2D Down-sampling matrix

Goodevening.

I want to compute a 2D down-sampling matrix (the matrix with which a matrix will be multiplied in order to be downsampled)

    d=downsample(eye(initial_size),downsampling_factor);
    downs=d(1:downsampling_factor:end,:);

But eye will not work with big number. For example if I want to compute the 170 x 340.000 down-sampling matrix with which a signal of 340.000 x 6 will be multiplied to provide the downsampled 170 x 6 signal.

I have found this alternative but I think is too "naive" with the use of for

    d=zeros(1,initial_size);
    d(1)=1;
    for i=1:sampling_factor
        downs(i,:)=circshift(d,[1,sampling_factor*(i-1)]);
    end

Thank you in advance.

Answer 1

[Jongwoo Hong]

Hi, I recommend the simple method using transpose.

If you want to downsample M x N matrix A into M/10 x N/5 matrix B (not exact number),

B = transpose(downsample(transpose(downsample(A,10)),5))

will be helpful.