Spectral Reconstruction of the Pixels in a Satellite Image through Mixing Process

This research introduces a technique based on the spectral mixing i.e. something opposite to the well known unmixing technique. This is done by rearrangement and reconstruction of the pixels surface materials in order to make it spectrally look like some predefined pixels. The method is called Pixel MAKeup (PMAK). This means that it is tried to find a way to select materials where when spreading them out on the pixel surface, the resultant spectral reflectance of the pixel gets similar to some predefined surface covers. The PMAK method was run for two subsets of one LISS-III P6 image in three bands where one of them is taken as the primary (the one that is intended to be reconstructed) and the other one is the secondary (the one that the reconstructed primary subset should be looked like). The RMSE between the secondary and PMAK output was found to be 0.0061, 0.0057 and 0.0035 for bands 2, 3 and 4 respectively. The number of materials used for this pixel reconstruction were as little as 3. Of course if we increase the number of bands, then the number of these materials may increase substantially. However the simple concepts of the PMAK method would make it applicable for any sensor with any number of bands.

Keywords:

unmixing , makeup , endmember , pixel , image