Color-Based Segmentation Using K-Means Clustering

Image Processing Toolbox™ offers a variety of techniques for image segmentation. When working with
color images, color-based segmentation is a common way to identify objects.
This example shows how to segment colors automatically using the L*a*b* color space and K-means clus-
tering in Statistics and Machine Learning Toolbox™.

Step 1: Reading the Image

Suppose you have a file hestain.png, which is an image of tissue stained with hemotoxylin and eosin
(H&E). This staining method helps pathologists distinguish different tissue types. Get started by reading
this image into MATLAB®.

he = imread('hestain.png');
imshow(he), title('H&E image');
text(size(he,2),size(he,1)+15,...
'Image courtesy of Alan Partin, Johns Hopkins University', ...
'FontSize',7,'HorizontalAlignment','right');
Step 2: Converting the Image from the RGB Color Space to the L*a*b* Color Space
How many colors do you see in the image if you ignore variations in brightness? There are three colors:
white, blue, and pink. Notice how easily you can visually distinguish these colors from one another. The
L*a*b* color space (also known as CIELAB or CIE L*a*b*) enables you to quantify these visual differences.
The L*a*b* color space is derived from the CIE XYZ tristimulus values. The L*a*b* space consists of a lumi-
nosity layer L*, chromaticity layer a* indicating where color falls along the red-green axis, and chromaticity
layer b* indicating where the color falls along the blue-yellow axis. All of the color information is in the a*
and b* layers. You can measure the difference between two colors using the Euclidean distance metric.
Using the functions makecform and applycform, you can convert the image to the L*a*b* color space.

cform = makecform('srgb2lab');
lab _ he = applycform(he,cform);

Step 3: Classifying the Colors in the a*b* Space Using K-Means Clustering
Clustering is a way to separate groups of objects. K-means clustering treats each object as having a location
in space. It finds partitions such that objects within each cluster are as close to each other as possible, and
as far from objects in other clusters as possible. K-means clustering requires that you specify the number of
clusters to be partitioned and a distance metric to quantify how close two objects are to each other.
Since the color information exists in the a*b* space, your objects are pixels with a* and b* values. Use the
kmeans function in Statistics and Machine Learning Toolbox to cluster the objects into three clusters
using the Euclidean distance metric.

ab = double(lab _ he(:,:,2:3));
nrows = size(ab,1);
ncols = size(ab,2);
ab = reshape(ab,nrows*ncols,2);

nColors = 3;
% repeat the clustering 3 times to avoid local minima
[cluster _ idx, cluster _ center] = kmeans(ab,nColors,'distance',...
'sqEuclidean','Replicates',3);

Step 4: Labeling Every Pixel in the Image Using the Results from kmeans
For every object in your input, kmeans returns an index corresponding to a cluster. The
cluster _ center output from kmeans will be used later in the example. You can label every pixel in
the image with its cluster _ index.
pixel _ labels = reshape(cluster _ idx,nrows,ncols);
imshow(pixel _ labels,[]), title('image labeled by cluster index');

Step 5: Creating Images That Segment the H&E Image by Color

Using pixel _ labels, you can separate objects in hestain.png by color, which will result
in three images.

segmented _ images = cell(1,3);

rgb _ label = repmat(pixel _ labels,[1 1 3]);

for k = 1:nColors
color = he;
color(rgb _ label ~= k) = 0;
segmented _ images{k} = color;
end
imshow(segmented _ images{1}), title('objects in cluster 1');

imshow(segmented _ images{2}), title('objects in cluster 2');

imshow(segmented _ images{3}), title('objects in cluster 3');

Step 6: Segmenting the Nuclei into a Separate Image

Notice that there are dark and light blue objects in one of the clusters. You can separate dark blue from
light blue using the L* layer in the L*a*b* color space. The cell nuclei are dark blue.
Recall that the L* layer contains the brightness values of each color. First, find the cluster that contains the
blue objects. You can extract the brightness values of the pixels in this cluster and threshold them with a
global threshold using imbinarize.
You must programmatically determine the index of the cluster containing the blue objects because
kmeans will not return the same cluster _ idx value every time. You can do this using the
cluster _ center value, which contains the mean a* and b* value for each cluster. The blue cluster has
the smallest cluster _ center value (determined experimentally).

mean _ cluster _ value = mean(cluster _ center,2);

[tmp, idx] = sort(mean _ cluster _ value);
blue _ cluster _ num = idx(1);

L = lab _ he(:,:,1);
blue _ idx = find(pixel _ labels == blue _ cluster _ num);
L _ blue = L(blue _ idx);
is _ light _ blue = imbinarize(L _ blue);
 You can use the mask is _ light _ blue to label which pixels belong to the blue nuclei, then display the
blue nuclei in a separate image.

nuclei _ labels = repmat(uint8(0),[nrows ncols]);

nuclei _ labels(blue _ idx(is _ light _ blue==false)) = 1;
nuclei _ labels = repmat(nuclei _ labels,[1 1 3]);
blue _ nuclei = he;
blue _ nuclei(nuclei _ labels ~= 1) = 0;
imshow(blue _ nuclei), title('blue nuclei');

Learn More About Image Processing

• clusterImg (download)
• clusterData (download)
• SegmentTool: An Interactive GUI for Segmenting Images (download)
• Segment Images Interactively and Generate MATLAB Code (download)
• Image Processing Toolbox (product trial)

© 2016 The MathWorks, Inc. MATLAB and Simulink are registered trademarks of The MathWorks, Inc. See mathworks.com/trademarks for a list of additional trademarks.
Other product or brand names may be trademarks or registered trademarks of their respective holders.

mathworks.com

93001v00 05/16