An algorithm is presented which replaces the pixel-mapping operation in the Hough transform. Instead of performing the complete mapping, it estimates a range of parameter space and only performs the mapping in this range. It is somewhat able to adapt to image noise, and in high noise cases reverts to the full Hough transform with little additional overhead. The expression to estimate the range was generated with immune programming, and the performance of various clonal selection algorithms was compared and analysed on its components. |