A novel particle swarm optimisation approach to detecting continuous, thin and smooth edges in noisy images

• We present two PSO-based algorithms for detecting continuous edges in noisy images.
• We compare them with a modified version of Canny and a robust rank order algorithm.
• The proposed algorithms are more robust than Canny and RRO under noisy conditions.
• The performance of the new algorithms are also compared with each other.
• The algorithm equipped with a penalising method is more efficient than the other one.

Detection of continuous edges is a hard problem and most edge detection algorithms produce jagged and thick edges particularly in noisy images. This paper firstly presents a novel constrained optimisation model for detecting continuous, thin and smooth edges in such images. Then two particle swarm optimisation-based algorithms are applied to search for good solutions. These two algorithms utilise two different constraint handling methods: penalising and preservation. The algorithms are examined and compared with a modified version of the Canny algorithm as a Gaussian filter-based edge detector and the robust rank order (RRO)-based algorithm as a statistical-based edge detector on two sets of images with different types and levels of noise. Pratt’s figure of merit as a measure of localisation accuracy is used for the comparison of these algorithms. Experimental results show that the proposed edge detectors are more robust under noisy conditions and their performances are better than the Canny and RRO algorithms for the images corrupted by impulsive and Gaussian noise. The proposed algorithm based on the penalising method is faster than the algorithm using the preservation method to handle the constraints.