摘要
在图像处理领域,谱聚类算法被广泛应用于图像分割任务。但对于背景复杂、色彩界限不明显的图像,传统的谱聚类算法往往无法准确地将目标物体分割出来,同时需要耗费大量的计算资源和时间。因此设计一种基于多尺度形态学梯度重构的分水岭超像素的改进谱聚类算法,它利用多尺度形态学梯度重构的分水岭超像素算法作为谱聚类算法的预处理步骤,将超像素处理所得的结果作为谱聚类的输入,以达到谱聚类算法的运算量目的。验证算法实验结果表明,所提出的改进谱聚类算法在图像分割任务中具有较好的边缘贴合度,并且程序的运行速率得到了显著提升。
In the field of image processing,spectral clustering algorithms are widely used in image segmentation tasks.How⁃ever,for images with complex backgrounds and unclear color boundaries,traditional spectral clustering algorithms often cannot ac⁃curately segment target objects and require a large amount of computing resources and time.Therefore,an improved spectral clus⁃tering algorithm based on multi⁃scale morphological gradient reconstructed watershed superpixels is designed.It uses the multi⁃scale morphological gradient reconstructed watershed superpixel algorithm as a preprocessing step of the spectral clustering algorithm.The processed results are used as the input of spectral clustering to achieve the computational purpose of the spectral clustering algorithm.The experimental results of the verification algorithm show that the proposed improved spectral clustering al⁃gorithm has better edge fit in the image segmentation task,and the running speed of the program has been significantly improved.
作者
陈瑜萍
覃正优
Chen Yuping;Qin Zhengyou(School of Computer and Information Engineering,Nanning Normal University,Nanning 530100,China;Guangxi Key Lab of Human-Machine Interaction and Intelligent Decision,Nanning 530100,China)
出处
《现代计算机》
2024年第17期23-29,共7页
Modern Computer
基金
广西高校中青年教师科研基础能力提升项目(2022KY0378)。
关键词
图像分割
超像素算法
谱聚类算法
改进分水岭算法
多尺度形态学梯度重构
image segmentation
spectral clustering algorithm
superpixel algorithm
improved watershed algorithm
multi⁃scale morphological gradient reconstruction
