Generating a N x N block Toeplitz Matrix out of N vectors

Hello all,

I have a 2D-convolution problem, where i need to create a matrix, that is a block toeplitz matrix.

First of all I have N vectors with N entries, where i create those "blocks" with. (For example in a matrix: T1=toeplitz(k(1,1:N)), T2=toeplitz(k(2,1:N)) .....)

Secondly I'm going to arrange them, as if they're entries in a symmetric toeplitz matrix: T_final=[T1 T2 ... TN-1 TN; T2 T1 ... TN-1; ....; TN TN-1 ... T2 T1]

Is there any efficient and easy way without for-loops, that can create such matrix?

Here is an imagesc, how the matrix could look like for N=6.

Greeting Tiliam

[Tiliam]

It's also important to avoid any "conv" or "for loop" code.

[Sindar]

I think a similar question was asked recently. I can't find it just now, but I believe a solution involves converting between cells and arrays. Algorithm roughly:

• generate matrices to go into blocks
• wrap each matrix in a cell
• tile cells into correct locations
• expand cells and concatenate elements