Otsu Multi-Threshold Image Segmentation Algorithm Based on Improved Particle Swarm Optimization

The Maximum Inter-Class Variance Method (Otsu) is a commonly used threshold segmentation method in image segmentation. It has a significant effect for single-threshold segmentation, but for multi-threshold segmentation, the computational complexity is large. To reduce complexity, further optimization techniques can be employed to find the optimal multi-level threshold. This paper proposes an improved multi-threshold segmentation algorithm based on improved particle swarm optimization (IPSO) for finding the optimal threshold. The general particle swarm optimization algorithm has two major problems: dimension disaster and easy to fall into local optimum. Improved particle swarm optimization decomposes high-dimensional groups into multiple one-dimensional groups. These one-dimensional groups exchange information with each other to generate overall fitness values, and then in each one-dimensional group, particles with fitness values smaller than the average fitness of the entire population. Wavelet variability is performed to prevent particles from falling into local optimum. Finally, a simulation experiment is carried out to compare the results with the existing particle swarm optimization maximum interclass variance algorithm. Experimental results show that the method has a faster convergence speed and a better Otsu threshold.