PROPOSED MODEL TO MEASURE THE EFFECT OF DISCONTINUITY ADAPTIVE MRF MODELS IN FUZZY BASED CLASIIFIER ON SATELLITE IMAGES

Abstract

Presently wide ranges of remotely sensed data are

available from earth observation satellites. This data are

analyzed to prepare land use/ land cover maps using

different remote sensing techniques. Image classification is

one way to produce these land use/ land cover maps. Due

to continuous nature of real world phenomena, the image

classification to map land cover classes is a challenge.

Presence of mixed pixels decreases the efficiency of image

classification. Fuzzy classification technique such as Fuzzy

c-Means (FCM) can be used to handle mixed pixels.

Although FCM has the advantage of classifying mixed

pixels by assigning membership value, it does not

incorporate spatial contextual information of the pixels

into its classifying algorithm. Use of context eliminates the

problem of isolated pixels and improves the classification

accuracy. In this research work a contextual FCM

classifier has to be developed by using MRF models.

Smoothness prior and four discontinuity adaptive prior

have been used to incorporate contextual information with

FCM. The developed discontinuity adaptive contextual

FCM classifier would be tested both on coarse and fine

resolution dataset i.e. AWFIS and LISS-III with spatial

resolution 60 m and 20m respectively. It is expected that

the discontinuity adaptive prior models, improves the

overall classification accuracy by preserving the edges at

boundaries and the classified output is consistent with

spectrally and spatially.