基于Mean Shift的大批量遥感影像分割方法

Segmentation of large scale remote sensing image based on Mean Shift

由于收敛速度快、分割精度好,Mean Shift算法被广泛应用于影像分割中,但是处理大遥感影像时,Mean Shift算法存在速度慢、效率低下等问题。为此提出一种基于Mean Shift的分块并行无缝分割算法。首先在分块并行Mean Shift分割的基础上,通过标签影像的统一编码和重叠区域标签值建立对应关系,确定分块线的消除准则;然后进行标签影像的行和列拼接;最后对拼接后的标签影像进行矢量化,生成最终分割结果。实验表明,该算法相对于原始Mean Shift算法,在保证分割结果可靠性的同时大大提高了影像分割的效率,能够很好地解决大批量遥感影像的分割问题。

Mean Shift algorithm has been widely used in image segmentation because of its fast convergence speed and good segmentation accuracy.However,when large scale remote sensing images are processed,Mean Shift algorithm has some problems,such as slow speed and low efficiency.In this paper,a parallel seamless segmentation algorithm based on Mean Shift is proposed.The algorithm is based on block parallel Mean Shift segmentation.The elimination criterion of block lines is determined by uniform coding of label images and establishing corresponding relations between label values of overlapping regions.Then,the row and column directions of the label image are stitched together.Finally,the segmented label image is vectorized to generate the final segmentation result.Compared with the original Mean Shift algorithm,the algorithm put forward in this paper can not only ensure the reliability of segmentation results but also greatly improve the efficiency of image segmentation,and can also solve the problem of large scale remote sensing image segmentation.