Image Compression via K-means Clustering (Color Quantization)

Overview

This uses the Lloyd's K-means Clustering algorithm with RGB tuples in order to find the k colors that best represent the input image. The K-means Clustering algorithm was implemented by me for a better understanding of how the algorithm worked. In addition, this program also produces scatter plots of the colors to better visualize the clusters.

Running the Program

NOTE: Assuming you are using Python 3.x

python image_k_means.py <image_name> [-k <default: 10>] [--iterations <default: 20>] [--save-scatter]

• -k is the number of clusters you want to find
• --iterations is the number of iterations you want to run Lloyd's Algorithm
• --save-scatter is a flag to choose to save a Scatter Plot of the data (WARNING: Very slow)

Results

Input Image	5-means Clustering (with 20 iterations)
10-means Clustering (with 20 iterations	20-means Clustering (with 20 iterations)

Input Scatter Plot by Color	20-means Clustering Scatter Plot by Color