Page 54 -

28 Chapter 2 ■ Edge-Detection Techniques

/* Measure the Normally distributed noise in a small region.
Assume that the mean is zero. */

#include <stdio.h>
#include <math.h>
#define MAX
#include “lib.h“

main(int argc, char *argv[])
{
IMAGE im;
int i,j,k;
float x, y, z;
double mean, sd;

im = Input_PBM (argv[1]);
/* Measure */
k= 0;
x = y = 0.0;
for (i=0; i<im->info->nr; i++)
for (j=0; j<im->info->nc; j++)
{
x += (float)(im->data[i][j]);
y += (float)(im->data[i][j]) * (float)(im->data[i][j]);
k+=1;
}

/* Compute estimate - mean noise is 0 */
sd = (double)(y - x*x/(float)k)/(float)(k-1);
mean = (double)(x/(float)k);
sd = sqrt(sd);
printf (“Image mean is %10.5f Standard deviation is %10.5f\n“,
mean, sd);

/* Now assume that the uniform level is the mean, and compute the
mean and SD of the differences from that! */
x = y = z = 0.0;
for (i=0; i<im->info->nr; i++)
for (j=0; j<im->info->nc; j++)
{
z = (float)(im->data[i][j] - mean);
x+= z;
y += z*z;
}
sd = (double)(y - x*x/(float)k)/(float)(k-1);
mean = (double)(x/(float)k);
sd = sqrt(sd);
printf (“Noise mean is %10.5f Standard deviation is %10.5f\n“,
mean, sd);
}

Figure 2.6: A C program for estimating the noise in an image. The input image is sampled
from the image to be measured, and must be a region that would ordinarily have a
constant grey level.