This project performs colour quantization with a k-means clustering algorithm.
Run this code in the commandline by calling
$ python main.py IMAGE-PATH K-VALUE
It will display the results of the algorithm in a popup window using the Pillow image processing library.
Colour quantization reduces the amount of colours in an image while maintaining its detail. It can be thought of as repainting a picture with a smaller palette of colours.
K-Means Clustering is a process that finds the natural clusters inside data. In this case, we're using data in the form of RGB pixels in an image, so the algorithm finds a palette of K-VALUE distinct colours that roughly approximate the image.
Ordinarily K-Means doesn't go further than producing the clusters. However this project then performs colour quantization by recolouring each pixel to the nearest colour on the palette generated. There are some very effective domain-specific algorithms for colour quantization of images, such as scolorq.