Entropy values of all the bands in a given hyperspectral image.

I have a dataset of hyperspectral image (Pavia University). I was trying to calculate the entropy values of all the respective bands in the image. The number of bands the image has is 103. The entropy values I am getting are all zeros. I want to know whether these values are correct or not. If not, then how to calculate?

Answer 1

[Bjorn Gustavsson]

It seems it is not the most robust functionality, If I run the test of entropy from the help of the entropy-function I get rather peculiar results:

I = imread('circuit.tif'); 
E = entropy(I);         
E2 = entropy(histeq(I));
E3 = entropy(double(I));
E4 = entropy(histeq(double(I)));
E5 = entropy(uint8(double(I))); 
disp([E E2 E3 E4 E5])                   
       6.9439       5.9354            0            0       6.9439

So maybe you're having a data-type problem - that you might dodge by scaling-n-typecasting to uint8.

("Entropy" - in my humble opinion have gotten far to much mileage out of its name, it's achievements are undeniable in statistical physics, but as a measure of anything sensible in imaging are often unjustified or overblown. Something like the "nobility of statistics" - shown too much reverence)

HTH

[Ramya Gutta]

Thank you so much for your answer.