Title

A Region Based Image Matching Method With Regularized Sar Model

Abstract

In this paper, we propose a new region-based image matching method to find the user defined regions in other images. We use color histogram and SAR (simultaneous autoregressive) model parameters as matching features. We characterize the spatial structure of image region with its block features, and we match the image region in target images with spatial constraints. SAR model was usually used to characterize the spatial interactions among neighboring pixels. But the spectrum of the transition matrix G in the SAR model is not well distributed. Therefore in this paper, we use a regularized SAR model to characterize the spatial interactions among neighboring image blocks, which is based on the solution of a penalized LSE (Least Squares Estimation) for computing SAR model parameters. The experimental results show that our method is effective. © Springer-Verlag Berlin Heidelberg 2004.

Publication Date

1-1-2004

Publication Title

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

Volume

3331

Number of Pages

263-270

Document Type

Article

Personal Identifier

scopus

DOI Link

https://doi.org/10.1007/978-3-540-30541-5_33

Socpus ID

35048867891 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/35048867891

This document is currently not available here.

Share

COinS