High Compression of Infrared Data

ABSTRACT: Over the past decades, the capabilities of sensors to gather data has greatly outstripped the abilities of communication links to transmit that data. Accordingly, leveraging advanced sensors requires effective techniques for data compression. In particular, modern satellites provide a vast amount of real-time information that is essential to decision makers, both military and commercial. Unfortunately, space platforms have extremely limited space, weight, and power budgets. Making the problem even more challenging, these vast data sources must be communicated over low-bandwidth communication links and the data that arrives must be trustworthy. The key novelty of our work is that our techniques achieve compression ratios that are similar to (if not better than) those achieved by standard lossy compression schemes while at the same time maintaining an a priori rigorous error bound profile for each pixel. Intuitively speaking, a prescribed error tolerance provides a small amount of freedom that may be used to improve the compression ratio. Our key innovation is the development of algorithms that take advantage of this freedom in a computationally efficient manner while guaranteeing that no error bounds are ever violated. BENEFIT: Presently, there are a number of commercial vendors that provide image compression solutions on a dedicated hardware chip. However, these implementations are not necessarily designed for space applications that have radiation hardening, and specialized data transmission requirements. The primary mission of military space-based IR sensors is to detect and track point targets, e.g., the OPIR mission. To ensure that all targets of interest (some dim) are detectable, the compression algorithm must faithfully preserve this data. Otherwise, the detection algorithm, which runs in the ground station, does not have the information that it requires to function. Thus, a unique requirement of the OPIR mission (the primary application) is that the quality of each pixel in the IR image must be preserved to a specific degree (which is determined, e.g., by the detection algorithm). Numerica differentiates its solution from these offerings by providing a unique compression algorithm that is able to guarantee an error-bound on each individual pixel. We intend to develop a plan for implementing this specialized algorithm on a space processor architecture, which will ultimately provide a superior solution to other offerings available today.