Image Segmentation Based On Particle Swarm Optimization And Maximum Entropy Multi-threshold Segmentation Method

Image segmentation is an important part of the target detection and recognition, which is one of the basic and key technology in computer vision. Its purpose is to the part of our interested from the image, and provide the basis for subsequent processing and recognition. There are many methods of image segmentation, including the use of the most common is the threshold segmentation method, which is simple, easy to segmentation. Now we have a pair of complex image of a multi peak distribution, If we want to effectively segment the image, and find the best combination of multi threshold to image segmentation, There will be a large amount of calculation, time-consuming problem. Multithreshold segmentation often can not meet the requirements of real-time application. The accuracy of threshold selection determines the quality of image segmentation. How to search the combination of multi threshold quickly and accurately is the difficulty of the image segmentation technology research. This article will improved particle swarm algorithm, and combined with the maximum entropy method, try to put forward an efficient and accurate algorithm to solve multi threshold segmentation problems. The specific contents are as follows:Quantum particle swarm algorithm inspired by quantum theory, and put forward the Quantum-behaved Particle Swarm Optimization, It comes from the particle swarm algorithm development. It is based on the particle convergence properties of particle swarm algorithm, the different: In the update equation, Quantum-behaved Particle Swarm Optimization algorithm without velocity vector. With less parameters than particle swarm algorithm, easy to implement, and easier to control. It also has a strong global search ability. Quantum-behaved Particle Swarm Optimization is a new intelligent optimization algorithm.1: To make improvements of the particle swarm algorithm, introducing the relative matrix algorithm and expansion model, the particle swarm algorithm combined with the maximum entropy method. First, the objective function of optimization is obtained by maximum entropy method. The improved particle swarm optimization algorithm optimize the objective function, which can search for the best threshold combination, and achieve multi threshold image segmentation.2 In order to improve the efficiency of image segmentation and segmentation accuracy, Put forward the maximum entropy multi threshold segmentation based on the Quantum-behaved Particle Swarm Optimization.First, the objective function of optimization is obtained by maximum entropy method, the Quantum-behaved Particle Swarm Optimization optimize the objective function. Division of pixels by the optimal threshold, Realization the Multi threshold image segmentation.3 The above two algorithms are applied to the multi threshold image segmentation, Experiments show that: The two improved algorithm with good segmentation effect, and they can find the best combination of thresholds quickly and accurately, The two algorithm is suitable for the complex image of multi peak histogram, they have a wide range of applications.