Browse Prior Art Database

SHAPE MATCHING IN PYRAMIDS

IP.com Disclosure Number: IPCOM000128465D
Original Publication Date: 1984-Jan-01
Included in the Prior Art Database: 2005-Sep-16
Document File: 11 page(s) / 47K

Publishing Venue

Software Patent Institute

Related People

Grosky, William I.: AUTHOR [+4]

Abstract

This paper presents a technique for region matching. An image is represented as a function f(X,Y) and a region in an image is approximated by an elliptical paraboloid. The coefficients of the elliptical paraboloid are used in matching regions. It is shown that the coefficients can be easily computed in a pyramid type architecture using proposed pyramid linking algorithms. Our experiments demonstrate that the proposed approach is more robust than moment-based approaches for region matching. Another feature of the approach is that it can be used for region splitting and merging, if required.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 10% of the total text.

Page 1 of 11

THIS DOCUMENT IS AN APPROXIMATE REPRESENTATION OF THE ORIGINAL.

SHAPE MATCHING IN PYRAMIDS

William I. Grosky and Ramesh Jain

CRL-TR-2-84

THE UNIVERSITY OF MICHIGAN COMPUTING RESEARCH LABORATORY

JANUARY 1984

Room 1079, East Engineering Building
Ann Arbor, Michigan 48109
USA
Tel: (313) 763-8000

SHAPE MATCHING IN PYRAMIDS

William I. Grosky
Intelligent Systems Laboratory
Computer Science Department
Wayne State University
Detroit, Michigan 48202 and
Ramesh Jain
Department of Electrical and Computer Engineering University of Michigan
Ann Arbor, Michigan 48109

Abstract

This paper presents a technique for region matching. An image is represented as a function f(X,Y) and a region in an image is approximated by an elliptical paraboloid. The coefficients of the elliptical paraboloid are used in matching regions. It is shown that the coefficients can be easily computed in a pyramid type architecture using proposed pyramid linking algorithms. Our experiments demonstrate that the proposed approach is more robust than moment-based approaches for region matching. Another feature of the approach is that it can be used for region splitting and merging, if required.

[ Chapter ] 1. INTRODUCTION

Matching plays a vital role in many applications of image understanding. At a low level, token matching, commonly called correspondence, is used in stereo and structure from motion; object recognition using structural information uses graph isomorphism at a high level. The tracking of objects and change detection are required in many applications. Cross-correlation and statistical techniques have been used for matching regions in such situations. If the segmentation of images has been already performed, then correlation has been suggested as a particularly powerful technique for matching and displacement. A serious limitation of correlation, however, is that it does not work well in the case of rotation.

University of Michigan Computing Research Laboratory Page 1 Jan 01, 1984

Page 2 of 11

SHAPE MATCHING IN PYRAMIDS

In many applications, object recognition is performed based only on the regions, or borders of regions, representing the mask of the object. Usually, several masks for each object are stored and recognition is achieved utilizing matching of the unknown region with each model. In tracking, also, a region for an object is given, or obtained in a frame, and then unknown regions are matched to it.

Assuming segmentation has been already accomplished, many of the most wide used techniques for the shape description of regions are moment-based1. These methods extract features which are invariant with respect to the object's position, size, and orientation. The initial impetus in this area was the work of Hu2 in which 7 algebraic invariants, calculated from n-th order moments of the image, for 0≤n≤3, are described.

In this paper, we introduce a least-squares based technique which computes various features, and compare it with that of Hu3. We interpret a...