Extra group meeting at the beautiful Sunday afternoon, May 5th, 2013.
Test performance has been given for the filters, the first is the sober filter combined with the canny edge detection. And the result is not as good as we want.
Below is the save-ass sample for the image processing.
http://civil.iisc.ernet.in/~nagesh/rs_docs/ImageProcessing_Matlab.pdf
By using the Gaussian filter and the laplacian, the image has been sufficiently smoothed and the noise level has been reduced appropriately.
So far, the filters part came to the end , and k-mean is what we are supposed to do next.
First of all, determine the K value is the primary task. It was said by the instructor that in order to do the kmeans, we need to determine the centroid by randomly pick up one, and then figure out the K and repeat the process until the equilibrium. K can not be too small, or things will just end up in a mess; k cannot be too big, or things will just end up into separate parts. So figuring out the balanced K is essential to keep the result appropriate.
Here follows an example for using the kmeans:
http://faculty.ksu.edu.sa/alisaad/Documents/segPet%20project.doc
No comments:
Post a Comment