How to determine the number of classes in images segmentation

I'm working on remote sensing image segmentation. I applied several classification methods to our images (ex. Kmeans, FuzzyCmeans, Otsu). These methods faces an important problems : determination of a good number of clusters to be found in the grouping of data. the number of clusters vary irregularly for each image.

 can you help me please !!!

[Sean de Wolski]

The determination of the "right" number of clusters is something you will have to decide based on your application. We will not be able to help with that - especially without seeing an image.