结合光谱和纹理的高分辨率遥感图像分水岭分割

发布时间：2018-02-26 03:22

本文关键词： 遥感图像分割双边滤波 Gabor滤波梯度分水岭变换形态学膨胀　出处：《武汉大学学报(信息科学版)》2017年04期 　论文类型：期刊论文

【摘要】：针对遥感图像分割时仅利用光谱信息容易造成过分割和边缘定位不准的问题,提出一种结合光谱强度和纹理信息的遥感图像分水岭分割算法。首先分别提取图像的光谱梯度和纹理梯度,提出一种改进双边滤波模型,滤除图像中的噪声的周时,采用了一种局部的平滑尺度,能够有效消除纹理信息,借助于滤波算法,分别对原图像和Gabor纹理特征图像进行平滑处理,利用边缘检测算子得到光谱梯度和纹理梯度。最后利用形态学膨胀方法进行融合融合,使用分水岭变换对图像分割。用三幅高分辨率彩色遥感图像数据进行实验,并与JSEG(Joint Systems Engineering Group)和多分辨率分割方法进行比较,结果表明该方法具有较高的边界定位准确性,同时降低了过分割和欠分割现象。
[Abstract]:In view of the problem of over-segmentation and inaccurate edge location caused by only using spectral information in remote sensing image segmentation, A watershed segmentation algorithm based on spectral intensity and texture information is proposed. Firstly, the spectral gradient and texture gradient of the image are extracted, and an improved two-sided filtering model is proposed to filter the period of noise in the image. Using a local smoothing scale, texture information can be effectively eliminated, and the original image and Gabor texture feature image are smoothed with the help of filtering algorithm. The spectral gradient and texture gradient are obtained by edge detection operator. Finally, the morphological expansion method is used to fuse the image, and the watershed transform is used to segment the image. Compared with JSEG(Joint Systems Engineering Group and multi-resolution segmentation methods, the results show that the proposed method has high accuracy of boundary location and reduces the phenomenon of over-segmentation and under-segmentation.
【作者单位】：海军装备研究院;海军航空工程学院;
【基金】：泰山学者工程专项(ts201511020)~~
【分类号】：TP751

【参考文献】