Convert an image to grayscale using only matrix (No BGR2GRAY function) in Python

Question

I'm trying to convert an image to grayscale without using OpenCV or Numpy Example : The matrix of my image is

    [[[116 116 117]
  [115 115 116]
  [117 115 115]
  ..., 
  [135 138 142]
  [137 139 139]
  [137 139 139]]

 [[116 116 116]
  [116 116 116]
  [114 116 116]
  ..., 
  [135 139 140]
  [135 137 138]
  [135 137 138]]

 [[115 118 114]
  [115 118 114]
  [112 116 115]
  ..., 
  [132 141 141]
  [134 137 141]
  [133 136 140]]

 ..., 
 [[ 35  44 163]
  [ 31  40 159]
  [ 10  33 158]
  ..., 
  [ 14  48  53]
  [ 24  56  55]
  [ 27  59  58]]

 [[ 24  38 156]
  [ 19  33 151]
  [  7  28 145]
  ..., 
  [ 25  55  66]
  [ 15  59  61]
  [ 17  61  63]]

 [[  0  27 131]
  [  0  26 130]
  [  0  34 113]
  ..., 
  [ 11  39  55]
  [  6  28  56]
  [  4  26  54]]]

Using the cv2.COLOR_BGR2GRAY function the matrix is :

[[150 150 150 ..., 150 150 150]
[150 173 175 ...,  97  91  89]
[150 176 179 ...,  95  89  82]
..., 
[150  66  67 ..., 152 154 152]
[150  62  59 ..., 152 152 152]
[150  62  64 ..., 155 154 151]]

First, i don't get why it returns a 36 integers when i gave 108 and how can i do it manually

Thanks for your help :)

Answer 1

Question 1

You are getting 36 integers instead of 108 because your original image had 3 channels (red, green, blue, a.k.a RGB). When you convert it to grayscale, you are flattening it to only one channel. 108 / 3 = 36, seems legit!

Quesion 2

Converting it by hand: There are multiple ways to convert an RGB image to grayscale, but the most straightforward would be to take the average of the three channels, basically (red_values + green_values + blue values) / 3. So take your original image array, and run:

gray_image = (image[:,:,0] + image[:,:,1] + image[:,:,2]) / 3

The above is assuming that your image array is in the form image.shape = (pixel_height, pixel_width, number_of_channels), and that the first 3 channels represent your R,G, and B channels but that's pretty standard