Optimization of the Continuous Wavelet Transform for DSP Processor Implementation

The redundant wavelet transform is an effective tool when emphasis is on the analysis of non-stationary signals and on localization and characterization of singularities. Here we describe an optimized method to implement a B-spline based redundant wavelet transform (RWT) on a digital signal processor (DSP) for integer scales. Expressions are derived to give an exact operation count at any integer scale m for any B-spline of order n. Finally experimental results are given using cubic B-spline as scaling function and first- and second-order derivative of B-splines as wavelets. It has been shown that optimized method improves the execution speed over the standard method by 20-28%