SBIR Phase II: Parallel Hardware Implementation of the Split and Merge Discrete Wavelet Transform for Wireless Communication

This Small Business Innovative Research (SBIR) Phase II project proposes to develop the Intellectual Property (IP) core of a novel image compression / signal decomposition algorithm based on the discrete wavelet transform (DWT). This is a fully parallel, scalable, multi-resolution, and low power implementation of the JPEG2000 DWT engine and is particularly well suited for use in both consumer applications at one end of the spectrum (as in reduced bit-rate web browsing over wireless communications channels as found in the next generation of web enabled cell phones) as well as in high-end commercial applications at the other end of the spectrum (as in non-linear video editing accelerators for the movie industry). This particular implementation is a highly efficient implementation of the DWT transform and makes use of a novel Overlap-State wavelet decomposition algorithm which minimizes memory, I/O and computational requirements. Over the next decade, spiraling consumer demand for fast mobile communication of voice and IP over increasingly integrated terrestrial and satellite based systems plagued by a limited electro-magnetic spectrum allocation necessitates the pursuit and development of better compression algorithms that a visually pleasing at low bit rates. As a consequence of extensive research, transform coding techniques now dominate every single image and video coding scheme proposed to-date. Consequently, efficient software and hardware based transform coding system designs and implementations have become a high priority objective. In fact, it is widely accepted that JPEG2000 will become the universally accepted format for digital images and high quality video - whether on the web, cable, over wireless systems, in digital cameras, printers, faxes or remote sensors. With its wavelet based image-coding technology, it offers features previously impossible in JPEG. Compared with the old baseline JPEG, the new JPEG2000 spec poses formidable technology challenges for the myriad of developers and OEM's planning on using it. The new standard uses coding algorithms based on the discrete wavelet transform (DWT) which is fundamentally different from the discrete cosine transform (DCT) JPEG spec. In JPEG2000, the importance of computational and especially memory bottlenecks has clearly increased several fold over the old specification. In fact, various implementations of computationally efficientCE wavelet transforms have been reported in recent years.