基于Gabor纹理描述的遥感影像分割

本文引入了在不同影像层次上的Gabor纹理特征,采用分裂-合并加智能像素精致的方法,实现遥感影像的非监督分割。实验结果表明采用Gabor滤波器为基础的多分辨率分析来描述高分辨率遥感影像的纹理特征,可以明显地描述影像的高、低频特征,并且基于Gabor纹理特征进行遥感影像的分割是有效的;将本文方法的分割结果与经典的eCognition分割结果进行了对比试验,表明本文方法的分割结果较好。

Segmentation of remote sensing imagery based on Gabor texture description

The paper introduced the method of a split-merge plus pixel-wise refinement based on the hierarchical Gabor texture feature.Without a priori knowledge of imagery contents,the remote sensing imagery can be segmented into different regions,the unsupervised segmentation of remote sensing imagery was achieved.Experimental results showed that it could clearly describe the high,low-frequency characteristics of the imagery based on Gabor filter multi-resolution analysis,and it was effective to use the Gabor texture features in high-resolution remote sensing imagery segmentation.The paper compared the segmentations with the eCognition's.Experimental results showed that the method was better than the eCognition's.