基于纹理结构相似性和边缘信息的水平集纹理图像分割  

Level set method based on texture similarity and edge information for segmenting texture images

作　　者：邓星涛 闵海[1] DENG Xingtao;MIN Hai(School of Computer Science and Information Engineering, Hefei University of Technology, Hefei 230601, China)

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230601

出　　处：《合肥工业大学学报（自然科学版）》2022年第1期30-38,124,共10页Journal of Hefei University of Technology：Natural Science

基　　金：国家自然科学基金资助项目(61702154)。

摘　　要：通过提取纹理特征对纹理图像进行分割是一类重要的纹理分割方法。传统的纹理特征提取通常根据像素邻域的灰度统计来获取,因为纹理具有方向性和尺度差异性,所以用传统方法提取纹理特征分割纹理图像是不准确、不稳定的。文章提出使用结构相似度(structural similarity index,SSIM)分析加噪图像和原始图像的相似性提取纹理结构形态和使用旋转扭曲方法提取边缘信息,然后结合灰度信息并融入水平集进行纹理图像分割。通过对多种类型的复杂纹理图像进行对比实验,将该文提出的方法和已有的一些基于水平集的纹理分割方法进行比较,并使用Jaccard系数计算分割准确度,结果表明,所提出的方法能够更准确地将纹理图像中的目标前景分割出来且有更高的准确率。Segmenting texture images by extracting texture features is an important kind of texture segmentation methods.Traditional texture feature extraction methods are based on local area grayscale statistics.Since the texture has directional and scale differences,the traditional methods cannot be used to accurately extract texture features.This paper proposes using structural similarity index(SSIM)to analyze the similarity between the noise image and the original image to extract the texture structure shape and the rotation distortion method to extract edge information,and then the two methods are combined and merged into the level set to segment the texture images.Through comparative experiments on various types of complex texture images,the method proposed is compared with some existing texture segmentation methods based on level set,and the Jaccard coefficient is used to calculate their segmentation accuracy.It can be found that the method proposed can more accurately segment the target foreground in the texture images and has a higher accuracy rate.

关 键 词：图像分割 水平集 结构特征 结构相似度(SSIM) 旋转扭曲方法 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]