covariance of a matrix which is not a square matrix

we have a matrix of dimension (65536X12) how can we find covariance of such a matrix which is not a square matrix

Answer 1

[Andrew Newell]

As was true of your previous post, your question is not clear. Do you have 65536 measurements each of 12 variables, or the reverse? If the former, then

cov(M)

(where M is your matrix) will give you a 12 x 12 covariance matrix for the variables. If the latter,

cov(M')

will give you a 65536 x 65536 matrix. Again, I recommend reading this discussion of covariance.

[Anuradha Kumari]

yes,I am getting 12 x 12 covariance matrix, but m loosing the data in this case because the matrix M here is that of an image and later in the program i have to reconstruct it into 256 x 256 to get the image

[Walter Roberson]

Okay, and how are you using the covariance matrix to do the reconstruction to 256 x 256 ?

[Andrew Newell]

I don't suppose it's a coincidence that 65536 = 256^2. Are you trying to correlate each pixel with each other pixel?

[Andrew Newell]

Maybe I should also ask - why are you trying to do this?