结合形态学运算的谱抠图声纳图像分割法

摘要

针对数字抠图与图像分割之间的联系,提出一种结合形态学运算的谱抠图声纳彩色图像分割法。首先通过形态学中的顶帽（top-hat）变换和底帽（bottom-hat）变换去除声纳图像中阴影的影响和背景的不均匀性,并进行图像增强;其次,将数字抠图中的alpha值考虑到图像分割中,通过全局平均迭代阈值法获取阈值,对抠图所获alpha图像进行阈值处理得出分割结果;最后,与多种现有的分割方法进行仿真对比实验,实验结果证明了本文分割方法的有效性。%Based on the link between digital matting and image segmentation,a sonar image segmentation method using spectral matting and morphological operations was proposed.First,the effect of sonar image shadows of and background nonuniformity were removed and the image was enhanced through morphological top-hat transformation and bottom-hat transformation.Then,the alpha value of the digital matting was taken into consideration in image segmentation;the threshold value was obtained by the global average iteration method;and the segmentation results were obtained by threshold process of the alpha images.Finally,simulation experiments were conducted to compare the proposed segmentation method with existing popular segmentation methods.The results show that the proposed method is more suitable for sonar image segmentation and the output pictures are more reliable with more details than other methods.