For those of you wondering about the paper, its:
Rectangle Detection based on aWindowed Hough Transform
by Cl´audio Rosito Jung and Rodrigo Schramm (don't know how to do the ´ over the a in this text field).
Now according to the paper, the intersection points are expressed as polar coordinates, obviously you implementation maybe different (the only way to tell is to show us your code).
Assuming you are being consistent with his notation, your peaks should be express as:
H_1 = (\rho_1,\theta_1)
H_2 = (\rho_2,\theta_2)
...
H_m = (\rho_m,\theta_m)
you then must preform peak paring given by equation (3) in section 4.3 or
\delta\theta = |\theta_i - \theta_j| < T_\theta
\delta\rho= |\rho_i- \rho_j | < T_\rho
|C(\rho_i ,\theta_i ) - C(\rho_j ,\theta_j )| < T_L \frac{C(\rho_i ,\theta_i )+C(\rho_j ,\theta_j )}{2}
where T_\theta
represents the angular threshold corresponding to parallel lines
and T_L
is the normalized threshold corresponding to lines of similar length
Note: I used LaTeX notation for clarity