| Summary: | Spectral Imagery provides multi-dimensional data, which are difficult to display in standard three-color image formats. Tyo, et al. (2001) propose an invariant display strategy to address this problem. This approach is to mimic the dynamics of human perception. The dimensionality of the data are reduced by using a Principal Component (PC) transformation, and then displayed by making used of a Hue, Saturation, and Value (HSV) display transform. This study addresses the PC transformation strategy , looks for a global eigenvector via 3D visualization of HSV color space information, and examines the suggested algorithm to provide the most intuitive display. The user interface created in this thesis is capable of computing the necessary implementation of the proposed strategy, viewing selected Region of Interest (ROI) in HSV color space model in 3D, and viewing the 2D resultant image. A demonstration application uses Java language including Java2D, Xj3D Player, Document Object Model (DOM) Application Program Interfaces (API), and Extensible 3D Language (X3D). The Java2D API enables the user to load imagery, process data, and render results in a two-dimensional (2D) view. Xj3D and DOM APIs are introduced to visualize Tyo's invariant display strategy in three-dimensional (3D) views and then to save results as X3D scenes. These techniques appear to be inherently valuable and can serve as the basis for further research. Through this thesis, 3D visualization of the proposed algorithm successfully showed PC transformed data does form a conical shape in HSV color space. Also, a comparison of PC transformed data with HSV color space revealed the hue angle needed to be adjusted. The application of this adjustment to multiple scenes produced consistent results. However, this hue adjustment left other scene elements in non-ergonomic colors and brought up the issue of further enhancement of the algorithm.
|
|---|
