摘要： 水声图像常分为亮区、暗区和混响区，而传统的单阈值方法不能根据需要获得相应区域，针对该问题，提出一种基于修正的灰 度-梯度二维直方图的最大熵分割方法。该方法根据先验知识截取部分灰度-梯度二维直方图，并对其进行最大熵阈值分割。实验结果表明，该方法可以根据需要提取出感兴趣的区域，并且能得到更好的分割效果。

关键词: 水声图像, 图像分割, 阈值选取, 属性直方图, 最大熵

Abstract: Aiming at the problem that the underwater acoustic image is often divided into light areas, dark areas and reverberation areas, while it is unable to obtain the corresponding region through the traditional single thresholding method, the image segmentation based on the maximum entropy of the amended gray-gradient 2D histogram method is proposed. In this method, priori knowledge is used to intercept part of the gray- gradient 2D histogram, which is segmented through the maximum entropy. Experimental results demonstrate that the interested region can be distilled as needed, and better segmentation results can be achieved by adopting this method in acoustic image segmentation.

Key words: underwater acoustic image, image segmentation, threshold selection, attribute histogram, maximum entropy

中图分类号: 

• N945