Page 108 - Phase Space Optics Fundamentals and Applications
P. 108

Rotations in Phase Space    89


               of other phase-space rotators for signal processing tasks, which can
               be elegantly expressed in the framework of generalized convolution.
                 The convolution operation between signals f and h, Eq. (3.51), can
               be alternatively expressed via the Fourier transform as 9
                                        −1
                    C f,h (r) = ( f ∗ h)(r) = F  {F[ f (·)](u) F[h(·)](u)} (r)  (3.82)
               By analogy we can introduce the generalized (canonical) convolution
               (GC) operation as 6,8,56
                                                     T 2
                                       T 3
                   GC f,h (T 1 , T 2 , T 3 , r) = R {R [ f (·)](u) R [h(·)](u)}(r)  (3.83)
                                           T 1
                                                                  T
               where the FT operators F are substituted by the CT ones R .Inthe
                                T
                                                       U
               widely used GCs, R corresponds to the RCT R and can be denoted
               by GC f,h (U 1 , U 2 , U 3 , r).
                               6
                 It is easy to see that Eq. (3.83) reduces to common convolution,
               Eq. (3.82), if the ray transformation matrices correspond to the direct/
                                        −1
               inverse FT ones U 1 = U 2 = U  = U f ( /2,  /2). Besides that the GC
                                        3
               includes, as particular cases, the correlation operation

                                !                                      "
                 Cor f,h (r) = GC f,h ∗ U f  ,  , U f  − , −  , U f  − , −  r
                                     2 2          2   2         2   2
                                                                    (3.84)
                                                                     9
               used as a measure of similarity between two signals f and h; the
               fractional convolution
                         GC f,h [U f (  x ,   y ), U f (  x ,   y ), U f (  x ,   y ), r]  (3.85)
               applied for shift-variant filtering and pattern recognition; 4,15  the
               Wigner distribution

                          !
                    2GC f, f  ∗ U f    x +  ,   y +  , U f  −  x +  , −  y +  ,
                                    2      2             2       2

                                         "
                         U f  − , −   , 2                           (3.86)
                               2    2


                                                                2
                                                  2
                                                      2
                                                           2
               expressed in polar coordinates   = (  x + p ,  y + p ); and the
                                                      x
                                                                y
                                                      U
                                                             2
                                            −1
               RCT power spectrum GC f, f (U, U , I, r) =|R [ f ](r)| , correspond-
                                      ∗
               ing to the squared modulus of the RCT of the signal or Wigner distri-
               bution projection, which in the case U = U f (  x ,   y ) is denoted as the
               Radon-Wigner transform (see Chap. 4)
                                                      U f (  x ,  y )  2
                  GC f, f [U f (  x ,   y ), U f (−  x , −  y ), I, r] =|R  [ f ](r)|  (3.87)
                      ∗
                 The   generalized  convolution  GC f,h (U 1 , U 2 , U 3 , r)  of  two-
               dimensional signals f and h is a function of 2 variables (r) and 12
   103   104   105   106   107   108   109   110   111   112   113