CODEWITHSUNDEEP
rk-meansimage-segmentationindices
anna
anna

Reputation: 51

How to separate clusters after Kmean segmentation

I have segmented an RGB image into 3 clusters. I want to recreate images containing indices from each cluster separately- using R. How can I do this?

Upvotes: 0

Views: 344

Answers (1)

jmcastagnetto
jmcastagnetto

Reputation: 486

You are not including a minimal reproducible example or a sample dataset, so I will assume that you have done something along the lines of the article: https://alstatr.blogspot.com/2014/09/r-k-means-clustering-on-image.html

One change to the code, in the aforementioned article, that might help you with associating the cluster number to the each pixel, will be the following:

library(dplyr)

kClusters <- 3
# pick all distinct colors we will be using for clustering
unique_colors <- imgRGB %>% select(R, G, B) %>% distinct() 
kMeans <- kmeans(unique_colors, centers = kClusters)

df <- imgRGB %>% 
  left_join(
    # make the data frame to join with the original
    bind_cols(unique_colors,
              data.frame(
                cluster = kMeans$cluster
              )),
    by = c("R", "G", "B")
  )

str(df)

# 'data.frame': 420800 obs. of  6 variables:
# $ x      : int  1 1 1 1 1 1 1 1 1 1 ...
# $ y      : int  526 525 524 523 522 521 520 519 518 517 ...
# $ R      : num  0.00392 0.00392 0.00392 0.00392 0.00392 ...
# $ G      : num  0.00392 0.00392 0.00392 0.00392 0.00392 ...
# $ B      : num  0.00392 0.00392 0.00392 0.00392 0.00392 ...
# $ cluster: int  2 2 2 2 2 2 2 2 2 2 ...

Then you could use the cluster values to facet a ggplot:

ggplot(data = df, aes(x = x, y = y)) + 
  geom_point(colour = rgb(df[c("R", "G", "B")])) +
  labs(title = "Separated clusters") +
  xlab("x") +
  ylab("y") +
  coord_fixed() +
  facet_wrap(~cluster, nrow = 2)

which produces something like:

Separated clusters in image

Hope this gives you a lead on how to solve your problem.

Upvotes: 3

Related Questions