Detection of Small-Scale Mineral Deposits in Super-Resolved THEMIS TIR Data

Introduction: Super-resolution is the process of improving spatial resolution from that of the original data source (or native) resolution. A first-order approach is the fusion of original data with an additional source, which has the desired resolution. There are a variety of techniques that can be used to fuse these data sets; however, a trade-off has been noted between techniques that are the most visually appealing and those that are most radiometrically accurate [1]. The technique for super-resolution presented here is a modification of an algorithm [2] that was originally tested successfully using multi-resolution data from the Earth orbiting Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) instrument [3]. The spatial and spectral resolution of this instrument is broadly similar to the Mars orbiting Thermal Emission Imaging System (THEMIS) instrument [4]. The current study investigates the applicability of this technique to enhance the THEMIS thermal infrared (TIR) data using the instrument's visible (VIS) data, thus providing radiometrically-accurate data at an improved spatial resolution of 36 m/pixel. This provides an independent approach to traditional sub-pixel deconvolu-tion techniques, and it can be used in the search for small-scale temperature and/or compositional anomalies. Data and Methods: Data from the THEMIS instrument are organized into two wavelength regions (VIS and TIR), with the VIS having fewer bands but higher spatial resolution. First, VIS pixels in a given image are clustered using the previously described methodology [6]. Upon completion, a tree is created with the co-located TIR pixels on a per-cluster basis, the branches of which represent spectrally-similar pix-els. After completion of clustering, it is possible to super resolve the image. Pixel values are assigned and then a radiometric correction is applied to the entire image as previously described for ASTER data [6], such that the super-resolved image, when degraded back to the lower resolution, will match the original image. Thus, this approach is radiometrically accurate and fully reversible.