On analysis of bi-dimensional component decomposition via BEMD

This paper investigates how the bi-dimensional empirical mode decomposition (BEMD) behaves in digital images via two components. First, some definitions of components are given. Then the theoretical analysis is presented to show that in digital images there might be sparse extrema sets and the sifti...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Pattern recognition 2012-04, Vol.45 (4), p.1617-1626
Hauptverfasser: Guanlei, Xu, Xiaotong, Wang, Xiaogang, Xu
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:This paper investigates how the bi-dimensional empirical mode decomposition (BEMD) behaves in digital images via two components. First, some definitions of components are given. Then the theoretical analysis is presented to show that in digital images there might be sparse extrema sets and the sifting process of BEMD converges to the true component provided that the sparse extrema sets do exist. Also, the feature of 2D component in digital images is explored via composite two-component signals. The three-dimensional cubes disclosing the performance of BEMD are presented, which turn out to be in good agreement with intuition and physical interpretation. This paper has also shown that the sampling period, the cross-angle, the amplitude ratio and the frequency ratio between the components will affect the separation of them and some instructive conclusions are achieved as well. The theoretical analysis is provided for analyzing the observed behaviors and supported by numerical experiments. The main aim of this study is primarily to contribute to a better understanding of the possibilities and limitations offered by BEMD in digital images. ► The performance of BEMD is given. ► Influences of the sampling period, cross-angle, amplitude and frequency ratios are discussed. ► Constructive conclusions are given in BEMD.
ISSN:0031-3203
1873-5142
DOI:10.1016/j.patcog.2011.11.004