Calculating the average pixel intensity for neighboring pixels only for specific points of an image and storing inside a n-dimensional array

Question

Im working on a face recognition task where i would like to extract face features but just from specific key locations in the face image. But for such task, i would need to calculate the average pixel value in the neighboring region of that specific part. I am doing so by hand, due to having no algorithm. This is an exhaustive process. This is how i am doing it:

img = plt.imread(path)
img[25, 40] = 0
img[25, 41] = 0
img[25, 39] = 0
img[26, 40] = 0
img[26, 39] = 0
img[26, 41] = 0
img[24, 39] = 0
img[24, 40] = 0
img[24, 41] = 0
img[25, 110] = 0
img[25, 111] = 0
img[25, 109] = 0
img[24, 109] = 0
img[24, 110] = 0
img[24, 111] = 0
img[26, 109] = 0
img[26, 110] = 0
img[26, 111] = 0
img[25, 170] = 0
img[25, 171] = 0
img[25, 169] = 0
img[24, 170] = 0
img[24, 171] = 0
img[24, 169] = 0
img[26, 170] = 0
img[26, 169] = 0
img[26, 171] = 0
img[40, 40] = 0
img[40, 41] = 0
img[40, 39] = 0
img[41, 40] = 0
img[41, 41] = 0
img[41, 39] = 0
img[39, 40] = 0
img[39, 39] = 0
img[39, 41] = 0
img[50, 110] = 0
img[50, 111] = 0
img[50, 109] = 0
img[51, 110] = 0
img[51, 111] = 0
img[51, 109] = 0
img[49, 110] = 0
img[40, 170] = 0
img[40, 171] = 0
img[40, 169] = 0
img[39, 170] = 0
img[39, 171] = 0
img[39, 169] = 0
img[41, 170] = 0
img[41, 171] = 0
img[41, 169] = 0

plt.imshow(img)

What i would like to do basically is have a better way to calculate the average value of a neighboring 3x3 pixels around the center for 20 coordinates in a image and store inside a n-dimensional vector. Or for more clarification: for a choosen keypoint 1 calculate the average value in a 3x3 neighborhood, store this value. keypoint 2 calculate the average value in a 3x3 neighborhood, store the value. For any given keypoint that corresponds to a coordinate x and y in the image.

Each keypoint is an average value of a grid that has 3x3 pixels to which i need to grab the average value and store inside a 20d array. Whats the best way to do it?

Answer 1

In Python/OpenCv, just use cv2.blur() to do a 3x3 block average of the whole image. Then just get the pixel values in the blurred image at the points you desire.