材料期刊网

为减少肺组织分割算法的运算时间,提出了一种基于粒子群优化的 Otsu 肺组织改进自动分割算法。针对传统粒子群优化的二维 Otsu 算法中二维直方图计算量大、粒子搜索容易陷入局部最优解的缺陷,使用灰度级-梯度二维直方图减少二维直方图的计算量,并减小粒子搜索范围,采用基于粒子空间对称分布的改进粒子群获取最佳阈值。算法实现过程中利用孔洞填充算法去除背景,基于形态学操作去除噪声、修补病变区域产生的孔洞。仿真实验结果显示,本文算法对图像尺寸为512像素×512像素 CT 图像的阈值分割时间约为0.2 s,比基于灰度级-邻域均值二维直方图的粒子群优化的 Otsu 算法的阈值分割速度提高了约16%。较好地实现了胸腔 CT 图像的肺组织自动分割,与传统算法相比较,本算法在保证分割精度的基础上分割速度明显提升。

This paper develops an automatic segmentation method of lung image using two dimensional Otsu method based on particle swarm optimization in order to reduce operation time.Aiming at the shortage of the traditional two dimensional Otsu based on particle swarm optimization,which the cal-culating quantity is large and standard particle swarm algorithm is easy to fall into local optimum,an improved lung segmentation based on 2D Otsu optimized by PSO is proposed in this paper.Using the grayscale-gradient two-dimensional histogram,not only reduces the amount of histogram’s calcula-tion,but also narrows the searching area of particles.The algorithm uses the improved PSO which based on diversity of particle symmetrical distribution to search optimal threshold.In the process of algorithm,the region filling algorithm is used to remove background in order to make the threshold segmentation of lung better,and the morphology operations to remove noise and repair holes which in the target image.The threshold segmenting time in this algorithm is about 0.2 s,increased about 1 6% than the speed of the traditional Otsu threshold segmentation optimized by PSO,the size of the CT images is 5 12 × 5 12 pixels in this experiment.The segmentation algorithm in this paper can seg-ment the lung in CT image automatically,not only ensures the accuracy of the segmentation,but also improves the speed of the segmentation in comparison with conventional algorithms.