摘要:
影像分割是面向对象影像分析中的重要步骤.为了提高高分辨率遥感影像(high-resolution remote sensing image,HRI)分割算法的性能,提出一种新的影像分割算法,包含种子确定、基于种子区域生长(seeded region growing,SRG)的过分割(advanced SRG,ASRG)和层次区域生长(hierarchical region growing,HRG)3个步骤.利用Gabor纹理特征定义纹理均匀性,将种子自动放置在HRI中同一纹理组成区域的中心位置;在SRG阶段,将HRI光谱信息与斑块形状信息相结合,提出了一种新的合并规则,以提高SRG过分割的精度与分割结果中各个斑块排列的紧凑性;在HRG阶段,提出了一种自适应的阈值,可以更好地保持多尺度分割的特性;在实验部分,采用3景HRI验证了上述方法.利用监督的影像分割评价方法定量评价了该方法的分割精度,并与另外2种主流的遥感影像分割算法进行了对比.结果表明,该方法可以得到令人满意的分割效果.
Abstract:
Image segmentation plays an important role in object-based image analysis.In order to enhance the performance of segmentation method for hierarchical region growing (HRG),this paper proposes a new image segmentation algorithm.The new method consists of three steps: seed determination, seeded region growing (SRG)based over-segmentation (advanced SRG, ASRG) and HRG.To improve the automation and precision of seeds determination, the authors used Gabor texture feature and defined textural homogeneity, attempting to place the seeds at the center of the regions composed of the same texture.At the stage of SRG, spectral information of HRI was combined with shape cues to form a new merging rule to raise the segmentation accuracy and segments compactness of SRG over-segmentation.At the HRG step, an adaptive threshold was used to better retain the multi-scale segmentation property.In the experiment, three scenes of HRI were utilized to validate the proposed method.A supervised segmentation evaluation method was adopted to quantitatively assess the segmentation accuracy of the proposed algorithm, and two state-of-the-art segmentation methods were compared with the proposed method.The experimental results show that the new algorithm proposed in this paper can produce satisfying segmentation.
下载: