Hyperspectral Data Processing: Algorithm Design and Analysis

29.5 Real Image Experiments

So far, we have discussed much about the utility of hyperspectral signature self-clarification using wavelet discrete decomposition, which can perform self-clarification given a relatively pure and clean signal as a reference. Intuitively, the farther the two hyperspectral signatures, the more the number of iterations required. So this can be used to judge if one hyperspectral signature is much closer to the another and further develop a technique to perform subpixel identification. In the following experiments, two different ways to choose reference were considered, one using the five average panel signatures, , as the reference, and the other using the average of the first two column panel signatures as the reference. Subpixel panels were the signatures to be identified.

Experiment 29.5.1 (WSCA-SSC)

In this experiment, five subpixel panels p₁₃, p₂₃, p₃₃, p₄₃, and p₅₃ in the last column were considered as contaminated signatures of the five panels signatures p₁, p₂, p₃, p₄, and p₅ in Figure 1.16 that were used as references for subpixel panels p₁₃, p₂₃, p₃₃, p₄₃, and p₅₃ for self-correction. In other words, WSCA-SSC was applied here to recover the original five panels signatures p₁, p₂, p₃, p₄, and p₅ from their corrupted versions, here considered as ...