kaizen123/imagesegmentation-13

image segmetation using kmeans method

ImageSegmentation

Image segmetation using kmeans method

Ideas from https://cn.mathworks.com/help/images/examples/color-based-segmentation-using-k-means-clustering.html

Image in this example from https://ss1.bdstatic.com/70cFuXSh_Q1YnxGkpoWK1HF6hhy/it/u=1357968263,3722616166&fm=27&gp=0.jpg

There's some problems in this example

1. We need to give the number of classes before running test1.m;
2. Initial value of cluster center is very crucial, improper value may influency the final result, so this program is not stable very well, needs to be adjusted later;
3. Validate methods haven't been used in these program.

The innovative point in this program

1. Image segmentation using kmeans has advantages compared to threshold method;
2. We first flatten our image into one dimension before input it into kmeans program;
3. Function reshape in matlab has been used to memorize the location information of data.
4. The order of classify can disturb the final result without adding extral term, details in test1.m

The results of our program

original image

when choosing 3 classes

when choosing 4 classes