k-Means Classifier
Features:
Performs classification of 3D datasets using k-Means clustering-based volume segmentation algorithms. Two types of k-Means classifiers are available:
Slice-based clustering k-Means classifier (frame-to-frame classifier)
Global clustering k-Means classifier (entire datacube).
Slice-based frame-to-frame classifier
1. Open the dataset in the IDCube format.
2. Click the Machine Learning tab on the menu bar, select k-Means Classifiers, and select k-Means Slice-Based Clustering Classifier.
3. The new dialogue window k-Means parameters will open. Enter the number of expected clusters (classes). The default number is 2. Enter the number of iterations (usually <100).
4. Each color in the updated image corresponds to a different class (1 out of 10 in this example). Pixels with the color corresponding to label 1 on the colorbar belong to the first cluster, label 2 belong to the second cluster, and so on for each of the k clusters.
5. Use Band/Channel sliders and other functions to further optimize the image.
Note: The number of the labels on the colorbar is identified automatically and could be equal or less than the number of input number of clusters.
k-Means Global Clustering Classifier
1. Open the dataset in the IDCube format.
2. Click the Machine Learning tab on the menu bar and select k-Means Classifiers then k-Means Global Clustering Classifier.
3. The new dialogue window k-Means parameters will open. Enter the number of expected clusters (classes). The default number is 2. Enter the number of iterations (usually <100).
4. Each color in the updated image corresponds to a different class. Pixels with the color corresponding to label 1 on the colorbar belong to the first cluster, label 2 belongs to the second cluster, and so on for each of the k clusters.
Since this algorithm is global across all bands/channels, the changes in the Band/Channel slider positions will not affect the image.
Note: The number of the labels on the colorbar is identified automatically and could be equal or less than the number of input number of clusters.
Additional Information:
k-Means clustering is an unsupervised machine learning algorithm. It is the fastest and most efficient algorithm to categorize data points into groups even when very little information is available about data. IDCube employs k-Means++ algorithm that chooses the initial values (or "seeds") for the k-Means clustering algorithm. The procedure initializes the cluster centers before proceeding with the standard k-Means optimization iterations.
References:
Arthur, David, and Sergei Vassilvitskii. “K-Means++: The Advantages of Careful Seeding.” In Proceedings of the Eighteenth Annual ACM-SIAM Symposium on Discrete Algorithms, 1027–35. SODA ’07. USA: Society for Industrial and Applied Mathematics, 2007.
