I am extracting the features like mean,varia​nce,kurtos​is,etc for images in a database.Here after extracting the features for an single whether normalization has to be done or after extracting the features for all the images normalization has to done

I am extracting the features from the database images.

Answer 1

[Image Analyst]

If you are doing automated classification, it's recommended you normalize your feature numbers. Let's say you have two feature classes that have values around 1 and 3, and another feature class that has values around a million. If you do automated classification, it might think that there are only two classes because the first two are so close together that they look like a single class when comparing it to a class that has values of a million. But if you can normalize them all so that they all have values around 0-1, then it will probably recognize that there are 3 distinct classes.