next up previous contents
Next: Fast backprojection Up: The filtered backprojection algorithm Previous: Standard fan beam geometry

Linear fan beam geometry

Here, the detector positions within a fan with vertex b are evenly spaced on the line perpendicular to b. We need the explicit form of the inversion formula mainly for the derivation of the FDK algorithm in 3D cone beam tomography in the next section. With g the function in (2.8), the sampled data is

eqnarray1665

The coordinates tex2html_wrap_inline3460 , y are related to the parallel coordinates tex2html_wrap_inline3302 , s in the representation tex2html_wrap_inline3652 of the rays by

  equation1670

Hence,

displaymath3654

Substituting tex2html_wrap_inline3460 , y for tex2html_wrap_inline3302 , s in (2.2) leads to

displaymath3664

where (2.12) has to be inserted for tex2html_wrap_inline3302 , s. As in the standard fan beam case a direct implementation of this formula results in an algorithm whose complexity is not competitive. Again we can circumvent this problem by exploiting the homogeneity properties of v. A lengthy but elementary computation shows that

displaymath3672

displaymath3674

From (2.11) it follows that

displaymath3676

yielding

eqnarray1694

As in the standard fan beam case we make the approximation tex2html_wrap_inline3678 . Again this is justified if tex2html_wrap_inline3576 , e.g. tex2html_wrap_inline3682 . Then,

displaymath3684

where tex2html_wrap_inline3686 . Defining

  equation1717

this can be written as

  equation1723

The implementation of (2.13), (2.14) can now be done exactly as in the standard case, leading to a filtered backprojection algorithm which needs O(p) operations for each reconstruction point x.



Frank Wuebbeling
Thu Sep 10 10:51:17 MET DST 1998