多元纹理图像构造方法及应用

Construction method and application of multivariate texture image

中文关键词  纹理特征  多元图像分析  图像分割  计算机视觉

英文关键词  texture feature  multivariate image analysis  image segmentation  computer vision

中文摘要

      多元图像分析方法在图像展开阶段会丢失像素间空间关联关系, 导致其对图像纹理特征的分析能力不足. 针对此问题, 本文提出了一种基于图像纹理特征的多元图像构造方法, 并应用于图像分割. 首先, 结合滑动窗口法 和灰度共生矩阵求取图像各通道的纹理特征影像, 叠加纹理特征影像构造多元图像. 然后, 应用多元图像分析方法 对所得多元图像进行分析, 分割出感兴趣区域. 最后, 利用分割结果构造决策树模型, 以完成对同类感兴趣区域的分 割. 在图像数据集上进行仿真实验, 实验结果表明, 本文所提方法的均交并比(MIoU)与同类方法相比有10%左右的 提升.

英文摘要

      In the image expansion stage, multivariate image analysis method may result in the loss of spatial correlation between pixels, which limits its application in the analysis of image texture features. To address this problem, an image texture features based multivariate images constructing method is proposed and applied to image segmentation in this work. Firstly, the texture feature images of each channel of the image are obtained by combining the sliding window method and the gray level co-occurrence matrix. Then, the multi-image analysis method is used to analyze the multi-image and segment the region of interest. Finally, based on the segmentation results, a decision tree model is constructed to segment the regions of interest of the same class. Simulation experiments on image datasets are carried out, the results show that the mean intersection over union (MIoU) of the proposed method is about 10% higher than that of similar methods.