PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Feature Point Matching Using a Hermitian Property Matrix
Muhammad Haseeb and Edwin Hancock
In: SIMBAD - Similarity-Based Pattern Recognition - First International Workshop, September 28-30, 2011, Venice, Italy.

Abstract

This paper describes the computation of feature point correspondences using the spectra of a Hermitian property matrix. Firstly, a complex Laplacian (Hermitian) matrix is constructed from the Gaussian-weighted distances and the difference of SIFT [10] angles between each pair of points in the two images to be matched. Matches are computed by comparing the complex eigenvectors of the Hermitian property matrices for the two point sets acquired from the two images. Secondly, we embed the complex modal structure within Carcassoni’s [12] iterative alignment method to render it more robust to rotation. Our method has been evaluated on both synthetic and real-world data.

EPrint Type:Conference or Workshop Item (Paper)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Machine Vision
ID Code:7370
Deposited By:Edwin Hancock
Deposited On:13 February 2012