ImageJ PCA

Posted by aszia787 on Apr 11, 2007; 6:40pm
URL: http://imagej.273.s1.nabble.com/ImageJ-PCA-tp3699745.html

Dear all,

I have a question in regards to the ImageJ plugins for PCA. I am trying to extract the principal components of a feature set that comprises of 56 features. I used the ImageJ plugin for calculating the PCA for a stack of images and in the end i get a graph with the eigen values for the respective images. Initially i used 14 stacks to calculate the PCA, of them i got the eigen values maximum for the images i loaded first in the stack. The images that were loaded in the end returned 0 or minimal eigen values. I tried to reverse the process to validate if the PCA calculated is correct by loading the same images which returned 0 eigen values in the previous exercise first, and loaded those images that returned the maximum eigen values in the end. This time i got the opposite result, the images loaded first had the maximum eigen values and the images loaded in the end returned 0 eigen values. I assume the images that are loaded first, their eigen values somehow always return the maximum value. Also there is no projection graph of the PCA. Is the application of PCA in the ImageJ plugins useful to my case in which i want to reduce the feature set?

Thanks in advance.

Regards,

Ali.