摘要
针对现有分割算法对高噪声侧扫声呐图像分割准确率低的问题,提出了一种综合利用NSCT(non-subsampled contourlet transform)分解图像、局部标准差和均值组合增强图像和多重分形判断图像奇异性的侧扫声呐图像分割方法。首先,借助NSCT分解图像,获得滤除高频噪声且保留轮廓信息的低频图像和一系列高频方向子带图像。然后,基于侧扫声呐图像中目标及其阴影伴随出现的特点,计算低频图像的局部标准差与均值的组合特征,获得分别突显目标及其阴影的特征图,使用多重分形分割方法分割特征图,获得低频图像分割结果;利用图像差分和非极大值抑制方法分割高频方向子带图像,获得高频分割结果;融合高低频分割结果获得目标及其阴影的精细边缘。最后通过试验验证了本文方法的有效性。
To improve the accuracy of current segmentation algorithms for the side scan sonar image with high noise,a side scan sonar image segmentation method is proposed that comprehensively utilizing of decomposing image by NSCT(non-subsampled contourlet transform),enhancing image by combination of local standard deviation and mean,estimating the singularity of image by multi-fractal.Firstly,NSCT is used to decompose images to obtain low frequency image which filtered out high frequency noise and retain contour information and a series of high-frequency direction sub-band images.Then,based on the feature that target shadow appeared with the target in side scan sonar images,it is calculated that the low-frequency image feature combined the local standard deviation and mean to obtain the feature images that highlight the characteristics of the target and its shadow respectively,use the multifractal method to segment the feature image to get the result of low-frequency image segmentation.The image difference and non-maximal suppression methods are used to segment the high-frequency direction sub-band images and obtain the high-frequency segmentation results.Finally,it is obtained that the fine edge of the target and its shadow by combing of high and low frequency segmentation result.The validity of this method is verified by experiments.
作者
何义才
赵建虎
张红梅
阮世伦
HE Yicai;ZHAO Jianhu;ZHANG Hongmei;RUAN Shilun(School of Geodesy and Geomatics, Wuhan University, Wuhan 430079, China;Shool of Power and Mechanical Engineering, Wuhan University, Wuhan 430072, China;Institute of Marine Geophysics, VAST, 18 Hoang Quoc Viet, Ha Noi 10000, Viet Nam)
出处
《测绘学报》
EI
CSCD
北大核心
2020年第2期162-170,共9页
Acta Geodaetica et Cartographica Sinica
基金
国家重点研发计划(2016YFB0501703)~~
