Re: adding noise to an image

[Paul Davis <pjdavis engineering uiowa edu>]

Foster, Gareth wrote:

> > Don't you just set rgb to the same value for grey scale?
> >
> >  uint8 * p_buf = get_buf_blah_pseudo();
> >  uint8 by_val = rand()%std::numeric_limits<uint8>::max_val;
> >  std::fill(by_val, by_val + 3, by_val);
> >
> >    
>
> Gawd, how embarrising ...
>
>  uint8 * p_buf = get_buf_blah_pseudo();
>  uint8 by_val = rand()%std::numeric_limits<uint8>::max_val;
>  std::fill(p_buf, p_buf + 3, by_val);
>
>
> Awoooo! Howling!
> _______________________________________________
> gtkmm-list mailing list
> gtkmm-list gnome org
> http://mail.gnome.org/mailman/listinfo/gtkmm-list
>  
If I'm not mistaken, that would create an image of one pixel value.

Some assumptions:
img_size = image_width * image_height
get_image returns a 1D array (  generally the best way for storing image 
data )
max_noise_magnitude is an unsigned char that is controlled by the slider.

MAX_VAL is the maximum value per pixel( 255 in this case using unsigned 
char )

An implementation would be something more like:

unsigned char* img = get_image() ;

//Assuming a grey scale image ( ie one value per pixel )
for( int i = 0 ; i < img_size ; i++ )
{
   //I read somewhere once that you should use the higher order bits 
for better randomness.
   //Hopefully this gives you a random value from -max / 2 to max / 2 ( 
or there abouts )
   char alter_val = ( rand() >> 24 ) % max_noise_magnitude - ( 
max_noise_magnitude / 2 ) ;
   if( img[ i ] + alter_val < 0 )
   {
      img[i] = 0 ;
   }
   else if( img[i] + alter_val > MAX_VAL )
   {
      img[i] = MAX_VAL ;
   }
   else
   {
      img[i] += alter_val ;
   }
}

//Should have our noisy image here.

Then again if you really want to get into fun stuff, adding white noise 
( ie noise of in a wideband frequency range ) you should look at doing 
an FFT and adding noise in the frequency domain.  But thats probably a 
bit more drastic than the need for this application.

Adding noise is one of those "there is no right way" things.  Its kind 
of like, does this look noisy to you?  There are different types of 
noise and different ways to add it.  But then again I'm not really heavy 
into image processing so I'm not sure what to say is the recommended 
way.  Go read about image processing concepts.  Those people know what 
they're talking about a good deal of the time.

Cheers,
Paul