Help for GAUSNOIS

PURPOSE:

GAUSNOIS generates an image with random pixel values where the pixel values
are distributed according to a gaussian (normal or bell shaped) probability
distribution.  This type of a noise image can be useful in tests.


EXECUTION:

GAUSNOIS  NOISE.IMG  NL=200 NS=300  MEAN=100 SIGMA=25 FORMAT=BYTE  SEED=7382382

This is an example using all of the parameters:  it produces a 200x300 byte
image with mean 100 and standard deviation 25, and starts the random number
generator with that particular seed.
OPERATION

The SEED parameter is provided so that the same random images can be generated.
If it is defaulted, the seed for the random number generator comes from the
system time.

Convenient seed numbers for test scripts have been: 16777215, 1073741969,
-7382382, 4617316,-74526, 94151, 97, -768825576, and -6.

Then two calls to rangen are used to generate x and y which plug into a formula

u = sqrt((-2.)*log(x)) * cos(2*3.1415927*y)

which, in turn are used to generate each sample value in the buffer.

sample = sigma*u + mean

All of the parameters except the output image can be defaulted.
PERFORMANCE:

    The following performance was directed towards VAX/VMS users:
	The ported version of GAUSNOIS is approximately half the speed of
the original GAUSNOIS.  The primary reason for this is that the random
number generator, RANGEN, is half the speed of the VAX/VMS routine RAN.
Steve Pohorsky agrees that this is unavoidable without a significant
rewrite to the program, including writing a seperate vectorized version of
the random number generator.

    The Linux version uses a call to the system time, in seconds passed
1 Jan 70 to initialize the seed.

HISTORY:

VMS Original Programmer :	Frank Evans	August 1985
Cognizant Programmer:   Ray Bambery

Revisions:
  1994-06    - Steve Hwan  - Ported to UNIX
  2011-04-15 - Ray Bambery - Fails under 64-bit linux
                    Internal call to get_seconds somehow corrupts
                    sigma parameter to always be zero. Fixed by
                    reordering xvp calls. This shouldn't be but
                    I could not find what was tromping on sigma.    
  2011-05-10 - Ray Bambery - Fixed failure to create file
                    on gfortran 4.4.4
  2012-06-06 - Ray Bambery - gfortran 4.6.3 revealed that
                    rangen parameters are (long,float) so 
                    created integer*8 dseed replacement for
                    integer*4 seed - sigma always became zero
                    when seed was entered into rangen
  2012-06-18 - Ray Bambery - get_seconds parameter is also
                    integer*8
                    rangen and get_seconds parameters were what
                    was causing sigma to be 0 in 4-15-2011 comments
  2013-08-11 - Ray Bambery - changed internal calls from
                   get_seconds to get_seconds_big and rangen to
                   rangen_big to prevent conflict with other
                   programs which call these routines
  2016-06-08 - Walt Bunch - Removed "_big" from get_seconds and 
                   rangen. Migrated back to MIPL.