The contrast function

[John W. Eaton]

On 25-Mar-2008, David Bateman wrote:

| The contrast function should be easy to implement.. It just comes down
| to normalizing the data to the number of elements in the colormap and
| counting the number of elements in the data between each integer value..
| I've implemented a function that basically does this, this being
| 
| function map = contrast (x, n)
|   if (nargin == 1)
|     n = rows (colormap);
|   elseif (nargin == 2)
|     if (! isscalar (n))
|       error ("contrast: n must be a scalar");
|     endif
|   else
|     print_usage ();
|   endif
| 
|   x = x (:);
|   minx = min (x);
|   map = find (diff (sort ([round(n * ((x - minx) ./ (max(x) - minx)));
| [0:n]']))) - [1:n]';
|   minm = min (map);
|   map = (map - minm) ./ (max (map) - minm);
|   map = [map , map, map];
| endfunction
| 
| though I see a difference with the matlab behavior for the same
| function.. It appears that matlab doesn't define the first term of the
| colormap as [0,0,0] and so the maximum contrast can't be achieved as
| pure black is excluded from the colormap. This can be seen with
| 
| imagesc(reshape(1:100,10,10))
| map = contrast (1:100,10)
| colormap(map)
| 
| It seems to me better to use both [0,0,0] and [1,1,1] as the limits of
| the colormap for contrast. Does it make sense to have this inconsistency
| with matlab? If so I'll propose a patch to add contrast based on the above

It does seem like a bug that 1 is always included but you can only
approach 0 as N -> Inf (if I understand correctly).  Unless there is
some good reason for doing that, I'd say go with what you have.

jwe