import numpy as np
[docs]
def most_common_rgb(input_array):
"""Determine the most common RGB values based on an image-derived array.
The input
Parameters
----------
input_array : ndarray(dtype=float, ndim=3)
Array containing data from a cropped image
Returns
-------
tuple
Most common RGB values.
Examples
--------
>>> most_common_rgb(img_arr)
(8, 181, 213)
"""
# check that the input range is between 0 and 255
if np.min(input_array) < 0 or np.max(input_array) > 255:
raise ValueError("Input array must have values between 0 and 255")
# check that the input are integers (i.e., whole numbers) with no decimal values
if np.issubdtype(input_array.dtype, np.integer) == False:
raise TypeError("Input array must be of integer type")
# check that the input has the correct dimension i.e., [:, :, 3]
if input_array.shape[2] != 3:
raise ValueError("Not an RGB image array. Check that the third dimension is of size 3")
# input from cv2 returns BGR array - need to flip it to return a RGB array
input_array = np.flip(input_array, axis=2)
# reshape array to create a tuple with the rgb values, i.e., (r, g, b)
width, height = input_array.shape[0], input_array.shape[1]
reshaped_array = input_array.reshape(width * height, 3)
rgb_array = list(map(tuple, reshaped_array))
# count the RGB values of each pixel and return the most frequent colour
rgb_dict = {}
for i in rgb_array:
if i not in rgb_dict:
rgb_dict[i] = 0
rgb_dict[i] += 1
most_frequent_colour = max(rgb_dict, key=rgb_dict.get)
return most_frequent_colour