Date of Submission

9-28-2008

Date of Award

9-28-2009

Institute Name (Publisher)

Indian Statistical Institute

Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy

Subject Name

Computer Science

Department

Advance Computing and Microelectronics Unit (ACMU-Kolkata)

Supervisor

Bhattacharaya, Bhargab Bikram (ACMU-Kolkata; ISI)

Abstract (Summary of the Work)

Several problems of characterizing a digital object, and particularly, those related to boundary description, have been studied in this thesis. New algorithms and their applications to various aspects of image analysis and retrieval have been reported. A combinatorial technique for constructing the outer and inner isothetic covers of a digital object has been developed. The resolution of the background 2D grid can be changed by varying the grid spacing, and this procedure can be used to extract shape and topological information about the object. Next, an algorithm has been designed for constructing the orthogonal (convex) hull of a digital object against a background grid by employing a combinatorial technique. On the application side, it has been shown how the shape code derived from the isothetic cover with varying resolutions can be used in image analysis and retrieval. The nature of the isothetic cover also provides a measure of the shape complexity of the underlying object. Several properties of isothetic cover are shown to be useful in defining features of handwritten characters. An algorithm for polygonal approximation of a thick digital curve based on its outer and inner isothetic covers has also been developed.To describe the contour of a digital object by an unordered set of points of optimal or sub-optimal size, a new concept called the pointillist ensemble, is introduced in this thesis. This is derived from the end points of digital straight pieces describing the curve satisfying certain neighborhood properties. This is based on the idea of pointillism, a style of painting with dots, developed by the Neo-Impressionist painters of France. The procedure of reconstructing the object from the ensemble is also presented. Finally, other topological properties of an object, namely, those based on the Euler number have been used to devise an algorithm for efficient image indexing. This has been formulated deploying a novel concept of randomized spatial masking of the digital object.The algorithms, the proofs of their correctness, and the relevant experimental results on diverse databases have been reported to substantiate the theoretical findings presented in this thesis.

Comments

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:28843765

Control Number

ISILib-TH365

Creative Commons License

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

DOI

http://dspace.isical.ac.in:8080/jspui/handle/10263/2146

Download

Included in

Mathematics Commons

Share

COinS