K-nearest neighbors in segmentation of gray image

Hi ,

I need help to implement k-nearest neighbors algorithm to segment a gray image.

I dispose of a Training image, and the original image to be segmented. But I don't have the knn algorithm to use it. KnnSearch corresponds to what I am looking to implement but I have Matlab r2009b thus it is not implemented.

please do you have the knn matlab code for 2 dimensionnal images ?

Answer 1

[Image Analyst]

Do you mean like a knn filter for denoising where you replace a pixel by the mean of the brightest or darkest N values? If so, you can do that all with a few calls imfilter() or conv2() if you're clever about it.

[Makrim]

no, that's not what I mean. I would like to use knn as supervised classifier, to classify pixels of an image based on training data

[Image Analyst]

What do the axes of your classes represent?

[Makrim]

it is okey, I implemented myself : http://www.mathworks.fr/matlabcentral/fileexchange/46117-k-nearest-neighbors-for-image-segmentation

[Mohammed Abdul Wadood]

in KNN filter is the size window is 3 by 3 or I can determine the size from k ?

do you have the code for KNN filter please ?

[Image Analyst]

Who are you asking? @Makrim says he uploaded his code to the File Exchange.

In MATLAB, knnsearch is a function in the Statistics and Machine Learning Toolbox.

[Mohammed Abdul Wadood]

I asking you @Image Analyst, Sorry, I didn't notice the term (knnsearch).

I working on denoising satellite image with gaussian noise. So, I asket about KNN filter.

thank you

[Image Analyst]

Attached is my only demo of knn.