Abstract: Markov Random Field (MRF) model is a classical image texture analysis method, which has been widely used in image texture simulation and segmentation. However, when the classical simulated annealing algorithm is used to compute a global optimal solution, it cannot satisfy the performance requirements of texture analysis and processing due to less efficiency. A fitness function which could determine texture class was designed, and the use of Particle Swarm Optimization (PSO) was proposed to calculate the optimal solution of it. Based on the proposed method, texture segmentation of remote sensing images was carried out. Compared with the simulated annealing algorithm, the experimental results show that PSO algorithm can reduce the computational complexity and it is a more effectively optimal method of image segmentation.

Key words: Markov Random Field (MRF), Gibbs distribution, Particle Swarm Optimization (PSO) algorithm, texture segmentation

摘要： 影像纹理的马尔可夫随机场(MRF)模型是一种分析纹理较为经典的方法，已被广泛用于影像纹理的模拟和分割。由于传统的模拟退火算法在计算全局最优解时，处理效率较低，无法满足纹理分析与处理的性能要求。设计了一种判定纹理类别的适应度函数，提出了利用粒子群优化算法计算适应度函数的最优解，应用该算法对遥感影像数据进行了纹理分割实验。实验结果表明，该算法与模拟退火算法比较，具有寻优速度快的优点，是一种有效的图像分割优化方法。

关键词: 马尔可夫随机场, 吉布斯分布, 粒子群优化算法, 纹理分割