Chord to Triangular Arm Angle (CTAA), a more accurate version of the CTAR corner detector

Corner detection is a vital process in many image processing and computer vision applications. Amongst all the previously published contour-based detectors described in this paper, Chord to Triangular Arms Ratio (CTAR) has been reported as the best corner detector in term of average repeatability. In this paper, we propose an alternate way of using the triangular theory originally proposed by CTAR. We demonstrate through experimental results that using the corner angle, estimated using triangular theory, is a better measure of curvature than the ratio of arm lengths use by CTAR. We compare the performance of our proposed corner detector with CTAR and CPDA. The results demonstrate that our proposed change is simple yet effective at detecting corners more reliably and accurately.