k-means image segmentation

Clustering pixel intensities into k groups by alternating nearest-centroid assignment and centroid recomputation.

K-means image segmentation treats each pixel's intensity (or feature vector) as a point in feature space, then alternates two steps until convergence: assign every point to its nearest of $k$ centroids, then recompute each centroid as the mean of the points now assigned to it. Replacing each pixel with its own cluster's centroid value partitions the image into $k$ regions, chosen by proximity in intensity space rather than a fixed cutoff rule.