Data fidelity-oriented spatial-spectral fusion of CRISM and CTX images

Abstract

The Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) is a Mars-dedicated compact reconnaissance imaging spectrometer that captures remote sensing data with very fine spectral resolution. However, the spatial resolution of CRISM data is relatively coarse (18 m), limiting its application to regional scales. The Context Camera (CTX) is a digital camera equipped with a wide-angle lens, providing a finer spatial resolution (6 m) and larger field-of-view, but CTX provides only a single panchromatic band. To produce CRISM hyperspectral data with finer spatial resolution (e.g., 6 m of CTX images), this research investigated spatial-spectral fusion of 18 m CRISM images with 6 m CTX panchromatic images. In spatial-spectral fusion, to address the long-standing issue of incomplete data fidelity to the original hyperspectral data in existing methods, a new paradigm called Data Fidelity-oriented Spatial-Spectral Fusion (DF-SSF) was proposed. The effectiveness of DF-SSF was validated through experiments on data from six areas on Mars. The results indicate that the fusion of CRISM and CTX can increase the spatial resolution of CRISM hyperspectral data effectively. Moreover, DF-SSF can increase the fusion accuracy noticeably while maintaining perfect data fidelity to the original hyperspectral data. In addition, DF-SSF is theoretically applicable to any existing spatial-spectral fusion methods. The 6 m CRISM hyperspectral data inherit the advantages of the original 18 m data in spectral resolution, and provide richer spatial texture information on the Martian surface, with broad application potential.