An enhanced multilevel thresholding Image segmentation using a novel Hybrid metaheuristic approach A thesis

Abstract

Multilevel thresholding plays a vital role in image segmentation newlineapplications. It establishes multiple thresholds for a given image and newlinesegments it into distinct brightness regions corresponding to a single newlinebackground and multiple objects. However, as the number of threshold levels newlinegrows significantly, the computational cost of traditional multilevel newlinethresholding increases exponentially, necessitating the use of metaheuristic newlineoptimisation to determine the optimal number of threshold levels. newlineNonetheless, a single metaheuristic algorithm often has some drawbacks, such newlineas low accuracy, poor generalisation efficiency and poor local optimisation newlinecapability, when searching for optimum threshold levels in a search area. newlineIn recent years, there has been a significant increase in interest in newlinehybrid metaheuristics in the field of Multilevel thresholding image newlinesegmentation. The hybrid algorithm combines the advantages of two newlinealgorithms and balances local exploitation and global exploration when newlinesearching the optimal solution, which decreases the probability of being newlinetrapped in local optima and improves search accuracy. Therefore, this research newlinework proposes an efficient and novel hybrid algorithm called WCMFO for an newlineenhanced multilevel thresholding image segmentation, which is based on the newlinehybridisation of water cycle and moth-flame optimisation algorithms. The newlineWCA has enhanced its capability for solution space exploration by the use of newlineits streams and river formations and MFO has a great potential to exploit newlinebecause of localised flames. On the basis of that, the stream positions are then newlineupdated according to the spiral movement produced by the MFO algorithm, newlineand WCMFO offers an improved raining process that utilises levy flight to newlineincrease randomisation newline