双阈值Ostu算法的侧扫声纳图像分割

Double Threshold Ostu Side- scan Sonar Image Segmentation Algorithm

为了利用侧扫声纳进行水下目标的探测和识别,首先必须将声纳图像分为亮点、阴影和背景三部分。由于多数侧扫声纳图像各部分灰度对比较明显,所以适合采用阈值分割的方法进行图像分割。本文针对基本的Ostu阈值提取算法,提出了改进的双阈值Ostu算法,从而满足将图像分为三部分的需求,得到分割阈值。在Ostu算法中,除了使用灰度均值,还使用了灰度方差作为特征量对阈值进行了自动提取。然后根据侧扫声纳图像的特征提出了一种快速分割的算法,将提取的阈值应用到该算法中,成功地将侧扫声纳的亮点和阴影在复杂的背景噪声中分割提取出来。并且发现方差比均值更适合用于Ostu算法进行图像分割,得到的分割效果更好,提高了算法的正确性和合理性。

Sidescan sonar image must be segmented into regions of object highlight,shadow and background before underwater object can be detected and recognized automatically and accurately. Due to the obvious contrasts of the grayscales of every region,the threshold segmentation algorithm is used. Based on the traditional Ostu threshold extracting algorithm,the improved double threshold Ostu algorithm,which is aimed to segment the image into three regions,is proposed. Besides the grayscale average,the variance is also used in the Ostu algorithm as a characteristic for threshold extracting. After this,the quick- segmentation algorithm is presented,which uses the thresholds as parameters. Then the highlight,the shadow and the background of the sidescan sonar image will be segmented successfully. Meanwhile,it is proved that variance is more useful for threshold extracting in the Ostu algorithm,it makes the segmentation better,more accurate and more reasonable.