How can I convert an integer matrix to a real-valued matrix to perfrom SVD

A = imread('image');

SVD = svd(A)

Answer 1

[John D'Errico]

help double

 DOUBLE Convert to double precision.
    DOUBLE(X) returns the double precision value for X.
    If X is already a double precision array, DOUBLE has
    no effect.
 
    DOUBLE is called for the expressions in FOR, IF, and WHILE loops
    if the expression isn't already double precision.  DOUBLE should
    be overloaded for all objects where it makes sense to convert it
    into a double precision value.
 
    See also SINGLE, DATATYPES, ISFLOAT, ISNUMERIC.

    Documentation for double
       doc double

    Other functions named double

       categorical/double      duration/double      quaternion/double
       codistributed/double    filtstates/double    StaticModel/double
       dataset/double          gpuArray/double      string/double
       datetime/double         opaque/double        tabular/double
       dlarray/double

[Jorge Arturo Clares Pastrana]

A = imread('image');

B = double(A);

B still gives me a 3d matrix

[John D'Errico]

Um, so? Your point is what? I presumed that since you knew what you wanted to do, then you understood enough about the SVD to know that it is meaningless to try to apply it to a 3-d matrix. Therefore your image must already be a single channel image.

So let me go back. You cannot compute the SVD of a 3-d matrix. That is meaningless. So YOU need to decide what you will compute the SVD on. And only you can decide that.

If, for example, you wanted to compute the SVD of only the red channel, that would be trivial. Or the green, or the blue channels. Just extract the indicated channel, then convert to double, then compute the SVD.

Up to you, entirely so. Or, perhaps you wanted to compute the SVD of the image as a greyscale matrix. So convert to grey FIRST, and THEN convert to a double array. At that point, the image will be only a single channel, so amenable to a SVD.

Or you might decide to convert the entire image to a color space like LAB. THEN compute the svd of L*, etc. Again, your choice.

Or, you might decide to reshape the matrix in some way to be a 2-dimensinal matrix. Again, only you know what and why you are doing this operation.

I'm sure there are other ways you might decide to transform the image into a 3D matrix. Just give me 10 seconds and I could think of a few more. But that would be again, your choice, because I'm pretty sure it is you who knows what it is you want to compute the SVD on, and why you are doing exactly that.