Consider a $6 \times 6$ grayscale image, represented as a $6 \times 6 \times 1$ matrix because there are no separate RGB channels. To detect vertical edges, we construct a $3 \times 3$ filter (or kernel). Cross-correlating the $6 \times 6$ input matrix with the $3 \times 3$ filter produces a $4 \times 4$ output matrix. To compute the upper-left element of this output, we overlay the $3 \times 3$ filter onto the top-left $3 \times 3$ region of the input image and sum the element-wise products. For instance, this calculation might be: $3(1) + 1(1) + 2(1) + 0(0) + 5(0) + 7(0) + 1(-1) + 8(-1) + 2(-1) = -5$. To compute the second element, we shift the filter one step to the right on the $6 \times 6$ image, perform the same element-wise products and additions, yielding $-4$, and so on. To compute the next rows, we shift the filter one step down and repeat the procedure.