Mathematical methods for mapping image and data compression transforms to adaptive computing systems

The efficient computation of high-compression image transformations is the key to implementing video transmission in real-time applications that employ low-bandwidth communication links. In practice, such image compression applications could benefit from developments in adaptive computation, such as high-capacity field-programmable gate arrays (FPGAs) and reconfigurable SIMD meshes. It is noted that repetitive block structure in many compression algorithms tends to facilitate SIMD implementation. In contrast, the recursive data dependencies and variable partition size of pyramidally structured transforms such as wavelet-based compression may yield more effective processing with reconfigurable logic hardware such as FPGAs. In this paper, work-in-progress is presented that concerns mapping of image compression transforms to reconfigurable computing devices. The discussion begins with an overview of several compression algorithms in common use and their classification. The structure and function of two reconfigurable compression implementations (SIMD and FPGA) is described at a high level. The analysis emphasizes time and space complexity of each transformation and its associated implementations. Applications include exploitation of underwater acoustic channels for digital video telemetry