Remote-Sensing Image Compression Using Embedded Multicore Platforms With Energy Consumption Measurement

Abstract

This letter was motivated by a practical problem in the optical image compression scope captured by remote-sensing satellites with energy restriction. A large number of embedded CPU architectures have a multicore capability. On the other hand, the most used algorithms for image compression were developed to perform the compression task using a single core. Therefore, this letter proposes a scheme to allow the most used compressor algorithms to run in parallel to perform the compression task more efficiently (Consultative Committee for Space Data Systems (CCSDS), Set Partitioning in Hierarchical Trees (SPIHT), and JPEG2000). In addition, the most used image compression algorithms were developed without considering the energy expenditure as an issue. Thus, this letter will use the algorithm's energy expenditure as one of the evaluation criteria, whose analysis can be extended to other systems that operate using batteries. Considering that the processing time and energy expenditure depend on the platform and may have different magnitudes, this letter will use two processing platforms for the analysis. The results will be analyzed in terms of rate–distortion–energy with the help of an external simple energy measurement tool. To avoid working only with the well-known rate–distortion curves of the image compressors mentioned previously, the scheme proposed herein allows using more than one algorithm to perform the image compression. The results show several gains in the processing time, especially for the SPIHT algorithm, which uses only 5% of the original processing time and 12.8% of energy expenditure (for a compression rate of 1.5 b per pixel), for the platforms evaluated using this scheme.