paddnoise

Synopsis

Description

paddnoise adds artificial noise to the input image im_in. Several distribution laws exist either additive or multiplicative and can be chosen from the parameter law. The output image im_out is built as follows:

• additive law: im_out = im_in + im_noise;
• multiplicative law: im_out = im_in * im_noise;

where im_noise is a noise image built from the specified distribution law.

Let u1 and u2 be two random variables uniformly distributed on the space [0..1]. The noise value for pixel p is computed as follows:

• gaussian law (Box Muller algorithm):

    z0=sqrt(-2.0*log(u1))*cos(2.0*M_PI*u2);
    im_noise[p] = std_dev*z0 + mean;
  

• exponential law (inversion method):

    z0=-1.0*log(u1);
    im_noise[p]= std_dev*z0 + mean;
  

• uniform law (rem : std_dev = (max-min)/sqrt(12)):

    z0=(u1-0.5)*sqrt(12.0);
    im_noise[p] = std_dev*z0 + mean;
  

• triangular law:

    z0=(u1+u2-1.0)*sqrt(6.0);  
    im_noise[p] = std_dev*z0 + mean;
  

Parameters

• law is an integer which specifies the distribution law:
  • 1: additive gaussian noise.
  • 2: additive exponential noise.
  • 3: additive uniform noise.
  • 4: additive triangular noise.
  • 11: multiplicative gaussian noise.
  • 12: multiplicative exponential noise.
  • 13: multiplicative uniform noise.
  • 14: multiplicative triangular noise.
• mean is a real which defines the mean of the distribution.
• std_dev is a real which defines the standard deviation of the distribution.

Inputs

• im_in: an image.

Outputs

• im_out: an image of the same type as im_in.

Result

Returns SUCCESS or FAILURE in case of bad parameter values.

Examples

Adds gaussian noise with mean 0 and standard deviation 1.5 to tangram.pan image and then computes the PSNR for the meanfilter smoothing operator:

   paddnoise 1 0 1.5 tangram.pan a.pan
   pmeanfilter 2 a.pan i1.pan
   ppsnr 255 tangram.pan i1.pan
   pstatus

See also

C++ prototype

Version française

Génération de bruit aléatoire sur une image.