Python Lesson 44: Image Color Spaces (AI pt. 10)

Hello everybody,

Michael here, and today’s post will cover how to understand color spaces in images.

Granted, I’ve previously discussed various colorscales you can find in computer programming in this post-Colors in Programming-but in this post, we’ll take a deeper dive into the use of colors in images.

But first, what is a color space?

Well, as the header above asks, what is a color space? In the context of images, a color space is a way to represent a certain color channel in an image.

Still confused? Let’s take the image we used in our first computer vision lesson (it can be found here Python Lesson 42: Intro To Computer Vision Part One-Reading Images (AI pt. 8)). Assuming we’re analyzing the RGB image of Orange Boy, the color spaces simply represent the intensities (or spaces) of red, blue and green light in the image.

And now let’s analyze colorspaces in OpenCV

As the header says, let’s examine color spaces in Open CV! Here’s the image we’ll be using for this tutorial:

This is a photo of autumn at Bicentennial Capitol Mall State Park in Nashville, TN, taken in October 2022.

Before we start exploring colorspaces, let’s read in this image to our IDE using the RGB colorscale (which means you should remember to convert the image’s default colorscale):

import cv2
import matplotlib.pyplot as plt
park=cv2.imread(r'C:\Users\mof39\Downloads\20221022_101648.jpg', cv2.IMREAD_COLOR)
park=cv2.cvtColor(park, cv2.COLOR_BGR2RGB)
plt.figure(figsize=(18, 18))
plt.imshow(park)

Great! Now that we have our RGB image, let’s explore the different color channels!

First off, let’s examine this image’s red colorspace! How can we do that? Take a look at the code below:

B, G, R = cv2.split(park)
plt.figure(figsize=(18, 18))
plt.imshow(R, cmap='Reds')

plt.show()

In this example, I used the first line of code (the one with B, G, R) to split the image into three distinct colorspaces-blue, green and red.

Aside from the standard plt.figure() functions, I did make a slight modification to the plt.imshow() function. Instead of simply passing in the park image, I passed in the R variable so that we see the image’s red colorspace AND passed in the cmap parameter with a value of Reds to display the red colorspace in, well, red.

Now, how can we show the green and blue colorspaces? We’d use the same logic as we did for the red colorspace, except swap the R in the plt.imshow() function for G and B for the green and blue colorspaces and change the cmap values to Greens and Blues, respectively.

Here’s the image’s blue colorspace:

plt.figure(figsize=(18, 18))
plt.imshow(B, cmap='Blues')
plt.show()

And here’s the image’s green colorspace:

plt.figure(figsize=(18, 18))
plt.imshow(G, cmap='Greens')
plt.show()

As you can see from all three of these color-altered images, the sky, park lawn, and buildings in the background are ceratinly more coloed than the trees, which look bright-white in all three color-altered images.

A little more on colorspace

Now that we’ve examined image colorspaces a bit, let’s see how we can find the most dominant color in an image! Take a look at the code below (which uses the park image):

from colorthief import ColorThief
colorthief =
ColorThief(r'C:\Users\mof39\Downloads\20221022_101648.jpg')
dominantColor = colorthief.get_color(quality=1)
print(dominantColor)

(120, 94, 72)

Granted, you could realistically use a package like numpy to find the most dominant color in an image, but the colortheif module is a much more efficient (and more fun) approach.

  • In case you didn’t know, you’ll need to pip install the colortheif module.

After creating a ColorTheif object (and passing in the image’s filepath on your computer), you’ll then need to use the get_color() method and pass in quality=1 as this method’s parameter. Using the quality=1 parameter will extract the most dominant color in an image.

  • You can certainly use a variable to store the most dominant color like I did here (I used the dominantColor variable) but that’s completely optional.

Once you print the dominant color, you’ll notice you don’t get a color name, but rather a 3-integer tuple that represents the frequency of red, blue and green in the image (the tuple is based off of the RGB colorscale). In this case, our most dominant color is RGB(120, 94, 72). What does that translate to?

In plain English, the most dominant color in this image is a very desaturated dark orange. If you take a look at the original RGB image, it makes sense not only because of the color of the park lawn but also due to all the trees and buildings in the image.

What if you want to know not only the most dominant color in an image, but also its color palette? The colortheif module can help you there too! Here’s how:

palette = colorthief.get_palette(color_count=5)
print(palette)

[(120, 94, 72), (179, 192, 208), (130, 160, 197), (28, 31, 32), (182, 141, 108)]

Just as colortheif did with the most dominant color in an image, all colors are represented as RGB 3-integer tuples. The get_palette() function helps returns the top X colors used in the image-the X is represented by the value of the color_count parameter. In plain English, five colors used in this image include:

  • very desaturated dark orange (the most dominant color)
  • grayish blue
  • slightly desaturated blue
  • very dark almost black blue
  • slightly desaturated orange.

This feature is like imagining a painter’s palette in Python form-pretty neat right! As you can see, our painter’s paletter for the park image has a lot of blues and oranges.

Thanks for reading!

Michael

Leave a Reply