Please use this identifier to cite or link to this item:
Title: Best-fit Ellipse Concept from the Second Order Moments for Object Retrieval
Authors: Lau, KL
Siu, WC 
Law, NF 
Keywords: Moment Invariants
Object Searching
Pattern Recognition
Shape Analysis
Issue Date: 2002
Source: Proceedings of the First International Conference on Information Technology and Applications (ICITA 2002), 2002, p. 397-402 How to cite?
Abstract: Hu moment invariants are a popular tool for shape description in areas such as object identification and character recognition. The original formulation consists of a set of seven moment invariants for characterizing object features, which are independent of translation, scale, and rotation. In order to provide a meaningful description of object shape, it is important to understand the meaning of these moments. However, there are few pertinent studies. This paper firstly provides an interpretation of the second order-based moment invariants using the best-fit ellipse formulation. Then a new formulation is suggested for image retrieval. Our theoretical analysis shows that the first invariant measures the total spread of the shape relative to its area square while the second invariant measures the degree of elongation of a best-fit ellipse on the shape. Six different shapes and objects with Gaussian noise were used to confirm the findings. The ratio of major and minor axes from the best-fit ellipse were applied for describing the global information of the shape. Results of the experimental work show that this is an efficient framework for object searching and the computational complexity can be significantly reduced.
Description: Proceedings of the First International Conference on Information Technology and Applications (ICITA 2002), Bathurst, 25-28 November 2002
ISBN: 1864671149
Appears in Collections:Conference Paper

View full-text via PolyU eLinks SFX Query
Show full item record

Page view(s)

Last Week
Last month
Citations as of Jan 14, 2019

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.