Unsupervised Segmentation of SPOT images

SYNOPSIS

• Initial SPOT image

Extraction of a textural feature adapted for urban areas:

• Estimation of the temperature associated with a 4-connected MRF based on a renormalization approach.
• Result

Unsupervised classification using the CVC algorithm

• Result on the initial SPOT image
• Result on the urban areas parameter image

Bayesian segmentation using a Markovian Prior and a MPM criteria

• Fusion data for the first order potentials
• Prior parameter estimation using an MCMCML algorithm
• Segmentation using a Potts model
  • Result
• Segmentation using a Chien model
  • Result

Concluding remarks