Hello folks,
I am using the SID method to classify a hyperspectral image along with a spectral library. I have noticed that the IDL code that I have implemented myself (based on Du et al. 2004) does not give the same results than the ENVI tool (Classification/Supervised/Spectral Information Divergence). I am quite surprised because the process is quite simple: for each spectrum in the image, (1) I calculate the SID distance for each class in the library and (2) I assign the class with the smallest SID. I am starting to think that ENVI does some kind of pre- or post- processing...
Has anybody experienced something similar?
Thanks!
|