An Improved Pansharpening Method for Misaligned Panchromatic and Multispectral Data.

Hui Li,Yunwei Tang,Linhai Jing,Haifeng Ding

doi:10.3390/s18020557

Hui Li, Yunwei Tang + Show 2 more

Open Access

https://doi.org/10.3390/s18020557

Copy DOI

Abstract

Numerous pansharpening methods were proposed in recent decades for fusing low-spatial-resolution multispectral (MS) images with high-spatial-resolution (HSR) panchromatic (PAN) bands to produce fused HSR MS images, which are widely used in various remote sensing tasks. The effect of misregistration between MS and PAN bands on quality of fused products has gained much attention in recent years. An improved method for misaligned MS and PAN imagery is proposed, through two improvements made on a previously published method named RMI (reduce misalignment impact). The performance of the proposed method was assessed by comparing with some outstanding fusion methods, such as adaptive Gram-Schmidt and generalized Laplacian pyramid. Experimental results show that the improved version can reduce spectral distortions of fused dark pixels and sharpen boundaries between different image objects, as well as obtain similar quality indexes with the original RMI method. In addition, the proposed method was evaluated with respect to its sensitivity to misalignments between MS and PAN bands. It is certified that the proposed method is more robust to misalignments between MS and PAN bands than the other methods.

Highlights

Due to the physical constraints of remote sensing imaging and limited bandwidth of satellite transmission, a large number of currently operating satellites, such as SPOT, IKONOS (IK), QuickBird (QB), and WorldView-2/3, provide both a single relative high-spatial-resolution (HSR) panchromatic (PAN) band and several low-spatial-resolution (LSR) MS bands
The improved methods with different values for k used the same haze values as those employed by the original RMI method, for each test image
In order to highlight the performances of the comparing methods on the fusion of dark pixels, a spectral angle mapper (SAM) value between fused and reference dark pixels is calculated from each fused image generated at the degraded scale

Summary

Introduction

Due to the physical constraints of remote sensing imaging and limited bandwidth of satellite transmission, a large number of currently operating satellites, such as SPOT, IKONOS (IK), QuickBird (QB), and WorldView-2/3, provide both a single relative high-spatial-resolution (HSR) panchromatic (PAN) band and several low-spatial-resolution (LSR) MS bands. Fusion of PAN and MS images is referred to as pansharpening [1]. It is an important pre-processing step for generating high quality HSR. The major problem encountered by current pansharpening methods is to reduce spectral and spatial distortions. The CS methods can provide fused products with good spatial quality in most cases, but they sometimes suffer from spectral distortions

Results

Discussion

Conclusion