简单线性迭代聚类的高分辨率遥感影像分割  

Image Segmentation of High Resolution Remote Sensing Imagery Based on Simple Linear Iterative Clustering

作　　者：董志鹏[1] 梅小明[1,2] 陈杰[1] 邓敏[1] 李昕[1] 

机构地区：[1]中南大学地球科学与信息物理学院,长沙410083 [2]东华理工大学江西省数字国土重点实验室,南昌344000

出　　处：《遥感信息》2016年第6期83-88,共6页Remote Sensing Information

基　　金：地理空间信息工程国家测绘地理信息局重点实验室经费资助项目(201322);江西省数字国土重点实验室开放研究基金资助项目(DLLJ201406);国家自然科学基金(41201428);测绘遥感信息工程国家重点试验室开放研究基金项目(13R01)

摘　　要：影像分割是面向对象影像分析的基础和关键。针对传统影像分割方法地物边界依附性差、易受影像噪声影响等问题,提出一种简单线性迭代聚类(Simple Linear Iterative Clustering,SLIC)的高分辨率遥感影像分割方法。该方法首先用SLIC算法对影像过分割生成SLIC超像素,之后根据相似性规则对SLIC超像素进行合并实现影像分割;然后通过构造Lab颜色空间下的五维特征参数度量影像像素的局部特征差异,并通过SLIC算法把具有相似性特征的像素聚类生成超像素,克服影像噪声对分割结果的影响;最后根据相似性合并规则以超像素为基本单元进行区域合并,从而达到分割目的。实验结果表明,所提出方法具有良好的高分辨率遥感影像分割结果。High resolution remote sensing image segmentation is a basis and key step for remote sensing image analysis. The traditional image segmentation methods have poor dependent boundary and are sensitive to image noises. A segmentation method for remote sensing image based on simple linear iterative clustering （SI.IC） super-pixels,which can effectively solve the problems with poor dependent boundary and image noise is proposed. The proposed method firstly constructs five feature parameters in the Lab color space,and measures difference of local feature of pixels in image by the five feature parameters in the Lab color space. Pixels having similar feature pixels are gathered into super pixels by SLIC algorithm to eliminate effect of image noise on segmentation. Super-pixels which are basic unit can be merged by similarity rule to achieve segmentation of the remote sensing image. The experimental results show that the proposed method can obtain better segmentation results.

关 键 词：高分辨率遥感影像 SLIC超像素 影像过分割 影像分割 超像素合并 聚类 

分 类 号：P237[天文地球—摄影测量与遥感]