Daubechies Wavelets

Fourier Analysis · Axiom Academy

From discontinuous Haar to smooth wavelets: understanding vanishing moments and compactness 1. The Need for Smooth Wavelets The Haar wavelet's discontinuous jumps are perfect for detecting edges but create unwanted artifacts when analyzing smooth, continuous signals. When we decompose smooth functions using Haar wavelets, the discontinuities introduce high-frequency components that don't represent the original signal. 2. Daubechies Wavelets: DbN (N Vanishing Moments) Daubechies wavelets, denoted DbN, are characterized by their number of vanishing moments N. A wavelet has N vanishing moments if it is orthogonal to all polynomials up to degree N-1. This means the wavelet "filters out" polynomial trends, focusing only on higher-order variations. 3. Tradeoff: Smoothness vs. Compactness The fundamental tradeoff in wavelet design is between smoothness and compact support. Increasing the number of vanishing moments N makes the wavelet smoother but also extends its support (the region where it's non-zero). Db1 is the Haar wavelet (support length 2), while Db10 has support length 20. 4. Comparing Db2, Db4, Db6 Wavelets Let's examine three popular Daubechies wavelets: Db2 (support length 3, good for piecewise linear signals), Db4 (support length 7, good for smooth signals), and Db6 (support length 11, excellent for very smooth signals like audio). As N increases, the wavelets become smoother and more sinusoidal in appearance.

This is the written version of the interactive lesson above. See the full Fourier Analysis course.