A concurrent spectral-screening PCT algorithm for remote sensing applications

The paper presents a concurrent algorithm for remote sensing applications that provides significant performance and image quality enhancements over conventional uniprocessor principal component transform (PCT) techniques. The algorithm combines spectral angle classification, principal component transform, and human centered color mapping. It is evaluated from an image quality perspective using images collected with the hyper-spectral digital imagery collection experiment (HYDICE) sensor, an airborne imaging spectrometer. These images correspond to foliated scenes taken from an altitude of 2000–7500 m at wavelengths between 400 nm and 2.5 μm. The scenes contain mechanized vehicles sitting in open fields as well as under camouflage. The algorithm operates with close to linear speedup on shared memory multiprocessors and can be readily extended to operate on multiple, low-cost PC-style servers connected with high-performance networking. A simple analytical model is outlined that allows the impact on performance of practical, application-specific properties to be assessed. These properties include image resolution, number of spectral bands, increases in the number of processors, changes in processor technology, networking speeds and system clock rates.