Multithreaded algorithm for band interleave conversion of hyperspectral images

This paper presents a high-performance data conversion algorithm of spectral bands (Band Interleave Conversion) of the sensor EO-1 Hyperion, providing the ability to run it on multiprocessor platforms in multi-threaded mode and ensures the effective imple­mentation on low I/O systems. Proposed implementation of proportional reading image data into RAM, followed by the placement of values (Digitals Number) in the target buffer arrays in multiple threads, calculated by the number of spectral channels or the number of lines of the image. Results of testing the algorithm presented with their basic counterparts software package Exelis ENVI. It is shown that the speed (second) of the multithreaded algorithm hundreds of times higher than for the native algorithm. In particular, the process of multithreaded band interleave conversion on the 8-core architecture with the use RAMDisk I/O-subsystem was only 45 seconds, while the native algorithms needed in similar environment of about 6800 seconds. The paper presents an Exelis ENVI extension, rea­lizing the developed algorithm based on the GUI-WIDGETS integration with packages Java SwingX using IDL-Bridge technology.