matrix normalization in matlab

hi.. i want normalize a matrix (129 x 128)..please help me with matlab codes

Answer 1

[Thorsten]

To normalize a matrix such that all values fall in the range [0, 1] use

Anorm = (A - min2(A))/(max2(A) - min2(A));

[moahaimen talib]

*how can i normalize the dwt2 coefficients *

[Irawati Roy]

You missed a dot before the division operation for a 2d matrix

[DGM]

There ./ is not needed here, since the denominator is a scalar.

That said, there is no function called min2() or max2(), nor can I find evidence that there were functions by those names which have since been removed. Maybe they used to be part of one of the toolboxes.

Alternatively, you could do:

Anorm = (A - min(A(:)))/(max(A(:)) - min(A(:)));

or you could do

Anorm = (A - min(A,[],'all'))/(max(A,[],'all') - min(A,[],'all'));

though the latter option wouldn't have worked circa 2013

Answer 2

[Matt J]

You need to decide how you want to define the amplitude of the matrix first. Maybe you want this

 normalized = A/norm(A);

or maybe this

 normalized = A/max(abs(A(:)))

[Jan]

Or perhaps "normalization" means a range of [0,1]?

[jenifer]

ya... but how can i get a single value from whole matrix...

[Jan]

Please, jenifer, we have asked you to define "normalization" exactly. It wastes time if we guess what you mean.

Answer 3

[Jing]

You can use 'normc' or 'normr' for normalization. 'c' or 'r' stands for columns or rows that you want to normalize.

Answer 4

[DGM]

Since R2018a, you can use normalize() to perform normalization in various different ways.

A = 1:6

A = 1×6

1 2 3 4 5 6

B1 = normalize(A,'range') % default range is [0 1]

B1 = 1×6

0 0.2000 0.4000 0.6000 0.8000 1.0000

B2 = normalize(A,'range',[-1 2]) % but you can specify any range

B2 = 1×6

-1.0000 -0.4000 0.2000 0.8000 1.4000 2.0000

B3 = normalize(A,'zscore','std') % zero-center and scale to have std=1

B3 = 1×6

-1.3363 -0.8018 -0.2673 0.2673 0.8018 1.3363

B4 = normalize(A,'norm',2) % normalize by vector p-norm (default p=2)

B4 = 1×6

0.1048 0.2097 0.3145 0.4193 0.5241 0.6290

[Sinan Islam]

The normalize function in matlab either works on columns or rows. We want a function that normalizes the matrix as a whole. We are not looking for normalizing features or observations. So the calculation must be made on global minimum and global maximum of the matrix. Thats why we flatten the matrix into one vector before doing any calculation.

[DGM]

As others have mentioned, "normalize" is a terribly vague word. If the goal is to rescale the entire array with respect to global statistics, then yes, reshaping the array is one way to do so.