How to calculate arithmetic mean of non aligned pixels?

My project is Segmenting retinal vessels. I use DRIVE dataset.

In my project, i have one doubt. Can anyone clarify this.

     [0 0 0 1 0 0 0;
      0 0 0 1 0 0 0;
      0 0 0 1 0 0 0;
      0 0 0 1 0 0 0;
      0 0 0 1 0 0 0;
      0 0 0 1 0 0 0;
      0 0 0 1 0 0 0 ]

This is one of the mask (T_d)(there are 12 masks like this). Green channel of a image is taken for the process.

T_di is centered on each pixel p of G and the following parameters are measured:

    The arithmetic mean, NAdi(p), of the gray-levels of all pixels of G included by T_di and non aligned along di.

Please clarify it. Thank you.

Answer 1

[Walter Roberson]

If you have a two-level mask, M, that selects "aligned" pixels whose mean is to be calculated by the other half of your question, with M being 1 for pixels that are included in the "aligned" and 0 for pixels that are not included as "aligned", then the mean of the "aligned" pixels is

mean_aligned = conv2(YourArray, M./sum(M(:)), 'same' );

Assuming that the non-aligned pixels means everything within reach of the mask but which is not selected as aligned by being 1 in M, then the mean of the "non-aligned" pixels is

nM = double(M~=1);
mean_non_aligned = conv2(YourArray, nM./sum(nM(:)), 'same');

Code that would be more efficient than double(M~=1) would be

nM = 1 - M;

Code that would be more efficient than sum(M(:)) would be nnz(M), and code that would be more efficient than sum(nM) would be nnz(nM)

[Kalai S]

Sir, Thank you so much for your help.