High level modular implementation of a lossy hyperspectral image compression algorithm on a FPGA

Abstract

In this paper, a Field Programmable Gate Array (FPGA) implementation of the LCE (Lossy Compression for ExoMars) algorithm is presented. This algorithm shows a good quality/compression ratio tradeoff for hyperspectral images, at the expenses of a higher complexity with respect to lossless algorithms. In order to deal with this complexity levels, high level synthesis (HLS) tools, such as Catapult C, have been used together with Precision RTL so that a final implementation was obtained in a Virtex-5 FPGA. The results show a performance slightly higher than a previous lossless/near-lossless algorithm implementation, in terms of operating frequency (87 MHz vs. 81 MHz), with a reduced number of memory blocks, at the expenses of an increase in the number of digital signal processing (DSP) slices used.