Importance of DA-MRF Models in Fuzzy Based Classifier

Mrinal Singha, Anil Kumar*, Alfred Stein, P.L.N. Raju, Y.V.N. Krishna Murty

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

3 Citations (Scopus)
18 Downloads (Pure)


This research considers the smoothness prior and four discontinuity adaptive Markov Random Field (DA-MRF) models to deal with discontinuity adaptation for the contextual fuzzy c-means (FCM) classifier. They were applied to classify AWiFS and LISS-III images from the Resourcesat-1 and Resourcesat-2 satellites. A fraction image from the high resolution LISS-IV image has been used as reference data. Quality of the classified AWiFS and LISS-III images was assessed by means of an image to image fuzzy error matrix (FERM). The classification accuracy increased by 1.5 to 6 % as compared to the conventional FCM. Classification accuracy increased with 0.5 to 8 % when comparing Resourcesat-2 with Resourcesat-1 data. The study showed that DA3-MRF model with FCM performed better than other MRF models, showing an improved overall classification accuracy as well as preserving the edges at boundaries.
Original languageEnglish
Pages (from-to)27-35
Number of pages9
JournalPhotonirvachak = Journal of the Indian society of remote sensing
Issue number1
Early online date11 Jul 2014
Publication statusPublished - Mar 2015


  • 2024 OA procedure
  • Discontinuity adaptive (DA)
  • Fuzzy c-means (FCM)
  • Markov random field (MRF) models


Dive into the research topics of 'Importance of DA-MRF Models in Fuzzy Based Classifier'. Together they form a unique fingerprint.

Cite this