Hyperspectral Data Processing: Algorithm Design and Analysis

5.5 Synthetic Image Experiments

The synthetic images used for experiments conducted in this section are the six scenarios (three TI scenarios, TI1, TI2, TI3, and three TE scenarios, TE1, TE2, TE3) designed in Chapter 4. A major advantage of using these synthetic images is to allow users to simulate complete ground truth to evaluate the effectiveness of algorithms to be designed and developed for quantitative study and comparative analysis. In these six simulated scenarios there are 100 pure panel pixels, 20 mixed panel pixels, and 10 subpixel panels simulated by five mineral signatures, alunite (A), buddingtonite (B), calcite (C), kaolinite (K), and muscovite (M) in Figure 1.12(c) along with image background simulated by the sample mean. These 130 panel pixels are either implanted into the image background to simulate the three scenarios TI1, TI2, and TI3, or embedded into the image background to simulate the three scenarios TE1, TE2, and TE3. So, technically speaking, there are only five spectrally distinct pure signatures and one spectrally distinct mixed background signature used to simulate synthetic images.

5.5.1 Data Characterization-Driven Criteria

The data characterization-driven criteria developed in Section 5.3 for VD estimation can be classified into two categories.

1. Category I: criteria involving no parameters and producing a single constant value regardless of applications

2. Category II: criteria involving one parameter, either error threshold ε or false alarm ...