3

I have an rgb image matrix (height*width*3) represented in doubles. After some manipulations on the matrix, some values went biger then 1 or smaller then 0. I need to normalize those valuse back to 1 and 0. Thanks.

Sanich
  • 1,739
  • 6
  • 25
  • 43

2 Answers2

6

Well, just use the indexing by condition. Let's say your matrix is called M. If you just want to set values bigger than 1 to 1 and smaller than 0 to zero, use:

M(M > 1) = 1;
M(M < 0) = 0;

However, if you want to proportionally normalize all the values to the interval [0; 1], then you have to do something similar to:

mmin = min(M(:));
mmax = max(M(:));
M = (M-mmin) ./ (mmax-mmin); % first subtract mmin to have [0; (mmax-mmin)], then normalize by highest value

You have to take account of the case when your matrix M is already in the interval [0; 1] and the normalization is not needed.

Marek Kurdej
  • 1,459
  • 1
  • 17
  • 36
  • Thanks, indeed it does what i asked. But the result is not good enough so i guess that i need realy to normalize the matrix. the bigst value should be 1, smalest 0 , and everything else betwin proportionaly. – Sanich Nov 13 '11 at 12:26
  • Sorry, I didn't get it right. I've edited my answer. Hope, it's what you wanted. – Marek Kurdej Nov 13 '11 at 12:35
  • @Curdeius: If you want to crop the values in the range [0,1] (your first solution), you can also write it vectorized as: `M = min(max(M,0),1)` – Amro Nov 13 '11 at 13:28
  • @Curdeius: Maybe the last action should be `(M = (M-mmin) ./ (mmax-mmin);`)? (instead of .*) – Sanich Nov 13 '11 at 21:37
0

if you just want to see the images you can use 

imagesc(M);

it takes care of the range itself.

If you want to change the values manually and have full control over it, 

M = M ./ max(M(:));

would do the trick if you only have positive values. To get a full contrast image you might want to:

m = m - min(m(:));
m = m ./ max(m(:));
Ali
  • 18,665
  • 21
  • 103
  • 138