Dual-Threshold Image Segmentation Method Based on Parallel Genetic Algorithms

In order to improve the efficiency of image segmentation, an optimal dual-threshold searching method based on parallel genetic algorithm is proposed. Each individual in the population is designed as a vector of one row and two columns, in which each element is encoded to binary and using variance ratio between clusters as its fitness value, and several sub-populations are generated and calculated parallelly. Meanwhile, in the genetic operations, the maximum fitness individual in each population is reproduce into the next generation and the mutation factor is increased properly. Conditioned best -reserve strategy is implemented in the environment of population variety in order to ensure that the algorithm converge to the best solution. Experiment result shows that the algorithm has not only good effect but also higher efficiency for image segmentation, the time cost for the parallel genetic algorithm is only 4.16% of general algorithm.