Fourier duality in integral geometry and reconstruction from ray integrals (Q485271): Difference between revisions

Given a \(k\)-dimensional variety \(\Delta\) of affine \((k-1)\)-dimensional subspaces of \(E^n\backslash 0\), the author assumes the variety \(\Sigma_k(\Delta)\) of all \(k\)-planes \(A\) that contains a \((k-1)\)-plane \(B\in \Delta\) is a complete family in \(E^n\backslash 0\). Thus, given any hyperplane \(H\) in \(E^n\backslash 0\) there exists an element of \(A\in \Sigma_k(\Delta)\) lying in \(H\), and consequently the corresponding polar planes satisfy \(H^0\subset A^0\). The polar \(A^0\) is the \((n-k-1)\)-dimensional subspace of \(E^n\backslash 0\) that is the set of solutions of the system \(\langle x,y\rangle+1=0\), \(\forall y\in A\). In Theorem 7, the author considers a function \(f\) on \(E^n\) such that \((1+ |x|^n)f(x)\in L^1(E^n)\), and its \(k\)-polar function, \(f^0(\xi)=|(1,\xi)|^{k-n-1}|(1,x)|^k \tilde{g}(1, \xi)\), where \(\tilde{g}\) is the Fourier transform of the homogeneous distribution \(g (dx_0\wedge dx)\) of degree \(k\), with \(g\) given by \(g(x_0,x)=|x_0|^{-n-1}|(x_0,x)|^kf(x/x_0)\), \(g(0,x)=0\). From the integral data of the Radon integral transformation of \(f\) on \(A\in \Sigma_k(\Delta)\) he concludes first that \(f^0\) can be recovered in all \(E^n\backslash 0\). He then shows that \(f\) can be reconstructed from its polar function via inverse Fourier transformation. In Theorem 6, a formula is given relating the Radon integral transformations \(Rf^0(A^0)\) and \(Rf(A)\) for such function \(f\) and any \(k\)-plane \(A\), \(0<k<n\), involving the distances functions of \(A\) and \(A^0\) to the zero point. Many results of the author's previous work [Reconstructive integral geometry. Basel: Birkhäuser (2004; Zbl 1063.44002)] are used in the proofs. If \(\Delta\) is a closed non-contractible curve \(\Gamma\) in the projective closure \(P^n\) of \(E^n\), in Theorem 8 the author concludes that any smooth function \(f\) on \(E\) with compact support can be reconstructed from the line transformation \(Rf(L)=\int_L f dV(L)\) of \(f\) for lines \(L\) that meet \(\Gamma\). In the case \(\Delta\) is a piecewise \(C^2\) curve \(\Gamma=y(s)\) in \(E^3\) satisfying the completeness condition at a given point \(x\in E^3\backslash \Gamma\), given a \(C^2\) function \(f\) on \(E^3\) with compact support away from \(\Gamma\), the author gives in Theorem 10 a reconstruction of \(f\) at \(x\) from a formula involving the ray transform \(g\) of \(f\), and the Fourier transform \(G\) of \(g(y,v)dv\), and a fixed bounded distribution function \(\epsilon(y,\xi)\), with \(y\in \Gamma, \xi\in S^2\). If \(f\) is a continuous function in \(E^2\backslash B\) such that \(|x|^{\delta} f(x)\) is bounded for each \(\delta \geq 0\) and for any line \(L\) not intersecting a fixed compact set \(B\), in the last Theorem 12 it is proved that the Fourier expansion of \(f\) restricted to a circle \(S\) enclosing \(B\) does not contain frequencies \(<\delta\). The proof uses a representation of \(f\) in sums of harmonics. This result implies Helgason's support theorem in dimension 2.