Converting RGB to BGR for image data

image_channel_order_cvt.py

import numpy as np

"""
In this gist, I demonstrate the simple ways to
convert between 2 colorspace representation of images 
mutually(RGB, BGR).
"""
CONVERSION_ORDER = [2, 1, 0]

def convert_channel(img):
  """
  Convert an image in the first convention to 
  the second convention.
  We just need to swap the order of the 1st and 3rd channel.
  - An image is represented by [height, width, channels]
    For a color image, [height, width, 3]
    The last dimension is reserved for color channels (R,G,B) or 
    (B,G,R) in cv2 convention. 
  """
  # Reverse the order in the last dimension
  return img[...,::-1]
  # or 
  return img[...,CONVERSION_ORDER]
  # or
  return np.stack([img[..., CONVERSION_ORDER[0], 
                   img[..., CONVERSION_ORDER[1], 
                   img[..., CONVERSION_ORDER[2]], 
                  axis=-1)