General test framework for straight-line detection by Hough transforms

The design and implementation of a general software framework is described for testing line-detection algorithms on digital images. This test framework is applied to compare the reliability and accuracy of different Hough transform algorithms, including a new hybrid. The software test-bed generates digital images containing lines that are drawn pseudorandomly according to parameters defined at run-time. A line-detection algorithm is applied to each simulated edge-image, and the parameters of the detected lines are compared with those of the known inputs. The errors in parametrization are evaluated for detected lines, and the numbers of undetected lines and false alarms are recorded. The Standard Hough Transform (SHT) is compared with variants that use different parametrizations and input-data sampling procedures, including the Counter-Based Hough Transform (CBHT), the Randomized Hough Transform (FIHT2), and an improved RHT developed as a result of these trials, the `rhotheta parametrized RHT´ (rtRHT)