How to shrink 64 pixels into 9?

Question

Machine learning

Good evening, I don’t understand how the HOG algorithm builds a 9-dimensional vector Suppose we have an 8 × 8 matrix:

10 20 30 40 50 60 70 80
90 15 25 35 45 55 65 75
85 95 100 110 120 130 140 150
160 170 115 125 135 145 155
175 185 185 180 180 140 120 100
80 60 215 185 165 145 125 105
85 80 70 60 40 40
20 10 85 75 65 55 45 35

, respectively, these 64 Pixel values ​​need to be compressed in 9 -dimensional vector, but I don't know exactly how to do it. Do I need to calculate x and y direction, magnitude and orientation for each pixel? And let's say we have a vector with cells from 0 to 160 with a step of 20: 0, 20, 40, 60, 80, 100, 120, 140, 160 What pixels should be entered in these cells? help me please