基于NSST和改进数学形态学的遥感图像目标边缘提取

发布时间：2018-04-22 17:39

本文选题：目标边缘提取 + 遥感图像　；参考：《图学学报》2017年04期

【摘要】：为了从遥感图像中提取出更为准确完整的目标边缘,提出一种基于无下采样Shearlet模极大值和改进数学形态学的目标边缘提取方法。首先采用无下采样Shearlet变换(NSST)将图像分解成边缘细节丰富的高频分量和边缘细节较少的低频分量;然后结合不同分解程度下边缘像素点处的系数关系,对高频分量的各个子带进行模极大值检测,再经过双层掩膜筛选后得到高频边缘提取结果;对低频分量采用改进的数学形态学方法,得到低频边缘提取结果;最后将上述两部分融合,使用区域连通方法去除孤立点,得到最终的目标边缘图像。大量实验结果表明,与Canny以及其他4种同类边缘提取方法相比,本文方法所得边缘定位准确、完整清晰、细节丰富,且抗噪能力强,为后续遥感图像目标特征提取与识别奠定更好基础。
[Abstract]:In order to extract more accurate and complete target edges from remote sensing images, a new method based on unsampled Shearlet modulus maximum and improved mathematical morphology is proposed. Firstly, the image is decomposed into high-frequency components with rich edge details and low-frequency components with less edge details by using non-down-sampling Shearlet transform. Every sub-band of the high frequency component is detected by the modulus maximum, then the high frequency edge is extracted by double mask screening, and the low frequency component is extracted by the improved mathematical morphology method. Finally, the two parts are fused and the final target edge image is obtained by using the region connectivity method to remove the outliers. A large number of experimental results show that compared with Canny and other four other similar edge detection methods, the proposed method is accurate, complete and clear, rich in details, and has strong anti-noise ability. It lays a better foundation for target feature extraction and recognition in the following remote sensing images.
【作者单位】：南京航空航天大学电子信息工程学院;浙江大学CAD&CG国家重点实验室;城市空间信息工程北京市重点实验室;南京水利科学研究院港口航道泥沙工程交通行业重点实验室;黄河水利委员会黄河水利科学研究院水利部黄河泥沙重点实验室;哈尔滨工业大学城市水资源与水环境国家重点实验室;
【基金】：国家自然科学基金项目(61573183) CAD&CG国家重点实验室开放基金项目(A1519) 城市空间信息工程北京市重点实验室开放基金项目(2014203) 港口航道泥沙工程交通行业重点实验室开放基金项目水利部黄河泥沙重点实验室开放基金项目(2014006) 城市水资源与水环境国家重点实验室开放基金项目(LYPK201304) 江苏高校优势学科建设工程资助项目
【分类号】：TP751

【相似文献】