mstGCluster: mstGCluster: A Graph Clustering Algorithm Based on Minimum...
In XiaoyuLiang/i2d: Convert image to digital data

Description Usage Arguments Author(s) References Examples

The minimum spanning tree-based graph clustering algorithm (mstGCluster) not only can be used to group the vertices into a specific number of clusters but also can filter the noise in the graph.

1	mstGCluster(dat=dat, k=2, filter=FALSE, filter.n=10, filter.d=0)

`dat`	data.frame or matrix of x-coordinate and y-coordinate.
`k`	integer: number of clusters. Default is `k=2`. If `k=0`, the function will return the path of minimum spanning tree.
`filter`	logical: if `FALSE` (the default), no vertices will be removed, otherwise the outliers will be filtered.
`filter.n`	integer: minimum number of vertices of each cluster. If `filter=TRUE`, it will be used as cluster size cutoff. This argument specifies filter the clusters that contain number of vertices less than specific number `filter.n`.
`filter.d`	a positive value: this argument specifies the minimum distance of vertices that not be filtered. If `filter=TRUE`, it will be used as weight/distance cutoff. If `filter.d=0` (the default), the cutoff will be calculated by k-means clustering algorithm.

Ying Hu yhu@mail.nih.gov

Chunhua Yan yanch@mail.nih.gov

Xiaoyu Liang xiaoyu.liang@yale.edu

Csardi, G., & Nepusz, T. (2006). The igraph software package for complex network research. InterJournal, complex systems, 1695(5), 1-9.

data(Noise_Filtering,package="i2d")
dat1 <- i2d(image=Noise_Filtering, p.n=500);
out1 <- mstGCluster(dat=dat1, k=2, filter=TRUE, filter.n=10);
col1 <- rainbow(10, alpha=0.5)[c(1,2,7,9)];

## The original image
EBImage::display(Noise_Filtering)
## Plot the 500 randomly selected vertices
plot(dat1, col="black", pch=19, main="Raw Data");
## Outliers detection
plot(out1$outlier, col="black", pch=19, main="Outliers Filtered Before Clustering");
## Plot the clusters after flitering noise
plot(out1$dat, col=col1[out1$clu$membership], pch=19, main="Cluster After Filtering")