Surrounded by transforms
Published:
An integral transform, or discrete transform, is a change of basis: an operation that re-expresses a function or signal defined on a “spatial” or “time” domain in terms of a different set of basis functions.
The purpose of such a transform is typically to diagonalize a linear operator. Operations that are difficult to work with in the original domain — differentiation, scaling, cyclic shifting, convolution — become pointwise multiplication in the transform domain. For a kernel $K(t, u)$, the transform $\mathcal{T}$ of a function $f$ is defined as
\[(\mathcal{T}f)(u) = \int_{t_1}^{t_2} f(t) K(t, u) \, dt\]In the discrete case, the integral is replaced by a summation and the kernel by a transformation matrix.
Criteria for “Fourier-Type” Transforms
A transform belongs to the Fourier-related family if it is rooted in Abstract Harmonic Analysis — that is, if it acts as a projection onto the character space of an underlying group.
Four algebraic properties define membership in this family:
- Group Domain ($G$): The input space forms a topological group under a symmetry operation (e.g., addition in $\mathbb{R}$, multiplication in $\mathbb{R}^+$, or modular arithmetic in $\mathbb{Z}_N$).
- Dual Group of Characters ($\hat{G}$): The kernels are group characters — homomorphisms mapping the group operation to multiplication on the complex unit circle.
- The Convolution Theorem: Group convolution $f * g$ maps to pointwise multiplication: $\mathcal{F}(f * g) = \mathcal{F}(f) \cdot \mathcal{F}(g)$.
- Plancherel Isometry: The transform preserves inner products and energy (the $L^2$ norm) between the spatial and frequency domains.
Properties 2 and 3 are stated here for the abelian case, which covers most entries below. For non-abelian groups (e.g. $S_n$, $SO(3)$, $SU(2)$), both properties generalize rather than fail: characters are replaced by the matrix coefficients of irreducible unitary representations, and the convolution theorem becomes pointwise matrix multiplication at each irrep rather than scalar multiplication. This is the form under which the Finite Non-Abelian Fourier and Peter-Weyl transforms remain part of the same family.
Timeline of Development and Family Tree
The table below indexes major continuous, discrete, geometric, and non-abelian transforms in chronological order of development.
| Year | Transform | Classification Type | Domain / Group ($G$) | Operator Diagonalized | Primary Application / Field |
|---|---|---|---|---|---|
| 1782 | Laplace Transform | Continuous | Non-negative reals $([0, \infty), +)$ | Differentiation operator $\frac{d}{dt}$ | Control theory, ODE stability, circuit analysis |
| 1782 | Spherical Harmonic Transform | Geometric / Spatial | 2-Sphere $S^2$ / $SO(3)$ | Angular Laplacian $\nabla_{S^2}^2$ | Planetary gravity, quantum orbitals, 3D graphics |
| 1805 | Discrete Fourier Transform (DFT) | Discrete / Finite | Cyclic group $(\mathbb{Z}_N, +)$ | Cyclic shift matrix | Asteroid orbit interpolation, digital signal processing |
| 1807 | Continuous Fourier Transform | Continuous | Real line $(\mathbb{R}, +)$ | Differentiation operator $\frac{d}{dt}$ | Wave mechanics, heat diffusion, quantum physics |
| 1869 | Hankel Transform | Geometric / Radial | Radial Euclidean space $(\mathbb{R}^n)$ | Radial Laplacian $\nabla_r^2$ | Cylindrical wave acoustics, optics, fluid dynamics |
| 1893 | Hadamard Transform | Discrete / Combinatorial | Sign matrix group ${+1,-1}^{N \times N}$ | Recursive orthogonal sign-flip basis (Sylvester construction) | Error-correcting codes, orthogonal design theory |
| 1896 | Fourier Transform on Finite Groups | Non-Abelian / Discrete | Finite groups ($S_n, D_n, GL_2(\mathbb{F}_q)$) | Left regular representation | Finite group representation theory, Diaconis card shuffling |
| 1896 | Mellin Transform | Continuous / Multiplicative | Multiplicative reals $(\mathbb{R}^+, \times)$ | Dilation operator $x \frac{d}{dx}$ | Analytic number theory, Dirichlet series, Riemann $\zeta$ |
| 1917 | Radon Transform | Geometric / Spatial | Hyperplane lines in $\mathbb{R}^n$ | Integral projection geometry | CT / PET medical tomography, seismic imaging |
| 1923 | Walsh Transform | Discrete / Quantum-adjacent | Hypercube $(\mathbb{Z}_2)^n$ | Bitwise XOR shift | Spread-spectrum coding, Simon’s quantum algorithm |
| 1927 | Peter-Weyl Transform | Non-Abelian / Continuous | Compact Lie groups ($SO(3), SU(2)$) | Lie algebra differential operators | Quantum angular momentum, particle physics |
| 1942 | Hartley Transform | Continuous / Real-Valued | Real line $(\mathbb{R}, +)$ | Shift operator on real fields | Real-valued signal processing without complex arithmetic |
| 1947 | Z-Transform | Discrete-Time | Discrete grid $\mathbb{Z}$ | Discrete linear time-shift | Digital filter design (IIR/FIR), sampled data control |
| 1965 | Fast Fourier Transform (FFT) | Fast Algorithm | Cyclic group $(\mathbb{Z}_N, +)$ | Fast $O(N \log N)$ matrix factorization | Modern digital telecom, spectral analysis, fast convolution |
| 1969 | Chirp Z-Transform (CZT) | Discrete / Complex | Spiral contours in $\mathbb{C}$ | Non-uniform Z-plane evaluation | Zoomed spectrum analysis, radar signal processing |
| 1971 | Number-Theoretic Transform (NTT) | Finite Field / Discrete | Galois fields $\mathbb{F}_q^\times$ | Modular polynomial shift | Exact integer convolution, post-quantum cryptography |
| 1980 | Fractional Fourier Transform (FrFT) | Continuous / Phase Space | Time-frequency plane | Quantum Harmonic Oscillator | Chirped optical signals, radar filtering |
| 1994 | Quantum Fourier Transform (QFT) | Discrete / Quantum | Cyclic group $(\mathbb{Z}_{2^n}, +)$ | Quantum phase shift gates | Shor’s factoring algorithm, quantum computation |
| 1998 | Higher-Order Fourier Analysis | Non-Linear / Combinatorial | Integers / Finite fields | Gowers Uniformity Norms ($e^{2\pi i P(n)}$) | Additive combinatorics, Green-Tao prime theorem |