基于改进OTSU算法的多时相立体卫星图像分割

Multi Temporal Stereo Satellite Image Segmentation Based on Improved OTSU Algorithm

针对传统OTSU算法仅适用于目标与背景分布均匀的图像,在处理多时相卫星图像时具有一定局限性,提出基于改进OTSU算法的多时相立体卫星图像分割方法。根据图像灰度均值,在一维OTSU算法基础上提出二维OTSU算法,计算图像中像素领域平均灰度值;通过图像邻域灰度得出邻域窗口取值,获取邻域灰度值二维直方图,得到图像均值矢量;将传统二维算法拆分成两个一维OTSU算法,用划分后两个一维OTSU算法得到阈值来替代之前阈值结果,完成图像分割。仿真结果表明,所提方法可高效分割多时立体卫星图像,且分割耗时较短,具有一定的适用性以及有效性。

Since the traditional Otsu algorithm is only applicable to images with uniform distribution of target and background,and has certain limitations in processing multi temporal satellite images,a multi temporal stereo satellite image segmentation method based on improved Otsu algorithm is proposed.Firstly,based on the mean value of image gray and one-dimensional OTSU algorithm,a two-dimensional OTSU algorithm was proposed to calculate the average gray value of pixels in the image.Secondly,according to the value of the neighborhood window obtained from the gray value of the image neighborhood,the two-dimensional histogram of the gray value of the neighborhood was obtained,thus getting the image mean vector.Finally,the traditional two-dimensional algorithm was divided into two one-dimensional OTSU algorithms.The thresholds obtained with the two one-dimensional OTSU algorithms replaced the previous threshold results to achieve the image segmentation.Simulation results show that the method has short segmentation time,excellent applicability and effectiveness.