基于过渡区的遥感图像分割

Transition-Region-based Remote Sensing Image Segmentation

许多遥感图像的目标和背景之间界定模糊,表现为过渡平缓,往往导致边界定位和目标的提取出现误差。首先利用有效平均梯度和特殊剪切来获得目标与背景之间的过渡区,由于目标的边界必在过渡区内,所以提取误差会被限制在一定的范围内。最后利用最小距离法将过渡区内的像素判决后划分到目标或背景中去,进一步保证目标的质量,解决了此类问题中目标提取的困难。实验结果表明了这种方法的有效性。

In a lot of remote sensing image, there are fuzzy boundary, smooth transition, causing the difficulty that the border is oriented and generating very great errors for the extract of object This paper solves the difficulty of object extraction to a certain extent Because the object that is extracted must be in the transition region, it can limit errors of extraction within certain range The pixels in transition region are divided into object and background respectively by Minimum Distance Rule in order to guarantee the quality of the object further The experimental result has shown the validity of this kind of method