[h] [w]; // Define 2D Array to store gradient direction values. Will use this during Step 3 // Non-Maximal Suppression instead of the gradient direction image // For loop to compute gradient direction image for (int i = 0; i < w; i++) { for (int j = 0; j < h; j++) { // Obtain the absolute value of each pixel in the following four floating point images v1 = Math.abs(G_xf_Ip.getf(i, j)); v2 = Math.abs(G_yf_Ip.getf(i, j)); v3 = Math.abs(G_45f_Ip.getf(i, j)); v4 = Math.abs(G_135f_Ip.getf(i, j)); // If statements to find the maximal response (strongest edge direction) if (v1 > v2 && v1 > v3 && v1 > v4) // Vertical edge orientation { GDir_Ip.putPixel(i, j, 0); GDir[j][i] = 0; // 0 stands for vertical edge orientation } else if (v3 > v1 && v3 > v2 && v3 > v4) // 45 degree edge orientation { GDir_Ip.putPixel(i, j, 1); GDir[j][i] = 1; // 1 stands for 45 degree edge orientation } else if (v2 > v1 && v2 > v3 && v2 > v4) // Horizontal edge orientation { GDir_Ip.putPixel(i, j, 2); GDir[j][i] = 2; // 2 stands for horizontal edge orientation } else if (v4 > v1 && v4 > v2 && v4 > v3) // 135 degree edge orientation { GDir_Ip.putPixel(i, j, 3); GDir[j][i] = 3; // 3 stands for 135 degree edge orientation } pixi = (int) (255.0 / 3.0) * GDir_Ip.getPixel( i, j); // Scale the gradient direction image so we can actually see it GDir_Ip.putPixel( i, j, pixi); // Place pixi in the (i,j) location of GDir_Ip...we can now view the gradient // direction image } } // OUTPUT GRADIENT DIRECTION IMAGE String DirTitle = "Gradient Direction"; ImagePlus GDir_Im = new ImagePlus(DirTitle, GDir_Ip); GDir_Im.show(); // OUTPUT GRADIENT MAGNITUDE IMAGE (need to convert to ByteProcessor to view first) GMagf_Ip.resetMinAndMax(); GMag_Ip.insert(GMagf_Ip.convertToByte(true), 0, 0); String MagTitle = "Gradient Magnitude"; ImagePlus GMag_Im = new ImagePlus(MagTitle, GMag_Ip); GMag_Im.show(); /** * ----------------------------------------------------- BEGIN STEP 2: COMPUTE GRADIENT * MAGNITUDE AND DIRECTION IMAGES ----------------------------------------------------- * */ /** * ----------------------------------------------------- BEGIN STEP 3: NON-MAXIMAL SUPPRESSION * -----------------------------------------------------* */ // Ignoring boundaries for convenience float Magnitude; // Storage for magnitude double Direction; // Storage for direction // FOR LOOP TO COMPUTE NON-MAXIMAL SUPPRESSION IMAGE for (int i = 1; i < w - 1; i++) { for (int j = 1; j < h - 1; j++) { Magnitude = GMagf_Ip.getf( i, j); // Get the magnitude in the (i,j) position of the gradient magnitude image if (Magnitude != 0) // If the magnitude is non-zero then we find the direction of the edge { Direction = GDir[j][i]; // Obtain the direction float n1GradMag = 0, n2GradMag = 0; // Initialize storage for the magnitude of the 2 neighboring pixels if (Direction == 0) // If Direction is 0 get the magnitude in the columns to the left and right { n1GradMag = GMagf_Ip.getf(i - 1, j); n2GradMag = GMagf_Ip.getf(i + 1, j); } else if (Direction == 1) // If Direction is 45 degree get the magnitude in adjacent pixels { n1GradMag = GMagf_Ip.getf(i + 1, j - 1); n2GradMag = GMagf_Ip.getf(i - 1, j + 1); } else if (Direction == 2) // If Direction is 2 get the magnitude in the rows above and below { n1GradMag = GMagf_Ip.getf(i, j - 1); n2GradMag = GMagf_Ip.getf(i, j + 1); } else if (Direction == 3) // If Direction is 135 degrees get the magnitude in adjacent pixels { n1GradMag = GMagf_Ip.getf(i - 1, j - 1); n2GradMag = GMagf_Ip.getf(i + 1, j + 1); } if (Magnitude > n1GradMag && Magnitude > n2GradMag) // Check to see if the magnitude of pixel under inspection is the // largest relative to its adjacent pixels { pix = GMagf_Ip.getf(i, j); // Store magnitude in (i,j) position in pix variable Edgef_Ip.setf( i, j, pix); // Place this pixel value in the (i,j) position of Edge Image (Edge Image is // output from non-maximal suppression) Copy_Edgef_Ip.setf( i, j, pix); // Place this pixel value in the (i,j) position of Copy Edge Image (this will // be used for thresholding with hysteresis) } else { } // Else do nothing since edge image was already filled with 0's at beginning of code } else { } // Do nothing if magnitude is 0 } } Copy_Edgef_Ip.resetMinAndMax(); Copy_Edge_Ip.insert(Copy_Edgef_Ip.convertToByte(true), 0, 0); // Display edge magnitude image but we need to convert to ByteProcessor first Edgef_Ip.resetMinAndMax(); Edge_Ip.insert(Edgef_Ip.convertToByte(true), 0, 0); String cTitle = "Non-Maximal Suppression"; ImagePlus Edge_Im = new ImagePlus(cTitle, Edge_Ip); Edge_Im.show(); /** * ----------------------------------------------------- END STEP 3: NON-MAXIMAL SUPPRESSION * -----------------------------------------------------* */ /** * ----------------------------------------------------- BEGIN STEP 4: THRESHOLDING WITH * HYSTERESIS -----------------------------------------------------* */ int count = 0; // Initialize counter to be used in while loop int IterationCount = 500; // Number of iterations to perform // Scan through all pixels in the edge image and mark all edges with magnitude above the high // threshold as a true edge otherwise if the magnitude is below the low threshold // then we delete that pixel (set it to zero) for (int i = 0; i < w; i++) { for (int j = 0; j < h; j++) { Magnitude = Copy_Edge_Ip.getPixel(i, j); // Obtain magnitude in edge image if (Magnitude > THigh) // If Magnitude is larger than the high threshold then make pixel white { Threshold_Ip.putPixel(i, j, 255); // True edge (Updates Threshold image to be output) Copy_Edge_Ip.putPixel( i, j, 255); // True edge (Updates edge image which will be used in the iterative // thresholding) } else if (Magnitude < TLow) // Else if magnitude is below the low threshold then make pixel black { Threshold_Ip.putPixel(i, j, 0); // Not an edge (Updates Threshold image to be output) Copy_Edge_Ip.putPixel( i, j, 0); // Not an edge (Updates edge image which will be used in the iterative // thresholding) } } } while (count < IterationCount) // Iterate again and again { // Ignore boundary pixels for convenience for (int i = 1; i < w - 1; i++) { for (int j = 1; j < h - 1; j++) { Magnitude = Copy_Edge_Ip.getPixel(i, j); // Obtain magnitude in edge image if (Magnitude == 255) // If we reach a true edge then we look at its 8 neighbors { // Obtain the magnitude in the 8 neighbors int n1, n2, n3, n4, n5, n6, n7, n8; n1 = Copy_Edge_Ip.getPixel(i - 1, j); n2 = Copy_Edge_Ip.getPixel(i - 1, j - 1); n3 = Copy_Edge_Ip.getPixel(i, j - 1); n4 = Copy_Edge_Ip.getPixel(i + 1, j - 1); n5 = Copy_Edge_Ip.getPixel(i + 1, j); n6 = Copy_Edge_Ip.getPixel(i + 1, j + 1); n7 = Copy_Edge_Ip.getPixel(i, j + 1); n8 = Copy_Edge_Ip.getPixel(i - 1, j + 1); if (n1 >= TLow) // If n1 is greater than or equal to the low threshold then we mark it as // an edge { Copy_Edge_Ip.putPixel(i - 1, j, 255); // Update edge image Threshold_Ip.putPixel( i - 1, j, 255); // Update threshold image (This is the output image from thresholding with // hysteresis) } if (n2 >= TLow) // If n2 is greater than or equal to the low threshold then we mark it as // an edge { Copy_Edge_Ip.putPixel(i - 1, j - 1, 255); // Update edge image Threshold_Ip.putPixel( i - 1, j - 1, 255); // Update threshold image (This is the output image from thresholding with // hysteresis) } if (n3 >= TLow) // If n3 is greater than or equal to the low threshold then we mark it as // an edge { Copy_Edge_Ip.putPixel(i, j - 1, 255); // Update edge image Threshold_Ip.putPixel( i, j - 1, 255); // Update threshold image (This is the output image from thresholding with // hysteresis) } if (n4 >= TLow) // If n4 is greater than or equal to the low threshold then we mark it as // an edge { Copy_Edge_Ip.putPixel(i + 1, j - 1, 255); // Update edge image Threshold_Ip.putPixel( i + 1, j - 1, 255); // Update threshold image (This is the output image from thresholding with // hysteresis) } if (n5 >= TLow) // If n5 is greater than or equal to the low threshold then we mark it as // an edge { Copy_Edge_Ip.putPixel(i + 1, j, 255); // Update edge image Threshold_Ip.putPixel( i + 1, j, 255); // Update threshold image (This is the output image from thresholding with // hysteresis) } if (n6 >= TLow) // If n6 is greater than or equal to the low threshold then we mark it as // an edge { Copy_Edge_Ip.putPixel(i + 1, j + 1, 255); // Update edge image Threshold_Ip.putPixel( i + 1, j + 1, 255); // Update threshold image (This is the output image from thresholding with // hysteresis) } if (n7 >= TLow) // If n7 is greater than or equal to the low threshold then we mark it as // an edge { Copy_Edge_Ip.putPixel(i, j + 1, 255); // Update edge image Threshold_Ip.putPixel( i, j + 1, 255); // Update threshold image (This is the output image from thresholding with // hysteresis) } if (n8 >= TLow) // If n8 is greater than or equal to the low threshold then we mark it as // an edge { Copy_Edge_Ip.putPixel(i - 1, j + 1, 255); // Update edge image Threshold_Ip.putPixel( i - 1, j + 1, 255); // Update threshold image (This is the output image from thresholding with // hysteresis) } } } } count++; // Update counter and continue iterations } // Display the threshold with hysteresis image. String ThreshTitle = "Threshold with Hysteresis"; ImagePlus Threshold_Im = new ImagePlus(ThreshTitle, Threshold_Ip); Threshold_Im.show(); /** * ----------------------------------------------------- END STEP 4: THRESHOLDING WITH * HYSTERESIS -----------------------------------------------------* */ /** * ------------------------------------------------------------------ END CANNY EDGE DETECTION * ------------------------------------------------------------------* */ }