What Is Wavelet Analysis?

Wavelet analysis is a mathematical framework for representing functions and signals in terms of localized wave-like basis functions called wavelets. Unlike the Fourier transform, which decomposes a signal into sinusoids that extend infinitely in time, wavelet analysis uses functions that are compact in both time and frequency. This dual localization allows wavelet analysis to characterize signals that are non-stationary, that is, signals whose frequency content changes over time, which makes it well-suited to a broad range of applications in signal processing, image analysis, and scientific computing. It draws its theoretical foundations from functional analysis, harmonic analysis, and filter bank theory.

Wavelet analysis was developed in its modern form during the 1980s, with contributions from Yves Meyer, Stéphane Mallat, Ingrid Daubechies, and others working across mathematics and engineering. Daubechies constructed the compactly supported orthonormal wavelet families that bear her name and remain widely used today. The field unified several pre-existing techniques, including subband coding in signal processing and the continuous wavelet transform used in geophysics, under a common mathematical structure.

The Wavelet Transform

The continuous wavelet transform (CWT) computes the inner product of a signal with a family of scaled and translated copies of a mother wavelet function. The scale parameter controls the width of the wavelet, effectively acting as an inverse-frequency measure: large scales correspond to low-frequency features spread over long intervals, while small scales capture high-frequency transients. The result is a two-dimensional time-scale representation of the signal that reveals when particular frequency components occur, which a power spectrum cannot. The CWT is computed by integration and is used primarily as an analysis tool rather than for compression or reconstruction.

The discrete wavelet transform (DWT), introduced in its multiresolution framework by Mallat in a foundational 1989 paper in IEEE Transactions on Pattern Analysis and Machine Intelligence, computes a dyadic (octave-scale) decomposition of the signal into approximation and detail coefficients using pairs of complementary low-pass and high-pass filters followed by downsampling by two. This filter bank structure makes the DWT computationally efficient: the full transform of an N-point signal requires O(N) operations, compared to O(N log N) for the fast Fourier transform. The DWT is the foundation of most practical wavelet-based signal and image processing systems.

Multiresolution Analysis

Multiresolution analysis (MRA) is the mathematical framework, introduced by Meyer and developed by Mallat, that gives wavelet analysis its theoretical coherence. An MRA consists of a nested sequence of approximation spaces, each capturing the signal at a particular level of resolution, together with complementary detail spaces that contain the information lost when moving from a finer to a coarser scale. Wavelets are the basis functions of these detail spaces. Each level of the MRA corresponds to one stage of the filter bank decomposition, so the MRA framework links the abstract functional analysis of wavelet theory directly to the practical implementation as a cascaded digital filter. An overview of wavelet-based multiresolution analysis from SIAM Review provides a mathematically accessible account of the MRA structure and its relationship to Daubechies wavelets and biorthogonal wavelet families.

Wavelet packets, an extension of the basic MRA, apply the decomposition recursively to both the approximation and detail subbands at each level, producing a more flexible time-frequency tiling that can be adapted to the structure of a particular signal class. A detailed treatment of wavelet families, filter design, and computational efficiency appears in the UC Irvine survey of wavelet analysis and its engineering applications, which covers both the mathematical foundations and practical implementation considerations for discrete wavelet systems.

Applications

Wavelet analysis has applications in a range of fields, including:

• Image compression, where wavelet-based coding forms the basis of the JPEG 2000 standard, offering superior compression ratios and graceful degradation compared to block-DCT methods
• Signal denoising in biomedical engineering, where wavelet thresholding removes noise from electrocardiogram and electroencephalogram recordings while preserving clinically important transients
• Geophysical data processing, where the continuous wavelet transform identifies reflection events and fault structures in seismic recordings
• Numerical analysis, where wavelet bases provide efficient representations for solving partial differential equations with localized features
• Feature extraction in machine learning, where wavelet coefficients serve as discriminative inputs for classification tasks in audio and image domains