Efficient and improved edge detection via a hysteresis thresholding method

Abstract

Hysteresis thresholding is a popular technique for automatic edge detection. However, calculating reasonably high and low thresholds using an unsupervized method remains an issue. Conventional low and high threshold-linking methods sometimes produce noisy edges and fail to detect some obvious edges. Here, a novel edge detection algorithm is proposed that provides efficient calculation of thresholds, and links edge maps for extraction of the final edge map at low complexity. The proposed method suppresses unwanted noisy edges while efficiently preserving obvious edges. The simulation results show that the proposed method provides better results in terms of performance and computation time. Thus, it can be applied to any feature image analysed by an edge detector.