How do I apply SVD (Singular Value Decomposition) to an image?

The syntax given for singular value decomposition is svd(x).

I tried it with my image, but it didn't work. Can you tell me how to work with svd for images please?

[David Young]

Please could you say what the error message was, and also show any other parts of your code that might be relevant.

[Andreas Goser]

While I agree with David on the need for specifics, my crystal ball tells my this is about data types and will craft an answer for that...

Answer 1

[Andreas Goser]

This sounds like it is about data types or sizes. Example

pout = imread('pout.tif');
svd(pout) % does not work
??? Undefined function or method 'svd' for input arguments of type 'uint8'.
svd(double(pout)) % works

I can however not comment on the mathematical sense of this. I you have another image format like here, you need to think about what you actually like to achieve

I = imread('board.tif');
svd(double(I))
??? Undefined function or method 'svd' for input arguments of type 'double' and
attributes 'full 3d real'.

[buddy]

thank u sir.. i understand it is bcos of datatype..

Answer 2

[meenakshi]

HELLO GOSER

                   i=imread('pout.tif');
                   i=im2double(i)
                   [u s v]=svd(i);

you can try like this.

                     k.meenakshi

[Walter Roberson]

That would not have any more success than svd(double(I)) if I is a truecolor (3D) image. Remember, images can be stored as pseudocolor (2D arrays in which the values indicate which index to use out of a color map), or as truecolor (3D arrays in which the values directly indicate the color information for each pixel without any map.) The problem is that svd() of a pseudocolor image is not meaningful, and svd() of a 3D array is not allowed. The only choice available to get anything useful out is to convert the image to grayscale and svd() the grayscale image.

Answer 3

[slama najla]

Hello, can some body help me with the code of SVD decomposition in 3d medical data in matlab please.

[Walter Roberson]

No. SVD is only for 2D.

You could take the svd() of each plane.

Answer 4

[slama najla]

But many approaches use it us decomposition for 3d data in watermarking,this is why i reask this question.thanks

[Walter Roberson]

SVD is *defined* in terms of rectangular matrices. There is no method to apply SVD to a 3D matrix. I looked at some of the articles about color image watermarking using svd, and of the ones I could access, not one of them attempted to apply SVD to a 3D matrix.

[Neelam Yadav]

Hello, can anyone help me how I get different image for U,S,V in svd

Answer 5

[Ayesha Iftikhar]

Hello can any one help me how to use SVD for feature extraction

Answer 6

[vasanthi Muniasamy]

i=imread('pout.tif');

i=im2double(i)

[u s v]=svd(rgb2gray(i));

[toktam sameie]

Thank you for your great help