Wavelet-based corner detection technique using optimal scale

In this paper we present a novel technique for wavelet-based: corner detect ion using singular value decomposition (SVD). Here SVD facilitates the sele ction of global natural scale in discrete wavelet transform. We define natu ral scale as the level associated with most prominent (dominant) eigenvalue . Eigenvector corresponding to dominant eigenvalue is considered as the opt imal scale. The corners are detected at the locations corresponding to modu lus maxima. Results indicate the suitability of the approach. Comparison wi th a recently proposed technique is also provided. (C) 2002 Elsevier Scienc e B.V. All rights reserved.