摘要
基于肺部CT图像灰度不均匀、纹理变化大的特点,文章提出一种超像素与随机森林相结合的肺部CT图像分割算法。该算法首先采用阈值和形态滤波的方法对图像进行预处理;再通过TurboPixels算法将图像分割为超像素;然后运用灰度共生矩阵提取超像素的纹理特征,并融合灰度特征形成特征矩阵;最后基于特征矩阵和随机森林算法获取分割图像。实验结果表明,该文提出的分割算法对肺部CT图像处理具有一定的有效性,健康肺部图像的分割准确率为98.07%,病变图像的准确率为96.23%,且该算法具有全自动、高准确率、鲁棒性好等特点。
Based on the characteristics of intensity inhomogeneity and texture changes in the lung CT images, a novel segmentation method for lung CT images with the combination of super pixels and random forest is proposed. Firstly, the threshold and morphological filtering methods are used to pre- process the image. Then the image is divided into uniform super pixels by using the TurboPixels method. The gray level co-occurrence matrix is used to extract the texture features of the super pix- els, and the feature matrix is formed by fusing the gray features. Finally, the segmentation image is obtained based on the feature matrix and random forest algorithm. The experimental results indicate that the proposed segmentation algorithm shows a promising performance when processing lung CT image. The segmentation accuracy rate of healthy and diseased lung images is 98. 07% and 96.23% respectively, and the algorithm has the characteristics of automatic, high accuracy, good robustness and so on.
作者
楚陪陪
魏本征
曲彦
杨凯
尹义龙
CHU Peipei WEI Benzheng QU Yan YANG Kai YIN Yilong(College of Science and Technology, Shandong University of Traditional Chinese Medicine, Jinan 250355, China School of Comput- er Science and Technology, Shandong University, Jinan 250101, China)
出处
《合肥工业大学学报（自然科学版）》
CAS
北大核心
2017年第3期332-339,共8页
Journal of Hefei University of Technology：Natural Science
基金
国家自然科学基金资助项目(61572300)
国家自然科学基金-广东省联合基金重点资助项目(U1201258)
山东省自然科学基金资助项目(ZR2015FM010)
山东高校科技计划资助项目(J15LN20)
关键词
CT图像
超像素
灰度共生矩阵
纹理特征
随机森林
CT image
super pixel
gray level co-occurrence matrix
texture feature
random forest
