A Bidimensional Empirical Mode Decomposition Method for Fusion of Multispectral and Panchromatic Remote Sensing Images

Weihua Dong,Zhilin Li,Xian'En Li,Xiangguo Lin

doi:10.3390/rs6098446

Abstract

This article focuses on the image fusion of high-resolution panchromatic and multispectral images. We propose a new image fusion method based on a Hue-Saturation-Value (HSV) color space model and bidimensional empirical mode decomposition (BEMD), by integrating high-frequency component of panchromatic image into multispectral image and optimizing the BEMD in decreasing sifting time, simplifying extrema point locating and more efficient interpolation. This new method has been tested with a panchromatic image (SPOT, 10-m resolution) and a multispectral image (TM, 28-m resolution). Visual and quantitative assessment methods are applied to evaluate the quality of the fused images. The experimental results show that the proposed method provided superior performance over conventional fusion algorithms in improving the quality of the fused images in terms of visual effectiveness, standard deviation, correlation coefficient, bias index and degree of distortion. Both five different land cover types WorldView-II images and three different sensor combinations (TM/SPOT, WorldView-II, 0.5 m/1 m resolution and IKONOS, 1 m/4 m resolution) validated the robustness of BEMD fusion performance. Both of these results prove the capability of the proposed BEMD method as a robust image fusion method to prevent color distortion and enhance image detail.

Highlights

With the development of remote sensing in observation technology, high multispectral resolution and spatial resolution of remote sensing images such as TM, SPOT, IKONOS, WorldView and GeoEye images are obtained by various types of sensors and applied in geographic condition monitoring, mapping, extracting and interpreting information and so on
Image fusion methods [35]
The IHS, PCA, Brovey, high-pass spatial filter (HPF) and Laplacian pyramidal decomposition (LPD) methods result in spectral distortions

Summary

Introduction

With the development of remote sensing in observation technology, high multispectral resolution and spatial resolution of remote sensing images such as TM, SPOT, IKONOS, WorldView and GeoEye images are obtained by various types of sensors and applied in geographic condition monitoring, mapping, extracting and interpreting information and so on. To enhance the quality of the fused images, researchers proposed image fusion methods in fusing panchromatic (PAN) and multispectral (MS) images. These methods can be divided into three levels: the pixel level, the feature level and decision level [1]. Within various pixel level fusion techniques, most algorithms can be classified in three categories: the component substitution fusion techniques, modulation-based fusion techniques and multi-resolution analysis-based (MRA) fusion techniques [2]. The most commonly used fusion method in remote sensing is the component substitution fusion technique, which has been integrated into many popular remote sensing image-processing software products, such as ENVI

Results

Discussion

Conclusion