Pansharpening with multi-CAE: impact of patch size and overlapping pixels on spectral and spatial distortion

[...]the relationship between the original PAN image and its degraded version is utilized to reconstruct the high-resolution MS image; in addition, an intensity component of MS image, which is obtained using an adaptive intensity-hue-saturation (AIHS), is reconstructed by utilizing the aforementioned relationship. A wide variety of pansharpening studies have been introduced in the literature, which can be grouped into three groups [9]: i) component substitution (CS) [10], [11]; ii) multiresolution analysis (MRA) [12], [13]; and iii) sparse representation (SR) [14], [15] based methods. The MRA based-methods such as indusion [21], wavelet transform [22], and generalized Laplacian pyramid (MTF-GLP) approach [23] leverages the detailed map by computing the variance between the PAN image and its low-resolution counterpart via PAN image decomposition. CAEs CAEs are considered unsupervised learning of convolutional filters. [...]any input would be used to extract features once a CAE has been learned [34].