Split and Merge Image Color Space Channels in OpenCV and NumPy


In this tutorial, we will show you how to split the image into it’s multiple channels by using the OpenCV cv2.split() function and also with Numpy. We will also show you how we can merge the single channel images to generate the multi-channel image with the cv2.merge() function.

Color Spaces and Channels

The colors of images can be organized in many ways and this scheme is known as color space. For example, a popular way of representing the image is in RGB color space, where the image consists of Red, Green, and Blue. To be more specific, every pixel of the image consists of three values (R,G,B) where each of R, G, and B ranges between 0 to 255. These attributes are also referred to as channels.

RGB Color Space Channel

For various image processing tasks, you may be required to split the colored image into its individual channels as shown in the below image. You may find it confusing why the Red, Green, and Blue channels are appearing in grayscale individually. This is because the split image has only a single channel whose pixel value varies between 0 to 255. This actually corresponds to grayscale colorspace hence the three images appear in gray.

On the flip side of the split, you can also merge single-channel images to produce the final multi-channel image.

RGB Color Space

RGB is just one of many colorspaces that are there, some of the others are HSV, BGR, CMYK, etc. The different color spaces have their own significance depending on the type of image processing task.

OpenCV Function – cv2.split()

In OpenCV, you can easily split the image into its individual channels by using the cv2.split() function.


The syntax of the OpenCV split function is as follows –


  • img – It is the image whose channels have to be split.
  • It returns the individual channels of the image.

Split BGR Image in OpenCV

It should be noted that OpenCV reads the colored image as BGR and not in RGB mode. This means the sequence of the channel is Blue, Green, and Red. It is however possible to convert it to RGB colorspace by using the cvtColor() function of OpenCV. But to keep it simple we will use the BGR colorspace only to show you how the cv2.split() function works in OpenCV.

Load Required Libraries

Let us start by loading the python libraries that will be required for our examples.


import cv2
import numpy as np

Read Sample Image

Next, we read and display the sample image that will be used throughout our examples. As discussed earlier, the OpenCV reads the image in BGR mode.

In [1]:

#read image

window_name='Dog Image'
cv2.namedWindow(window_name, cv2.WINDOW_NORMAL)



Split Image with cv2.split

Now we pass the image to the cv2.split() function that returns its three channels captured by blue, green, and red variables respectively.

In [2]:
blue,green,red = cv2.split(img)

Display Blue Channel

Let us see the blue channel of the image with the below code.

In [3]:

window_name='Blue Channel'
cv2.namedWindow(window_name, cv2.WINDOW_NORMAL)


cv2.split - blue channel


Display Green Channel

Next, let us see the green channel below.

In [4]:

window_name='Green Channel'
cv2.namedWindow(window_name, cv2.WINDOW_NORMAL)


cv2.split - green channel


Display Red Channel

Next, let us see the red channel below.

In [5]:

window_name='Red Channel'
cv2.namedWindow(window_name, cv2.WINDOW_NORMAL)


cv2.split - red channel


Split HSV Image in OpenCV

Creating HSV Image

Let us convert the earlier image to HSV color space with the help of cvtColor() function. We will then split this image with the cv2 split function in this example.

In [6]:

hsv_img = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)

window_name='HSV Image'
cv2.namedWindow(window_name, cv2.WINDOW_NORMAL)


HSV Colorspace


Split Image with cv2.split

HSV colorspace has three channels – Hue, Saturation, and value. In the below code, the HSV image is split into its three channels.

In [7]:

h_img,s_img,v_img = cv2.split(hsv_img)


Display Hue Channel

Let us see what the Hue channel looks like.

In [8]:

window_name='Hue Channel'


cv2 split - Hue channel


Display Saturation Channel

Next, let us see the saturation channel of the image.


window_name='Saturation Channel'


cv2.split - Saturation channel

Display Value Channel

Finally, let us see the value channel of the image.

In [10]:

window_name='Value Channel'


cv2 split - Value channel

Split Image in NumPy

Although the split() function of OpenCV makes it really easy to split the image into its different channels, its speed performance may not always be very efficient. In such cases, you can split the image with the help of NumPy operations with better speed.

For example, the image containing three channels B,G, and R can be split with NumPy as follows –

In [11]:

B = img[:, :, 0]
G = img[:, :, 1]
R = img[:, :, 2]

It will produce the same output that we saw with OpenCV split function for the BGR image that we saw in the first example below. Hence not repeating the same again.


OpenCV Function – cv2.merge()

In OpenCV, you can use cv2.merge() function to merge the individual single channel images to produce the final multi-channel image.


The syntax of the OpenCV split function is as follows –

merge(img_channel1,img_channel2,… img_channeln)

  • img_channel – Single channel image.
  • It returns the multi-channel image.

Merge Image Channels in OpenCV to BGR

Let us merge the individual single-channel image that we had split in the first example of the tutorial. It produces back the original BGR image.


merge_bgr_img = cv2.merge([blue,green,red]);

window_name='BGR Image'


cv2 merge - BGR

Merge Image Channels in OpenCV to HSV

Next, let us merge the individual single-channel image that we had split in the second example of this tutorial. It produces back the original HSV image.

In [13]:

merge_hsv_img = cv2.merge([h_img,s_img,v_img]);

window_name='HSV Image'


cv2 merge - HSV



  • MLK

    MLK is a knowledge sharing community platform for machine learning enthusiasts, beginners and experts. Let us create a powerful hub together to Make AI Simple for everyone.

Follow Us

Leave a Reply

Your email address will not be published. Required fields are marked *