Date of Submission


Date of Award


Institute Name (Publisher)

Indian Statistical Institute

Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy

Subject Name

Computer Science


Electronics and Communication Sciences Unit (ECSU-Kolkata)


Majumdar, Dwijesh Dutta (ECSU-Kolkata; ISI)

Abstract (Summary of the Work)

Analysis and recognition of 2-dimensional shapes constitute an important problem in the fields of Pattern Recognition, Image Analysis, Computer Vision, Robotics and other related areas of research. If we look at the problem from the point of view of image or scene analysis, in a sense, the antire subject of automatic scene analysis might be defined as the problem of describing and recognizing the shape of the ob jects in an image. Shape is the primal intrinsic property for the vision system because we associate the definitions of objects with shape, rather than with colour or reflectivity, for example.According to Webster, the shape of an object is that quality of the object which depends on the relative position of all points composing its outline. This definition emphasizes the fact that uje are aware of shapes through outlines which may be visually perceived.In this thesis, the shape of an object is defined as that property of the object which is invariant under translation, dilation and rotation. Shape recognition is undoubtedly one of the most important facilities of our visual system. We know how important shape information can be extracted from images in early processing and segmentation. One of the ma jor challenges to computer vision is to, represent shapes, or the important espects of shapes, so that thay may be learned, matched against, recollected and used. This thesis deals with description and classification of the shapes of parts of a two-dimensional picture. For thia purpose, ue assume that parts of the picture have been isolated as meaningful entities. In other words, our assumption is that söme subsets of the picture plane, called the objects, have already been defined as a result of previous processing. These objects ars nothing but bilevel pictures in the plane. In this thesis, tuwo types of bilevel pictures are considered (line structures or curvea and reoions),Now any scheme to describe the shape of a bilevel picture (that is, of an object) should ideally satiafy the three following objectives : (1) It should preserve the information of interest, that is, it should not lose any relevant information about the some object while doleting the irrelovant ones (whether informatio is relevant .or not depends an the particular context), (2It should permit compact, storage and be convenient for display, (3) It should facilitate any required future processing. It is clear that the above three conditions are conflicting with one another in the sense that compact storage may lead to loss of some relevant information or preservation of all relevant information - may not be always good for subsequent proce- seing. So, a compromise has to be made to get an optimal strategy depending on speaific problems. It has been mentioned above that one desirable property o a description of an object is preservation of relevant informatic Since in practice it is not possible in general to preserve all the relevant information:, attempts are made to preserve as much of ft as possiblo. We will naw formalize the notion of what a more informative description o2 means. The more informative the description, the fewer number of objects there are that satisfy that description.


ProQuest Collection ID:

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.


Included in

Mathematics Commons