Feature Sensitive Level Set For Clustering.

Date of Submission

December 2004

Date of Award

Winter 12-12-2005

Institute Name (Publisher)

Indian Statistical Institute

Document Type

Master's Dissertation

Degree Name

Master of Technology

Subject Name

Computer Science


Electronics and Communication Sciences Unit (ECSU-Kolkata)


Mukherjee, Dipti Prasad (ECSU-Kolkata; ISI)

Abstract (Summary of the Work)

Level set analysis is an important tool for curve evolution. Assume a closed curve is embedded inside a 2D matrix. Level set is defined as a distance function of the closed curve where minimum distance of every matrix point from the curve is evaluated. These distances have positive values inside the curve and negative values outside the curve. Consequently, the embedded curve has zero distance (or level set function) value. The evolution of curve inside the matrix using level set analysis is iterative redefinition of the curve distance function based on certain PDE based energy minimization process. Therefore, the evolution of curve is equivalent to evolution of level set function.In this thesis, we have extended level set analysis for clustering regions in feature space. The energy function for clustering is based on classical definition of clustering in terms of minimizing intra-cluster distances and maximizing inter-cluster distances. Multiple cluster in the feature space are detected using multiple level set functions.In each iteration, the c are evolved using forces that engulf the nearby points. Again the curves are mutually 1epelled through maximization of distances between corresponding cluster centers. We have assumed that the number of clusters are known a priori. However, we have also proposed a heuristir introduce a new curve and corresponding level set function in case a set of pointsve a tendency to form a separate cluster. The efficacy of the clustering technique is demotrated through its clustering performance on both synthetic and real images. can


ProQuest Collection ID: http://gateway.proquest.com/openurl?url_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:dissertation&res_dat=xri:pqm&rft_dat=xri:pqdiss:28843122

Control Number


Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.



This document is currently not available here.