Median filtering is a nonlinear process useful in reducing impulsive, or salt-and-pepper noise. The median filter is also used to preserve edge properties while reducing the noise. Also, the smoothing techniques, like Gaussian blur is also used to reduce noise but it can’t preserve the edge properties. The median filter is widely used in digital image processing just because it preserves edge properties.
Approach :
- Store the pixel values of input image in an array.
- For each pixel value store all the neighbor pixel value including that cell in a new array (called window).
- Sort the window array.
- Median of window array is used to store output image pixel intensity.
Boundary Issues Example:
2D Median filtering example using a 3 x 3 sampling window:
Extending border values outside with values at the boundary.
Edge preservation :
All smoothing techniques are used to remove noise. The median filter is also one kind of smoothing technique like Gaussian filter, but the only difference between the median filter and Gaussian filter is that the median filter preserves edge property while Gaussian filter does not. Edge preservation is an important property because edges are important for visual appearance. For edge preservation property median filter is widely used in digital image processing.
C++ implementation of median filter algorithm.
C++
#include <iostream>#include <fstream>#include <sstream>using namespace std;/* Function to sort an array using insertion sort*/void insertionSort(int arr[], int n){ int i, key, j; for (i = 1; i < n; i++) { key = arr[i]; j = i - 1; /* Move elements of arr[0..i-1], that are greater than key, to one position ahead of their current position */ while (j >= 0 && arr[j] > key) { arr[j + 1] = arr[j]; j = j - 1; } arr[j + 1] = key; }}int array[2000][2000];int arr[2000][2000];int main(){ int window[9],row = 0, col = 0, numrows = 0, numcols = 0,MAX=0; ifstream infile("Saltpepper.pgm"); stringstream ss; string inputLine = ""; // First line : version getline(infile,inputLine); if(inputLine.compare("P2") != 0) cerr << "Version error" << endl; else cout << "Version : " << inputLine << endl; // Continue with a stringstream ss << infile.rdbuf(); // Secondline : size of image ss >> numcols >> numrows >> MAX; //print total number of rows, columns and maximum intensity of image cout << numcols << " columns and " << numrows << " rows" <<endl<<" Maximum Intensity "<< MAX <<endl; //Initialize a new array of same size of image with 0 for(row = 0; row <= numrows; ++row) array[row][0]=0; for( col = 0; col<=numcols; ++col ) array[0][col]=0; // Following lines : data for(row = 1; row <= numrows; ++row) { for (col = 1; col <= numcols; ++col) { //original data store in new array ss >> array[row][col]; } } // Now print the array to see the result for(row = 1; row <= numrows; ++row) { for(col = 1; col <= numcols; ++col) { //neighbor pixel values are stored in window including this pixel window[0] = array[row-1][col-1]; window[1] = array[row-1][col]; window[2] = array[row-1][col+1]; window[3] = array[row][col-1]; window[4] = array[row][col]; window[5] = array[row][col+1]; window[6] = array[row+1][col-1]; window[7] = array[row+1][col]; window[8] = array[row+1][col+1]; //sort window array insertionSort(window,9); //put the median to the new array arr[row][col]=window[4]; } } ofstream outfile; //new file open to store the output image outfile.open("Medianfilter.pnm"); outfile<<"P2"<<endl; outfile<<numcols<<" "<<numrows<<endl; outfile<<"255"<<endl; for(row = 1; row <= numrows; ++row) { for (col = 1; col <= numcols; ++col) { //store resultant pixel values to the output file outfile << arr[row][col]<<" "; } } outfile.close(); infile.close(); return 0 ;} |
Input Image
Output Image
