K-means & Gaussian Mixture Model Implementation

• KECE471 Computer Vision (Prof. Chang-Su Kim)
• C++ Implementation of 1-dimensional K-means & Gaussian Mixture Model for gray-scale image segmentation

Result

• The results of applying 1D K-means clustering & GMM method to each image with the number of clusters, $k$ = 2, 4, or 8, are as follows.
• Note that each cluster's pixel value is the average gray level.

Original Input Image

KU.raw

Gundam.raw

Golf.raw

1D K-means Clustering

$k = 2$, KU.raw

Image	Log

$k = 4$, KU.raw

Image	Log

$k = 8$, KU.raw

Image	Log

$k = 2$, Gundam.raw

Image	Log

$k = 4$, Gundam.raw

Image	Log

$k = 8$, Gundam.raw

Image	Log

$k = 2$, Golf.raw

Image	Log

$k = 4$, Golf.raw

Image	Log

$k = 8$, Golf.raw

Image	Log

Gaussian Mixture Model

$k = 2$, KU.raw

Image	Log

$k = 4$, KU.raw

Image	Log

$k = 8$, KU.raw

Image	Log

$k = 2$, Gundam.raw

Image	Log

$k = 4$, Gundam.raw

Image	Log

$k = 8$, Gundam.raw

Image	Log

$k = 2$, Golf.raw

Image	Log

$k = 4$, Golf.raw

Image	Log

$k = 8$, Golf.raw

Image	Log