摘要
针对海陆遥感图像存在对比度低、斑点噪声、弱边界等问题,提出一种基于随机游走的弱交互式水边线提取改进算法。将色彩和梯度特征引入到随机游走算法中,利用l2-范数与高斯权函数相结合构造一种新的相似度矩阵,以获得一个加权图。给出标签种子点,借助于电路模拟与组合Dirichlet问题计算图中节点间电势值,并根据电势值的大小进行分类,从而提取到高质量的水边线。仿真实验结果表明,与传统算法相比,改进的算法具有很强的抗干扰能力,对于遥感图像水边线的提取更精确,特别对于弱边缘的水边线有良好的提取能力。
A weak interactive waterside line extraction improved algorithm based on random walks is proposed to solve the pro- blems that the sea and land remote sensing images have the drawback of low contrast, speckle noise and weak boundaries. First, color and gradient features are introduced into random walks algorithm, and then 12 norm and Gaussian weighting function are used to creat a novel affinity matrix to obtain a weighted graph. Second, a small number of labeled pixels are given to calculate the potential of each node with the aid of circuit simulation and combinatorial Dirichlet problem. The classification is obtained ac- cording to the potential value between nodes, therefore a high-quality waterside line can be obtained. The simulation experiments demonstrate that the improved algorithm has a strong anti-jamming capability and is more accurate to the waterside line extrac- tion of remote sensing images compared with traditional algorithms. Especially for weak boundary of waterside line, it has the fine ability to extract.
出处
《计算机工程与设计》
CSCD
北大核心
2013年第4期1442-1445,1516,共5页
Computer Engineering and Design
关键词
遥感图像
弱交互式
随机游走
水边线提取
梯度信息
remote sensing images
weak interactive
random walks
waterside line extraction
gradient information
