Continuous wavelet transform with arbitrary scales and O(N) complexity
Introduction
The continuous wavelet transform (CWT) of a signal f with the wavelet ψ is defined as
It can be interpreted as the correlation of the input signal with a time-reversed version of ψ rescaled by a factor of a. For a 1-D input signal, the result is a 2-D description of the signal with respect to time b and scale a. The scale a is inversely proportional to the central frequency of the rescaled wavelet ψa(x)=ψ(x/a) which is typically a bandpass function; b represents the time location at which we analyze the signal. The larger the scale a, the wider the analyzing function ψa, and hence smaller the corresponding analyzed frequency. The output value is maximized when the frequency of the signal matches that of the corresponding dilated wavelet. The main advantage over the Fourier transform (FT) analysis is that the frequency description is localized in time. The advantage over the short-time Fourier transform (STFT) is that the window size varies; low frequencies are analyzed over wide time windows, and high frequencies over narrow time windows, which is more effective than to use a fixed-size analysis. Typical applications of the CWT are the detection and characterization of singularities [3], [14], pattern recognition [6], image processing [4], [15], fractal analysis [2], [12], [23], noise reduction [11] and the analysis of biomedical signals [7], [10], [25].
The main contribution of this paper is the development of a fast algorithm for the computation of the CWT at any real scale a and integer time localization b. Mallat's fast wavelet algorithm [12] uses the multiresolution properties of the wavelet to compute the CWT at dyadic scales a=2i and time shifts b=2ik, k∈Z [17]; it achieves an overall O(N) complexity. Other techniques compute the wavelet transform at dyadic scales and integer time points with an ‘à trous’ approach. Their complexity per scale is O(N), the same as Mallat's algorithm, but with a larger leading constant [3], [5], [9], [16].
Despite their speed, these methods may not be precise enough for some applications, since a dyadic scale progression cannot be finer than an octave sub-band decomposition. To achieve a better scale resolution, other approaches have been proposed, either based on M-band decomposition inside an octave [17], [24] or on a generalization of the two-scale relation to general integer N-scale relations [8], [22]. However, none of these algorithms can handle arbitrary scales.
Our purpose here is to develop a novel and fast algorithm that works for any real value of a. It takes advantage of a B-spline decomposition of the input signal and of the mother wavelet. The method exploits the fact that B-splines are compactly supported and that the convolution of two B-splines can be expressed analytically [13].
Section snippets
Operators and definitions
First, we introduce some operators and definitions that will be helpful to solve our problem.
We express a B-spline of degree n aswhere x+n=max(x,0)n is a one-sided power function; Δn+1 denotes the (n+1)-fold iteration of the finite difference operator Δ=δ(x)−δ(x−1). This latter operator also corresponds to a discrete convolution (digital filter) whose z-transform is Δ(z)=1−z−1. Likewise, Δn+1(z)=(1−z−1)n+1.
We have the equivalencewhere D−1 is the
Spline wavelets
Among all existing wavelet bases, B-spline wavelets have the advantage of possessing an explicit formula [1]; most wavelets are defined only implicitly by means of a refinement filter. For example, the well-known Haar wavelet is a weighted sum of two B-splines of degree 0. Other wavelets, such as the first derivative or the second derivative of a Gaussian (Mexican hat wavelet), can be closely approximated by linear combination of B-splines of sufficiently high degrees (n⩾2) [19].
The description
Results
Here, we discuss the implementation of our fast CWT algorithm and compare its execution time with a FFT-based implementation. As an example of application, we show the analysis of a biomedical signal.
Discussion: integer scale method
In a previous paper [22], Unser et al. describe a fast algorithm for the CWT computation at integer scales using B-splines as basis functions. Their method can be shown to be equivalent to ours when a is an integer. This follows from the identityNext we summarize the two ways of computing the CWT at integer scales. We start with the method of Unser et al. for which we write
Conclusions
We have presented a novel B-spline-based CWT algorithm that is able to compute the CWT at any real scale, making it possible to use arbitrary scale progressions. The computational complexity per computed coefficient is O(1), as is the case with the most efficient wavelet algorithms for dyadic or integer scales. The overall operation count only depends on the wavelet shape and on the degrees of the B-spline basis on which the wavelet and the input signal are described, but is independent of the
Acknowledgements
This work was funded by the Swiss National Science Foundation, Grant #2100-053540. We also thank the reviewers for their insightful comments.
References (25)
- et al.
A family of polynomial spline wavelet transforms
Signal Process. Mag.
(1993) - A. Aldroubi, M. Unser (Eds.), Wavelets in Medicine and Biology, CRC Press, Boca Raton, FL, USA,...
- J.P. Antoine, D. Barache, R.M. Cesar, L. da Costa, Multiscale shape analysis using the continuous wavelet transform,...
- K. Berkner, R.O. Wells, A fast approximation to the continuous wavelet transform with applications, in: Record of the...
Digital Image Processing
(1996)- Y.T. Chan, K.C. Ho, Filter design for CWT computation using the Shensa algorithm, in: Proceedings of the IEEE...
- et al.
Shape Analysis and Classification: Theory and Practice
(2001) - H. Goely, R.D. Jones, P.J. Bones, Continuous wavelet transform for the detection and classification of epileptiform...
Fast CWT computation at integer scales by the generalized MRA structure
IEEE Trans. Signal Process.
(February 1998)- K.C. Ho, Y.T. Chan, Optimum filter design for the ‘à trous’ algorithm, in: Proceedings of the IEEE-SP International...