摘要
针对传统Otsu多阈值分割方法对SAR图像分割存在对噪声敏感且计算量大的问题,提出了一种结合降斑各向异性扩散(speckle reducing anisotropic diffusion,SRAD)和自适应量子遗传算法的Otsu多阈值SAR图像分割方法。首先,利用SRAD对SAR图像进行滤波,滤除其相干斑噪声,并通过获取滤波迭代过程中图像间的平均结构相似度,有效地控制迭代过程;通过图像的直方图和阈值的组合来定义图像的类间方差。然后,将阈值的组合编码为量子染色体;设置若干量子染色体构成初始阈值组合种群,并对每个组合个体以定义的类间方差作为评价标准进行适应度评价。利用量子旋转门作用于量子染色体叠加态的基态实现其进化,并根据相邻两代量子染色体的差异,逐代地调整量子旋转角的大小;以最终演化的阈值组合种群中适应度最大的阈值组合个体作为最优阈值组合,实现SAR图像最优多阈值分割。为验证所提出的分割方法,对模拟和真实SAR图像进行了实验。定性和定量评价结果表明了该方法的可行性和有效性。
Aiming at the problem that the traditional SAR image segmentation based on Otsu multi-threshold segmentation method is sensitive to noise and large in calculation,an Otsu multi-threshold SAR image segmentation combining speckle reducing anisotropic diffusion(SRAD)and adaptive quantum genetic algorithm is proposed.Firstly,the SAR image is filtered by SRAD to remove the speckle noise,and the average structural similarity between the images in the filtering iteration process is obtained to control the iterative process effectively.The inter-class variance of the image is defined by the combination of histogram and thresholds.Then,the combination of the thresholds is encoded as a quantum chromosome;aplurality of quantum chromosomes are set to constitute an initial threshold combined population,and the fitness evaluation is performed for each of the combined individuals using the defined inter-class variance as an evaluation criterion.The quantum revolving gate is used to realize the evolution of the ground state of the quantum chromosome superposition state,and the quantum rotation angle is adjusted from generation to generation according to the difference of the adjacent two generations of quantum chromosomes.The optimal multi-threshold segmentation of SAR images is realized by using the individuals with the highest fitness in the final evolving threshold combination population as the optimal threshold combinations.In order to verify the proposed segmentation method,experiments were performed on simulated and real SAR images.The qualitative and quantitative evaluation results show the feasibility and effectiveness of the method.
作者
杨蕴
李玉
王玉
赵泉华
YANG Yun;LI Yu;WANG Yu;ZHAO Quanhua(School of Geomatics,Liaoning Technical University,Fuxin,Liaoning 123000,China)
出处
《遥感信息》
CSCD
北大核心
2019年第4期29-38,共10页
Remote Sensing Information
基金
国家自然科学基金(41301479、41271435)
辽宁省自然科学基金(2015020090)
关键词
SAR图像分割
多阈值
降斑各向异性扩散
最大类间方差
自适应量子遗传算法
SAR image segmentation
multiple threshold
speckle reducing anisotropic diffusion
maximum variance between-class
adaptive quantum genetic algorithm