K-means on a gray scale image

Hello everyone, I have a grayscale image 'lena.bmp' and I am trying to apply k-means color quantization. I found the following code but I have some points that are not clear in my mind: 1)Why we need to convert the image format to double? 2)While I am writing "pixel_labels = reshape(cluster_idx,nrows,ncols);" in the code what excatly it is doing? 3)Why did I need to write the square brackets inside this "imshow(pixel_labels,[])" ? The code is here:

Image=imread('lena.bmp');
Image = im2double(Image);
nrows = size(Image,1);
ncols = size(Image,2);
Image = reshape(Image,nrows*ncols,1); 
Colors = 4;
[cluster_idx, cluster_center] = kmeans(Image,Colors,'distance','sqEuclidean');  
pixel_labels = reshape(cluster_idx,nrows,ncols);
imshow(pixel_labels,[]);

Thank you so much.

Answer 1

[Adam]

1. kmeans only supports double or single inputs. It works in floating point so integers wouldn't make sense.

2.

doc reshape

will answer this. It does pretty much exactly what it says - it reshapes your output back to the number of rows and columns of your image so that you have an image back rather than just a vector of class indices.

3.

doc imshow

describes the different scaling you can do with imshow. This syntax is the 3rd in the list.

Answer 2

[Image Analyst]

See my attached demos for kmeans in both color and grayscale. Write back if you still have questions.

By the way, you can't do "color quantization" on a grayscale image, unless you're considering the gray levels as "colors".

[Iremsu Savas]

Thank you for your attached demos. I meant 8 bit grayscale images, is it a wrong mention?

[Image Analyst]

Not exactly sure what that means. But neither you or I mentioned anything "wrong".

[pooja s]

THANK U SO MUCH SIR . THIS HELPED ME A LOT

[Mnahil Khursheed]

How to save the result of segmentation is a folder for feature extraction further. Any one please help me.

[Image Analyst]

Call imwrite() or save().