Work

Concurrent execution of lossy compression and anomaly detection of hyperspectral images on FPGA devices

Year: 2025

Type: article

Abstract: Hyperspectral sensors capture a wide range of spectral data, making them crucial for Earth observation applications, but this fact poses significant challenges for embedded systems with limited resources. Nevertheless, most studies only perform one application at the same time, so multi-applications in the same device are not considered since high-performance and low hardware resources are limited. In this sense, this paper presents a hardware-friendly algorithm for concurrently execution of anomaly detection and lossy compression for hyperspectral imaging. The proposed algorithm reuses a hardware platform to perform both tasks in parallel, offering a validated hardware architecture designed for deployment on a cost-optimized FPGA device. The experimental results show that our hardware component can process hyperspectral images with a resolution of $$825\times 1024$$ 825 × 1024 pixels and 160 bands in 0.53 s (486 MB/s), with a power consumption of 1.08 watts (399 MB/W).

Source: Journal of Real-Time Image Processing

DOI: 10.1007/s11554-025-01692-0

Authors: Julián Caba, Jesús Barba, María Díaz, José Luis Mira, Sebastián López, Juan Carlos López

Institutions: University of Castilla-La Mancha, Oceanic Platform of the Canary Islands

Language: English

Topic: Advanced Data Compression Techniques

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Estimated APC paid: $3,290

Funders: Ministerio de Economía y Competitividad, Universidad de Castilla-La Mancha, NextGenerationEU