Tue Jul 1 11:12:56 CEST 2008

simplified HT for direction estimation

Essentially, a Hough transform H(r,t) of an image I(x,y) can be seen
as a likelihood function for the parameters r and t.

Now observe a typical 'sinogram', like the one here:

Given the Hough transform of a single line, take an r-spectrum for a
certain angle: it will will have less variance when the angle is close
to the correct one. (The integral of the r-spectrum sums to the
number of line-points that voted.)

It would be interesting to be able to ignore the fact that there are
multiple lines, and have them all contribute to a single parameter. In
fact, the first harmonic in the DFT of the r-spectrum will give an
estimate for the phase and offset.

Let's try this.