Robust detection of significant points in multiframe images

doi:10.1016/S0167-8655(98)00135-4

Pattern Recognition Letters

Volume 20, Issue 2, February 1999, Pages 199-206

https://doi.org/10.1016/S0167-8655(98)00135-4 Get rights and content

Abstract

Significant point (SP) detection is an important pre-processing step in image registration, data fusion, object recognition and in many other tasks. This paper deals with multiframe SP detection, i.e. detection in two or more images of the same scene which are supposed to be blurred, noisy, rotated and shifted with respect to each other. We present a new method invariant under rotation that can handle differently blurred images. Thanks to this, the point sets extracted from different frames have relatively high number of common elements. This property is highly desirable for further multiframe processing. The performance of the method is demonstrated experimentally on satellite images.

Introduction

Detection of significant points (SP) or landmarks is an important step in image processing and computer vision. It provides input information for further operations, such as image registration, image fusion, time-sequence analysis and object recognition. By significant points we understand the points that are easy to identify in the image, such as corners, line intersections, T-junctions, etc.

In this paper, we address a multiframe version of this problem: having two or more images of the same scene, the aim is to detect significant points in each of them. Multiframe SP detection methods must fulfill the condition of repeatability. This property means that the results should not be affected by imaging geometry, radiometric conditions and by additive noise and that the sets of points detected in all frames should be identical. Since the last requirement is not realistic in practice, “maximum overlap” is usually required instead of identity.

In this paper we assume that the individual frames may be rotated and shifted with respect one another, they may have different contrast, they may be degraded by a linear shift-invariant blur and corrupted by additive random noise. Our primary motivation comes from the area of remote sensing, where the registration of images with such kinds of distortions is a very frequent task. Having the SP detection method which works on differently distorted frames and which yields high repetition rate is a fundamental requirement.

Section snippets

Present state-of-the-art

Numerous methods for single-frame significant point detection in gray-level images have been published in last two decades. Most of them are known as corner detectors. A survey of basic methods along with a comparison of their localization properties can be found in (Rohr, 1994).

Kitchen and Rosenfeld (1982) proposed a corner detection scheme based on a differential operator that consists of first and second order partial derivatives of the image f(x,y): $K(x,y)= f_{x}^{2} f_{yy} −2f_{x} f_{y} f_{xy} +f_{y}^{2} f_{xx} f_{x}^{2} +f_{y}^{2},$ where

Description of the proposed method

Our newly proposed method for the detection of significant points uses a parameter approach to handle differently distorted images. Points, which belong to two edges with an angle from the interval [π/2−d_a,π/2+d_a] (d_a is user defined parameter) in between regardless of its orientation are understood here as significant points. The described method is based on this definition.

Information about the number of edges passing through each pixel and about the angle between them is acquired from the

Numerical experiments

In this section, practical capabilities of the proposed SP detection method are demonstrated and a comparison with the classical techniques (Kitchen and Rosenfeld, 1982; Förstner, 1986) is shown. Since the intended major application area is the area of remote sensing, the experiments are performed on satellite images.

A subscene covering the landscape near Prague (Czech capital city) of the size 180×180 pixels was extracted from the SPOT image of the central part of the Czech Republic. This

Conclusion

In this paper we proposed a novel method for detection of significant points – corners with high local contrast. The method works in two stages: all possible candidates are found first and then the desirable number of resulting significant points is selected among them by maximizing the weight function.

Although the method can be applied to any image, it is particularly devoted to SP detection in blurred images because it provides high consistence. We compared the performance of the method with

Acknowledgements

This work has been supported by the grant No. 102/96/1694 of the Grant Agency of the Czech Republic. Most of this work was done when Gabriele Peters and Jaroslav Kautsky were visiting the Institute of Information Theory and Automation, Academy of Sciences of the Czech Republic.

References (12)

L. Kitchen et al.
Gray-level corner detection
Pattern Recognition Letters
(1982)
S.-T. Liu et al.
Moment-preserving corner detection
Pattern Recognition
(1990)
R. Mehrotra
Corner detection
Pattern Recognition
(1990)
M. Trajkovic et al.
Fast corner detection
Image and Vision Computing
(1998)
X. Xie
Corner detection by a cost minimization approach
Pattern Recognition
(1993)
Beaudet, P.R., 1978. Rotationally invariant image operators. In: Proceedings of the International Joint Conference on...

There are more references available in the full text version of this article.

Cited by (52)

On the robustness of Harris detector in image watermarking attacks
2011, Optics Communications
This paper investigates the misplacement issue presented on the detection of feature points by using the well-known Harris detector, under several watermarking attacks. Harris detector, as a powerful tool in finding image's points of high importance, called “interest points”, has been widely used to locate specific image portions where watermarks are embedded, according to some watermarking insertion procedure. Although Harris detector constitutes a significant part of a typical feature-based watermarking process, no work dealing with its performance under certain attacks has been reported yet. This work studies the accuracy of Harris detector in finding the feature points, which define the watermarked regions of attacked images, firstly as an individual processing module and secondly as a part of a feature-based watermarking methodology. Detailed theoretical analysis followed by appropriate simulation experiments has shown that the incorporation of the Harris detector into the watermarking procedure presents misplacement issues that can lead the watermarking algorithm to unreliable results. In the way of decreasing these misplacement errors, a novel algorithm that extracts more stable interest points, with promising performance is also proposed.
Removing Atmospheric Turbulence Effects Via Geometric Distortion and Blur Representation
2022, IEEE Transactions on Geoscience and Remote Sensing
Improved Harris Corner Detection Algorithm Based on Canny Edge Detection and Gray Difference Preprocessing
2021, Journal of Physics: Conference Series
Grayscale-information-based Segmentation Registration for Fault Diagnosis of Train Components
2020, Proceedings of 2020 IEEE 9th Data Driven Control and Learning Systems Conference, DDCLS 2020
A new robust image feature point detector
2020, Proceedings of SPIE - The International Society for Optical Engineering
Translation compensation and resolution of ballistic target with precession
2019, Beijing Hangkong Hangtian Daxue Xuebao/Journal of Beijing University of Aeronautics and Astronautics

View all citing articles on Scopus

¹: Electronic Annexes available. See http://www.elsevier.nl/locate/patrec.

View full text

Robust detection of significant points in multiframe images1

Abstract

Introduction

Section snippets

Present state-of-the-art

Description of the proposed method

Numerical experiments

Conclusion

Acknowledgements

Pattern Recognition Letters

Pattern Recognition

Pattern Recognition

Image and Vision Computing

Pattern Recognition

Robust detection of significant points in multiframe images¹