In this paper, multivariate regression is adopted to improve spectral quality, without diminishing spatial quality, in image fusion methods based on the well-established component substitution (CS) approach. A general scheme that is capable of modeling any CS image fusion method is presented and discussed. According to this scheme, a generalized intensity component is defined as the weighted average of the multispectral (MS) bands. The weights are obtained as regression coefficients between the MS bands and the spatially degraded panchromatic (Pan) image, with the aim of capturing the spectral responses of the sensors. Once it has been integrated into the Gram-Schmidt spectral-sharpening method, which is implemented in environment for visualizing images (ENVI) program, and into the generalized intensity-hue-saturation fusion method, the proposed preprocessing module allows the production of fused images of the same spatial sharpness but of increased spectral quality with respect to the standard implementations. In addition, quantitative scores carried out on spatially degraded data clearly confirm the superiority of the enhanced methods over their baselines.

This paper aims at defining a new paradigm (hypersharpening) in remote sensing image fusion. In fact, due to the development of new instruments, thinking only in terms of pansharpening is reductive. Even though some expressions as hyperspectral (HS) pansharpening already exist, there is not a suitable definition when multispectral/hyperspectral data are used as source to extract spatial details. After defining the hypersharpening framework, we draw the readers' attention to its peculiar characteristics, by proposing and evaluating two hypersharpening methods. Experiments are carried out on the data produced by the updated version of SIM-GA imager, designed by Selex ES, which is composed by a panchromatic camera and two spectrometers in the VNIR and SWIR spectral ranges, respectively. Owing to the different resolution factors among panchromatic, VNIR and SWIR data sets, we can apply hypersharpening to fuse SWIR data to VNIR resolution. Comparisons of hypersharpening with “traditional” pansharpening show hypersharpening is more effective.

Multiresolution analysis (MRA) and component substitution (CS) are the two basic frameworks to which image fusion algorithms can be reported when merging multispectral (MS) and panchromatic (Pan) images (pansharpening), acquired with different spatial and spectral resolutions. State-of-the-art algorithms add the spatial details extracted from the Pan into the MS data set by considering different injection strategies. The capability of efficiently modeling the relationships between MS and Pan is crucial for the quality of fusion results and particularly for a correct recovery of local features with a consequent reduction of spectral distortions. Although context-adaptive (CA) injection models have been proposed in the MRA framework, their adoption in CS schemes has been scarcely investigated so far. In this letter, CA strategies are compared with global models by considering a general protocol in which both MRA- and CS-based schemes can be described. Qualitative and quantitative results are reported for three high-resolution data sets from two different sensors, namely, IKONOS and simulated Pleiades. The score gains of well-known and novel quality figures show that CA models are more efficient than global ones.

In this paper, the characteristics of multispectral (MS) and panchromatic (P) image fusion methods are investigated. Depending on the way spatial details are extracted from P, pansharpening methods can be broadly labeled into two main classes, corresponding to methods based on either component substitution (CS) or multiresolution analysis (MRA). Theoretical investigations and experimental results evidence that CS-based fusion is far less sensitive than MRA-based fusion to: 1) registration errors, i.e., spatial misalignments between MS and P images, possibly originated by cartographic projection and resampling of individual data sets; 2) aliasing occurring in MS bands and stemming from modulation transfer functions (MTF) of MS channels that are excessively broad for the sampling step. In order to assess the sensitiveness of methods, aliasing is simulated at degraded spatial scale by means of several MTF-shaped digital filters. Analogously, simulated misalignments, carried out at both full and degraded scale, evidence the quality-shift tradeoff of the two classes. MRA yields a slightly superior quality in the absence of aliasing/misalignments, but is more penalized than CS, whenever either aliasing or shifts between MS and P occur. Conversely, CS generally produces a slightly lower quality, but is intrinsically more aliasing/shift tolerant.