Topic:
Hardware Acceleration of the DWT based on the Fast Lifting Wavelet Transform

Team Members:
1. Singh,Vidhu Niti20
2. Kulkarni , Pritam

Introduction:

Wavelet transforms have been applied to a number of fields ranging from
pure mathematics to applied science.  Pure software implementations of
the Discrete Wavelet Transform, however, appears to be the performance
bottleneck in real-time systems in terms of performance.  In this
project, we will investigate the possibility of hardware acceleration of
the Discrete Wavelet Transform and compare against the pure software
implementation.  The DWT algorithm is based on the Fast Lifting Wavelet
Transform. 

General Purpose processors executes programs sequentially.  The amount
of parallelism is limited to a few functional units, while in the
hardware domain multiple parallel operating functional units can be
instantiated, which can lead to performance improvement.  Another
limitation of general-purpose processors is their comparatively small
number of registers.  This deteriorates the performance, as a memory
access is significantly slower than a register access.  In hardware, the
amount of data storage elements can be substantially larger, and can
hold the needed data for fast future access. 

Proposed work :

We will implement the Fast Lifting Wavelet Transform in software and
analyse in order to extract parts of the algorithm to be accelarated. 
Some of the concepts learnt in the class such as single assignment
transform, unfolding, etc.  will be used to improve the software
performance. 

A literature search for hardware accelaration for the lifting based DWT
has been done by us.  A thorough investigation of the ideas in the
papers show a considerable amount of operations that can be performed in
parallel.  We plan to implement an efficient hardware design which will
significantly reduce the amount of data transfer to and from the memory
by buffering data that once has been read from the memory.  In addition
to this we will introduce parallelism to speed up the performance.  The
coding will be done in verilog using Vsim and will be synthesized using
Synopsis Design Compiler. 

References:

[1]W.  Sweldens, 94Lifting Scheme: A New Philosophy in Biorthogonal
Wavelet Constructions,94 in Proc.  SPIE-2569, Wavelet Applications in
Signal and Image Processing III, editors A.  F.  Laine and M.  Unser,
pp.68-79. 

[2]G.  Fernandez, S.  Periaswamy, and W.  Sweldens,94LIFTPACK: A
software package for wavelet transforms using lifting94.  Wavelet
Applications in Signal and Image Processing IV, Proc.  SPIE 2825, 1996,