Шевченко В. Ю., Гнатушенко В. В., Тимошенко Ж. І.

Дніпропетровський національний університет імені Олеся Гончара


Pansharpening has been an actual research topic in the last few decades and numerous methods have been developed. These methods are generally categorized as arithmetic combination based (AMC) and component substitution (COS) techniques. The AMC methods involve direct arithmetic operation such as multiplication, addition division, weighted adding, etc. on the low resolution multispectral (MS) images to obtain high resolution images. The commonly known methods are Brovey method, Syntheic Variable Ratio (SVR) method, and high pass filtering. The COS-based substitution methods are performed after taking spectral or spatial transformation of the low resolution MS image. The popular COS approaches are the intensity-hue-saturation (IHS), the principal component analysis (PCA), and Multiresolution Analysis (MRA) basedpansharpening.

The PCA approach has been very commonly used for spectral transformation due to its ability to optimally compress the high dimension data. For this approach, the first principal component (PC) is substituted with the high resolution histogram-matched Pan image. However, the PCA approach is data dependent. For images with mostly vegetation/agricultural contents, this method yields very poor results with high spectral distortion. To alleviate this problem, we proposed a PCA-wavelet merger pan-sharpening method that took the advantage of the component substitution and the currently popular multiresolution approach. The transformation obtained by this method is very data dependent.

In this paper, we present new fusion alternatives based on the same concept, using the multiresolution wavelet decomposition to execute the detail extraction phase and the intensity-hue-saturation and principal component analysis procedures to inject the spatial detail of the panchromatic image into the multispectral one. The multiresolution wavelet decomposition has been performed using both decimated and undecimated algorithms and the resulting merged images compared both spectral and spatially. These fusion methods,as well as standard IHS-, PCA-, and wavelet-based methods have been used to merge Systeme Pour l’Observation de la Terre (SPOT) 4. We have estimated the validity of each fusion method by analyzing, visually and quantitatively, the quality of the resulting fused images. The methodological approaches proposed in this paper result in merged images with improved quality with respect to those obtained by standard.

In our opinion, the evaluation of a pan-sharpened image should be conducted under an application task, where we focus on the usefulness of the image data rather than its pixel value fidelity. The three applications on linear unmixing, detection, and classification explore the pixel spectral information within the spatial context of an image scene. This means that the spatial and spectral information are jointly evaluated. Based on the multiresolution wavelet decomposition experiments with different image scenes, we also conclude that the performance of a pansharpening technique may be varied with sensor and image content.

The alternative image-fusion methodological approaches presented in this paper, based on the intensity-hue-saturation transformation and the principal component analysis using the the multiresolution wavelet decomposition, allow to obtain merged images of higher quality than those obtained applying the IHS and PCA standard mergers. This higher quality is due to a selective incorporation into the multispectral image of just the spatial detail of the panchromatic image missing in the former, instead of performing a whole substitution. In addition, the injection of spatial detail extracted from the PAN image into the MS one just once when these methods are used results in images of higher spectral quality than those obtained applying standard wavelet-based merging methods where each MS band is fused with the PAN image separately.

As expected, artifacts are not detected in merged images when a translation-invariant undecimated algorithm is used to perform the MWD.

In the particular case of SPOT 4 images fusion, where the spectral bandwidth of the sensor mode does not overlap with the entire range of bandwidths. The bands, the methods based on PCA lead to better results than those based on the HIS transformation.