Performance optimization of JPEG 2000 decompression for ground observation satellite operations

Due to its high compression performance, bit-error resilience, and support for both lossly and flexible lossy modes, JPEG 2000 is an image compression standard which is widely used in satellite image processing, remote sensing and many other fields. With the recent advancements in camera and telecommunication technology, resolution and download rate of ground observation satellite imagery are continually increasing. This makes it crucially important to simultaneously speed-up all image processing stages in satellite operations, including JPEG 2000 decompression. Existing work in literature regarding the optimization of JPEG 2000 for GPU is insufficient in terms of applicability to ground observation satellite operations. This work primarily focuses on speeding-up the JPEG 2000 decoder for GPU using CUDA optimization techniques. Also a hybrid decoder has been proposed which involves utilization of CPU and GPU together at maximum efficiency. Lastly, a distributed homogeneous JPEG 2000 decoder architecture has been proposed which is highly flexible and scalable. This work is anticipated to provide a useful reference and perspective to future studies regarding the speed-up of other stages in satellite image processing.