Optical remote sensing relies on the use of panchromatic, multispectral and hyperspectral images with a high spectral resolution and spatial resolution to distinguish the observed scene.

Hypersperctral remote sensing has two distinct characters that the band number is more than multispectral remote sensing and the width of band is narrow.

In this paper, we propose an automatic multispectral remote sensing image classification technique based on improved probabilistic diffusion.