一种基于标记分水岭的高分辨率遥感影像分割方法  

A Marker-Based Watershed Segmentation for High Resolution Remote Sensing Image

作　　者：李珀任[1] 潘懋[1] 杜世宏[1] 

机构地区：[1]北京大学造山带与地壳演化教育部重点实验室,北京大学地球与空间科学学院,北京100871

出　　处：《地理与地理信息科学》2012年第5期10-15,共6页Geography and Geo-Information Science

基　　金：国家自然科学基金项目(40742015)

摘　　要：为避免分水岭分割高分影像的过分割现象并充分利用高分影像的特点,该文提出一种基于标记的分水岭分割方法,即基于影像先验知识指导分割的原则,对遥感影像进行各向异性扩散平滑滤波后,计算生成融合了光谱和纹理特征的梯度图像;从梯度图像中提取标记重建对象的边缘特征,并执行基于标记的分水岭变换,得到最终的分割结果。实验表明,该方法能够充分利用高分影像的特征信息,并能有效地抑制分水岭过分割现象。Watershed segmentation is usually adopted for image segmentation in object-oriented analysis, while it produces excessive over-segmentation as the main disadvantage. To remedy this over-segmentation problem for high resolution remote sensing image, this paper presents a novel approach to watershed segmentation. Mainly, with the priory knowledge of the image, a merged gradient image, including intensity gradient and texture gradient, is computed after smoothing the original image with Perona-Malik anisotropic diffusion filter. Then, a marker image can be extracted from the gradient image and used in the water- shed transform procedure. Experiments show that this approach can exploit the characteristics of a high resolution remote sens ing image to produce high quality segmentation, while effectively suppress the over-segmentation phenomenon.

关 键 词：图像分割 标记分水岭分割 纹理梯度 

分 类 号：TP750[自动化与计算机技术—检测技术与自动化装置]