基于纹理特征的自适应SLIC超像素分割算法  

An Adaptive SLIC Super-Pixel Segmentation Algorithm Based on Texture Features

作　　者：徐霞 XU Xia(The Chengdu University of Technology,Chengdu Sichuan 610059,China)

机构地区：[1]成都理工大学,四川成都610059

出　　处：《信息与电脑》2022年第21期173-175,共3页Information & Computer

摘　　要：由于简单线性迭代聚类算法(Simple Linear Iterative Cluster,SLIC)只考虑了颜色和空间信息导致分割不准确且边界附着度不高,且人工预设的超像素块数也会影响后续分割效果,提出了一种基于纹理特征的自适应SLIC超像素分割算法。先使用图像复杂度衡量图像分割的难易程度,根据自适应计算合适的图像分割块数,再基于SLIC算法把局部二值模式(Local Binary Patterns,LBP)纹理特征纳入相似性度量,提高SLIC算法分割精度。实验结果表明,本文方法与SLIC算法相比有更高的评价指标。Because Simple Linear Iterative Cluster(SLIC) only considers color and spatial information, resulting in inaccurate segmentation and low boundary adhesion, and the number of manually preset super pixel blocks will also affect the subsequent segmentation effect, an adaptive SLIC super pixel segmentation algorithm based on texture features is proposed. First, the image complexity is used to measure the difficulty of image segmentation, and the appropriate number of image segmentation blocks is calculated adaptively. Then,the Local Binary Patterns(LBP) texture features are included in the similarity measurement based on the SLIC algorithm to improve the segmentation accuracy of the SLIC algorithm. The experimental results show that the method in this paper has a higher evaluation index than SLIC algorithm.

关 键 词：简单线性迭代聚类算法 自适应 图像复杂度 局部二值模式(LBP)纹理特征 超像素分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]