改进的最小交叉Tsallis熵的小目标声呐图像分割

Improved minimum symmetric Tsallis cross entropy for segmentation of a sonar image from a small underwater target

利用一维属性直方图改进交叉Tsallis熵,在此基础上提出了一种基于一维属性直方图的对称最小交叉Tsallis熵水下小目标声呐图像分割方法。该方法的主要步骤是:①抑制水下小目标声呐图像的散斑噪声;②根据图像像素的灰度值和该像素邻域的灰度平均值的大小建立属性集,在属性集上建立与该属性集约束对应的一维属性直方图;③根据一维属性直方图的对称交叉最小Tsallis熵法确定灰度二值化阈值;④对二值化后的图像去除孤立区。实验结果表明:该方法适用于直方图为复杂非双峰形状的水下小目标声呐图像,而且与现有的属性直方图上的一维最大熵阈值化法比较,具有更强的抗噪能力。

A segmentation method on the sonar image of a small underwater target is proposed. In this method, the gray-level threshold for segmentation is acquired via the minimum symmetric Tsallis cross entropy that is based on one-dimensional bound histogram. In this method, first, the speckle noise of the sonar image is suppressed. Second, the bound set is constructed according to the restriction in the gray-level values of both pixels and their neighborhood averages, and the one- dimensional bound histogram corresponding to that bound set is established. Third, the gray-level threshold for segmentation is determined according to the minimum symmetric Tsallis cross entropy based on the one-dimensional bound histogram. Finally, the isolated areas in the threshold image are removed. Experimental results show that the proposed method is well adequate for the images with a nonideal bimodal histogram; it has better antinoise performance than the existing methods based on the entropy of the one-dimensional bound histogram.