摘要： 常用的阈值分割方法在对红外图像进行分割时，由于红外图像本身的特点，会出现准确性不高的问题。为此，提出一种基于灰 度-平均灰度-梯度直方图的三维Otsu法，使用稀疏矩阵和改进的粒子群优化(PSO)算法，并在标准PSO算法中加入判断早熟停滞的因子，进行最佳阈值的选取，以提高运算速度，保证准确度。实验结果表明，该方法能够快速有效地对红外图像进行分割。

关键词: 红外图像分割, 阈值选取, 三维Otsu法, 稀疏矩阵, 粒子群优化

Abstract: In view of the characteristics of infrared image, the veracity is not good when the commonly used threshold is used in infrared image segmentation. A 3D Otsu method based on gray level-average gray level-gradient histogram is proposed. The sparse matrix and improved Particle Swarm Optimization(PSO) algorithm are used to search for the segmentation threshold. An effective method that identifies premature stagnation is embedded to PSO. The operation velocity is improved and the veracity is better. Experimental results demonstrate the algorithm can segment infrared images quickly and effectively.

Key words: infrared image segmentation, threshold selection, 3D Otsu method, sparse matrix, Particle Swarm Optimization(PSO)

中图分类号: 

• N945