Proc. SPIE. 8871, Satellite Data Compression, Communications, and Processing IX
KEYWORDS: Hyperspectral imaging, Principal component analysis, Image processing, Remote sensing, Field programmable gate arrays, Data processing, Very large scale integration, Algorithm development, Matrix multiplication, System on a chip
The Pixel Purity Index (PPI) algorithm is one of the most successful algorithms for hyperspectral image endmembers extraction. But it has high computational complexity so it is hard to meet the real-time processing demands of some onboard application. In this paper, we present a novel Very-Large-Scale Integration (VLSI) architecture for PPI algorithm to meet the on-board demands. With parallelism and improved I/O communication strategy, our implementation is significantly time saving than other architectures in the same hardware resources. We evaluate our implementation using the well-known “Cuprite” scene and assess endmembers signature purity using the U.S. Geological Survey (USGS) library. It demonstrates that our hardware implementation can get endmembers in less processing time to meet the onboard demands.