Planning for a Real-Time JPEG2000 Compression System

Since the initiation of JPEG in the mid-to-late 1980's, the JPEG standard has remained a viable solution for image compression in many applications. However, many modern applications seek significant improvements in image compression, particularly for compressed image representations at higher compression gains. Coupled with the growing interest in improving compression technology is the fact that bigger image sizes have become increasingly popular. In addition to the combined challenges of higher compression gains, large image sizes, and extremely long image sequences (e.g. in persistent surveillance durations consisting of hours, days, or months), many applications desire real-time or near real-time availability of compressed images for viewing and distribution. JPEG2000 is an often sought compression solution and is being increasingly adopted for future compression systems. However, for real-time compression systems one of the major challenges JPEG2000 presents is its very demanding computational intensity. It is this high computational intensity that mitigates much of the current interest in expanding JPEG2000's use in embedded and real-time applications. Planning for the implementation of a JPEG2000 system requires addressing all of the challenges listed above, particularly the computational needs. Fortunately, this intensive processing can be partitioned into concurrent processing modules that are suitable for FPGA implementation. This paper focuses on providing an outline of the major components of a JPEG2000 encoder and the requirements they impose for an example embedded application.