How do i vectorize an image?

I am trying to break a square matrix of an image and create another vector which contains its column vector sub-blocks. I used the following code for it and getting an error

for i=1:50    a(i)=A(:,i)

The error says 'In an assignment A(I) = B, the number of elements in B and I must be the same.'

Please help me how i can solve my problem.

[Hiren Parmar]

A = imread('cameraman.tif'); % your image

a = A(:); % Image matrix is converted to column vector.

[Prashanna Jain]

thankyou for the input but i'm not trying to make a single vector. I'm trying to convert my NxN image into N column vectors, each of those vector have columns from the NxN image. i hope you get my point.

[Stephen23]

Is the image an RGB image, or indexed or grayscale? If the image is RGB then is has size M*N*3. Can you please tell us exactly the output of

size(A); % A is your image

[Varun Pai]

Then that will be same as your original NxN image. Can you please show your requirement by using a 3 x 3 matrix as an example ?

[Prashanna Jain]

if A is a 3x3 matrix with elements [1,2,3,; 4,5,6,; 7,8,9] then i want to create a vector B such that its first element has [1,4,7], second has [2,5,8], and third has [3,6,9]

[Prashanna Jain]

Stephen Cobeldick : how do i proceed if i want to do the same for an RGB image?

[Stephen23]

@Prashanna Jain: the answer depends on you. Each pixel of an RGB image is made up of the R, G and B values, so do you want to keep these values together (with each column that you are splitting the image into), or do you want to split each column of the image into three separate variables (one for each of R, G and B).

In the first case you will get a matrix of some kind, in the second case you get three vectors. Which do you want?

It is probably a lots easier to not split your data up anyway. Why do you need to split the image into vectors? Why not just use indexing and access the parts that you need to?

[Prashanna Jain]

i want to write an algorithm to find the KL Transform of the image, for which i need to vectorize the image to find the mean and covariance. converting it to grayscale and double will lead to loss in data..so i might have to use the coloured version.

[Walter Roberson]

mean(Array) gives a vector of column means, you do not need to split it up. Likewise with std()

Answer 1

[Walter Roberson]

a = mat2cell(A, size(A,1), ones(1,size(A,2)));

Now a{K} will be the column vector corresponding to A(:,K)

[Walter Roberson]

For RGB one solution would be

a = mat2cell(A, size(A,1), ones(1,size(A,2)), size(A,3));

This would create a{k} as being an N x 3 matrix, rather than creating the vectors that were asked for. Or you could use

a = mat2cell(A, size(A,1), ones(1,size(A,2)), ones(1,size(A,3)));

which would create a as an N x 3 cell array, each entry of which is a column vector.

[Prashanna Jain]

but this is easy to implement for a 3x3 matrix..i will have to apply it on an image of about 100x100..writing it for the entire thing would be tedious no?

Answer 2

[Thorsten]

Why don't you just get your vectors from A using

A(:,i);

I can see no reason why a you need to re-organize your data.

[Prashanna Jain]

i did try a(i)=A(:,i); under a for loop..it gave an error as mentioned in my problem above

[Prashanna Jain]

also this will only work on a 2D image..what do i do for an rgb one?