| In the field of image segmentation, the image segmentation method based on threshold is the most widely used. The threshold segmentation method can be divided into single threshold segmentation and multi-threshold segmentation. The multi-threshold segmentation method is derived from the single threshold segmentation method. Relatively speaking, single threshold segmentation method is more mature,but it is susceptible to noise interference. Meanwhile the multi-threshold segmentation method's main task focus on how to reduce the time consumption and improve image segmentation quality. It is popular to introduce the swarm intelligence algorithm into the image segmentation domain, and the traditional image segmentation algorithm is used as objective function, so that the problem of segmentation threshold is transformed into the global optimal solution problem of the objective function. Fireworks Algorithm (referred to FWA) has a unique explosion search mechanism, and it can simulate the fireworks explosion method to solve the solution space of the problem through multi point simultaneous explosion optimization, which makes it a new intelligent optimization algorithm that is different from other swarm intelligence algorithm. Considering its excellent global convergence ability to solve complex optimization problems, we introduced it into the field of image segmentation to improve the efficiency.The fireworks algorithm is a new intelligent algorithm proposed by the inspiration of fireworks explosion in the air. The algorithm can adjust the control parameters (spark number and explosion range) according to the fitness of the fireworks, and ensure the diversity of the population and fast convergence, which can easily get rid of the local optimal solution. This performance provides a new and effective approach for solving the noise sensitivity problem, and improving time efficiency which caused by threshold segmentation. In this paper, we mainly introduce the fireworks algorithm to solve the problem of threshold segmentation. The main works are as follows:Firstly, in order to solve the problem of noise interference, high time complexity and poor segmentation effect caused by the basic threshold segmentation method, an image segmentation algorithm based on fireworks algorithm is proposed,which is respectively combined with the maximum inter-variance method and maximum entropy method, and this method fully exert the fireworks algorithm's advantages of quick convergence and the ability of jump out of the local optimum to obtain the optimal segmentation threshold.By comparing the experiment results between single threshold and multi - threshold segmentation, it is verified that the algorithm has a good effect on the above problem in threshold segmentation.Secondly, an improved fireworks algorithm based on position offset is proposed for the sensitive problem of the basic fireworks algorithm in solving the objective function.The algorithm is no longer limitedto the function that the most optimal location is in the origin. Therefore, the application of the method can effectively shorten the optimization time, and improve the segmentation effect for the situations that the image segmentation threshold is usually not in the origin. The comparison experiment between improved and traditional fireworks algorithm shows that the former one can quickly find out the optimal threshold of images.Thirdly, in order to improve the accuracy of multi-threshold segmentation, a multi-threshold image segmentation method for fireworks algorithm based on memetic algorithm, which uses both the mechanism of global optimization and local optimization,is put forward. This method that use genetic algorithm as the global search algorithm can improves the global optimization effect of fireworks algorithm, therefor it strengthens the quality of image segmentation.Through the comparative analysis of the experiment demonstrates that the method we proposed can achieve better effect for multi-threshold image segmentation. |
PDF Full Text Request