基于梯度特征的图像自动分割方法

利用遥感图像进行海上目标检测、监视的关键技术是实现海陆分离，从而为目标检测、监视减少不必要的处理工作。传统的采用阈值分割方法分离陆地，但是这种方法对于陆地上灰度级较低区域容易造成误分割。而且，很显然的不能够充分利用图像中的灰度变化信息，即梯度。为此提出了一种基于梯度特征的海陆分割算法。首先利用边缘检测提取图像的梯度特征，梯度特征对灰度级并不敏感而对梯度的变化敏感，对于陆地灰度级较低的区域也能够提取边缘。对得到的边缘进行形态学处理，利用设定的面积阈值将图像分为主体陆地区域和孤立陆地区域，进一步可以屏蔽陆地。实验结果表明，该方法可以较好的分割陆地中阴暗区域，对于未经辐射校正图像有较好的适用性，并且有较快的处理速度。

An Automatic Image Segmentation Approach Based Gradient

Target detection and surveillance on the sea utilizing remote sensed images, segmentation should he do to shield the land area. Traditionally, thresholding method is used to segment remote sensed images. However, there are some shortcomings using this method, firstly threshold selecting is difficult, secondly for images have darker gray level this method results in error segmentation easily, thirdly it consumes more computation time. In this paper a new approach based gradient is proposed. First edge detection operator is used to pick-up edge of image, divided image into two parts, the sparseness area of edge and the denseness area of edge, which corresponding the sea area and the land area respectively. Then the binary edge image is processed using morphologic operator, so as to improve detecting land area easily and getting more accuracy coastline. Experiment results indicate that the new approach possesses better performance.