Bckground subraction using k means

I have tried background subtraction algorithm using k means with matlab. Step 4 and 5 are difficult to understand. kinldly check and do the needful

1: Idown ←Down-sample the image to 25% of its original size using

simple linear interpolation.

clc;

clear;

close all;

I = imread ('45.jpg');

Idown=imresize(I,[256,256]);

2: Get the S channel of Idown and consider it as an 1-d vector V of

pixel intensities.

I_hsv = rgb2hsv(Idown);

HSV_s = I_hsv(:,:,2);

HSV_s_1 = HSV_s(:);

3: Perform Dbin ←K-Means(V, k = 2).

Dbin = kmeans(HSV_s_1,2);

4: MapM ← Dbin back to image space. For that just do a linear scan

of Dbin.

please help this step ( how to do linear scan?)

5: Mup ←Up-sample the generated binary mapMback to the input

image size.

please help this step

6: Close small holes on Mup using the Closing morphological operator

with a disk structuring element of radius 7 pixels.

se = strel('disk', 7)

closeBW = imclose(Mup,se);

**********************************************************

Please check the above steps and do the needful

Answer 1

[Image Analyst]

Since the color of the objects is nearly the same color as the background, I would probably not use color segmentation or kmeans. I'd probably use stdfilt() to identify "rough" things. The background will be smooth so you can then threshold it away.

[vetri L]

Can you send me sample program for your method?

[Image Analyst]

Attached is a start. try playing with the parameters to see if you can optimize it.

[vetri L]

Thank you sir. I will try with your code