Publication Type : Conference Paper
Thematic Areas : Wireless Network and Application
Publisher : Proceedings of the 3rd International conference on Visual Information Engineering
Source : Proceedings of the 3rd International conference on Visual Information Engineering, Bangalore, India (VIE 2006), September 2006.
Url : https://ieeexplore.ieee.org/document/4286738
Campus : Amritapuri
School : School of Engineering
Center : Amrita Center for Wireless Networks and Applications (AmritaWNA)
Department : Wireless Networks and Applications (AWNA)
Year : 2006
Abstract : We extend non-canonical multiwindow Gabor representations to quincunx sampling and discuss the advantages of representation of textures using non-canonical discrete multiwindow Gabor expansions, with the time-frequency plane sampled in a quincunx fashion. We show that the quincunx sampling is more efficient in terms of placement of the tiling zones in the time-frequency and achieves better time-frequency localisation when compared to the usual rectangular sampling. To emphasise the advantage of quincunx sampling, we compare images reconstructed using multi-window Gabor frames using the canonical dual, a non-canonical dual, and finally a non-canonical dual under a quincunx sampling lattice and discuss our results. We show that the use of non-canonical duals permits a great deal of flexibility in the choice of analysis functions (and consequent generation of Gabor coefficients) and that duals can be optimised to the situation at hand.
Cite this Research Publication : N. K. Subbanna, and Y. Y. Zeevi, "Image Representations Using Non-Canonical Discrete Multiwindow Gabor Frames", in Proceedings of the 3rd International conference on Visual Information Engineering, Bangalore, India (VIE 2006), September 2006.