Page 320 - Biomedical Engineering and Design Handbook Volume 2, Applications
P. 320

298  DIAGNOSTIC EQUIPMENT DESIGN

                       transfer function O[u, v] (OTF). 22  This is a spatial frequency-dependent complex function with a
                       modulus component called the modulation transfer function M[u, v] (MTF) and a phase component
                       called the phase transfer function Φ[u, v] (PTF). The MTF is the ratio of the image-to-object mod-
                       ulation, while the PTF is a measure of the relative positional shift from object to image.
                         Taking a one-dimensional distribution for simplicity, we have
                                                                   v
                                                       ≡
                                                               []e
                                                 ℑ{ ( )} O [] = Mv  iΦ ()                (10.99)
                                                  hx
                                                          v
                       where M[v] and Φ(v) are MTF and PTF, respectively. It is customary to define a set of normalized
                       transfer functions by dividing by its zero spatial frequency value
                                                          +∞
                                                    O[] = ∫ −∞ h x dx
                                                             (
                                                              )
                                                      0
                       according to
                                                  ℑ {(
                                                   hx)}
                                                                          v
                                                            hx)} =
                                          Ov[] =         =ℑ {(     Mv e i (Φ v)
                                                                     [
                                                                      ]
                                            n     +∞         n      n                   (10.100)
                                                 ∫ −∞ hx dx
                                                    ()
                       where
                                                            hx()
                                                    hx() =  +∞
                                                     n
                                                          ∫ −∞ hx dx
                                                             ()
                       is the normalized line-spread function and
                                                            Mv[]
                                                      Mv[] =
                                                       n
                                                            O[]
                                                              0
                       is the normalized modulation transfer function. If the line-spread function is symmetrical, there is no
                       phase shift so that Im O [v] = 0, M [v] = Re O [v], and Φ(v) = 0. A useful parameter in evaluating
                                                         n
                                        n
                                                n
                       the performance of a system is the contrast or modulation, defined by
                                                       Iv    −  Iv
                                                        ()
                                                               ()
                                                  Γ [] =  max    min
                                                    v
                                                               ()
                                                        ()
                                                       Iv    +  Iv                      (10.101)
                                                          max    min
                       where I(v) max  is the peak of the profile and I(v) min  is the trough of the profile. Since the MTF is the
                       ratio of the image-to-object modulation, we may write, for sinusoids of period l and corresponding
                       frequency v = 1/l the expression
                                                          Γ  v ()
                                                    Mv[] =    image                     (10.102)
                                                      n
                                                          Γ  v () object
                       It is a common practice to use the MTF to quantify the resolution capability of linear systems asso-
                       ciated with image processing. This is because if the individual MTFs are known for the components
                       of a system, the overall MTF is often simply their product. A natural resolution standard would com-
                       prise a continuous series of parallel lines with sine wave intensity profiles of increasing frequency.
                       However, this is difficult to produce in practice and it is usual to adopt a simpler pattern known as
                       the Sayce chart, which has a parallel line, or rectangular, wave grating with a decreasing spatial
                                                                23
                       period d and hence increasing spatial frequency s = 1/d. The rectangular pattern profile L(x) may
                       be expressed in terms of its Fourier series components [Eq. (10.34)], according to
                                               ⎧ ⎪  4     ∞ ∞  () k+1       ⎫ ⎪
                                                            −
                                                             1
                                                       σ
                                                                    π
                                        Lx() =  L ⎨ 1 +  Γ 0 ( ) r∑  cos22 k −1 )σ ⎬    (10.103)
                                                                      (
                                              0
                                               ⎩ ⎪  π    k=1  2 k −1        ⎭ ⎪
   315   316   317   318   319   320   321   322   323   324   325