Linear canonical ripplet transform: theory and localization operators (Q2674421)

The authors study a new version of the ripplet transform for analyzing two-dimensional signals where the Fourier transform is replaced by the more general linear canonical transform. The two-dimensional linear canonical transform (LCT) is defined by \[ \mathcal L_{M}[f](\xi)= \int_{\mathbb R^2}\frac 1{2\pi B}\exp\left(\frac{i(A|\mathfrak t|^2-2\mathfrak t^{T}\xi +D|\xi|^2)}{2B}\right ) f(\mathfrak t)\, d\mathfrak t, \] where \(M=[A,B;C,D]\) is a real unimodular matrix and \(B\neq 0\) (and \(\mathcal L_{M}[f](\xi)=\sqrt{D}\exp(iCD|\xi|^2/2)f(D\xi)\) when \(B=0\)) so that one gets the Fourier-transform with \(M=[0,1;-1,0]\). A basic waveform \(\Upsilon_{a,\mathfrak 0,0}\) of the LCT-based ripplet transform is defined by \( \mathcal L_{M}[\Upsilon_{a,\mathfrak 0,0}](r,\omega) = \frac{a^{(1+d)/2d}}{\sqrt c}W(ar)V\left( \frac{a^{1/d}\omega}{ca}\right), \) in polar coordinates where \(W\) and \(V\) are radial and angular windows functions, respectively, with compact support that satisfy certain admissibility conditions. By letting \(\Upsilon^M_{a,\mathfrak 0,0}(\mathfrak t)=\exp(iA|\mathfrak t|^2/2B)\Upsilon_{a,\mathfrak 0,0}(\mathfrak t)\) this novel waveform can be expressed with the aid of the Fourier transform as \[ \mathcal F[\Upsilon^M_{a,\mathfrak 0,0}](r,\omega)= \frac{a^{(1+d)/2d}B}{\sqrt c} \exp\left(-\frac{iDBr^2}2\right ) W(aBr)V\left( \frac{a^{1/d}\omega}{ca}\right), \] and the novel family of linear canonical ripplets \(\Upsilon^M_{a, \mathfrak{b},\theta}\) is obtained by translations and rotations as \[ \Upsilon^M_{a, \mathfrak{b},\theta}(\mathfrak t) =\Upsilon^M_{a,\mathfrak 0,0}(R_\theta(\mathfrak t - \mathfrak b)). \] The ripplet transform of a function \(f\in L^2(\mathbb R^2)\) is \[\mathcal R^M_\Upsilon f](a,\mathfrak b ,\theta)= \int_{\mathbb R^2} f(\mathfrak t) \overline{\Upsilon^M_{a, \mathfrak{b},\theta}(\mathfrak t)} \, d\mathfrak t. \] The authors derive a reconstruction formula for this ripplet transform and (for functions \(f\) with \(\mathcal F[f](\xi)=0\), \(|\xi|< 2/a_0B\)) a version of Rayleigh's energy theorem and a Heisenberg-type uncertainty inequality. In addition the authors study a localization operator \(\mathcal L^M_\sigma\) defined on functions \(f\in L^2(\mathbb R^2)\) with \(\mathcal F[f](\xi)=0\), \(|\xi|< 2/a_0B\) by \[ \left\langle \mathcal L^M_\sigma f,g\right \rangle_2 = \int_{\mathbb X} \sigma (a,\mathfrak b,\theta) \left\langle f, \Upsilon_{a,\mathfrak b,\theta}\right \rangle_2 \left\langle \Upsilon_{a,\mathfrak b , \theta},g\right \rangle_2 d\mu, \] for all \(g \in L^2(\mathbb R^2)\) where \(\mathbb X= [0,2\pi)\times \mathbb R^2 \times (0,a_0)\) and \(d\mu =da\, d\mathfrak b \, d\theta/a^3\).