Restoration of Pansharpened Images by Conditional Filtering in the PCA Domain

Abstract

Pansharpening techniques aim at fusing low-resolution multispectral (MS) images and highresolution panchromatic (PAN) images to produce high-resolution MS images. Despite significant progress in the field, spectral and spatial distortions might still compromise the quality of the results. We introduce a restoration strategy to mitigate artifacts of fused products. After applying the Principal Component Analysis (PCA) transform to a pansharpened image, the chromatic components are filtered conditionally to the geometry of PAN. The structural component is then replaced by the locally histogram-matched PAN for spatial enhancement. Experimental results illustrate the efficiency of the proposed restoration chain.

fusionImage: An R package for pan‐sharpening images in open source software

“…This substitution maximizes the effect of the high-resolution PAN band on the fused bands resulting from the process (Shettigara, 1992). Although the PCA algorithm is one of the oldest and has been widely implemented in many commercial remote sensing packages, it is still used today with appropriate results when the objective is image classification (Gasparovic & Jogun, 2018) or in the restoration of pan-sharpened images (Duran & Buades, 2019).…”

Fusion of Panchromatic and Multispectral Images Using Multiscale Convolution Sparse Decomposition

“…However, spectral distortion can be always found in the fusion results due to the spectral response range differences between LR MS and PAN images. Thus, some methods [11]- [13] are proposed to alleviate the issue. For example, Kim et al [11] employed spatial principal component analysis (SPCA) to obtain more reasonable spatial structures from the bands of LR MS image.…”

Image Restoration for the MRA-Based Pansharpening Method

“…Major challenges related to image sensor and acquisition systems include two aspects: misregistration caused by differences between temporal or spatial coordinate systems of MS and PAN images, and aliasing of MS bands associated with optical properties of MS sensors, which usually results in spectral distortion and jagged edges [18], although the CS family has been proven to be less sensitive than the MRA family to these two problems in general [19], [20]. The pansharpening method, due to its own limitations, may inevitably cause some spectral distortion, artificial texture or excessive injection of spatial details.…”

“…Approaches mentioned in these studies all belong to restoration methods of preprocessing; there are also some restoration methods used as a postprocessing step for pansharpening. For example, conditional filtering in the PCA domain was used to enhance the spatial details of fusion results in the literature [18]. Correction of color distortion and noise near the edges was carried out based on the fusion results in the literature [24].…”