基于数据场模型的多光谱图像边缘检测

Edge detection for multispectral images based on data field model

数据场模型模拟了物理世界中力的相互作用,具有较好地描述数据分布、反映数据关系的优势.基于这一优势,提出了一种新的多光谱图像边缘检测算法.在特征空间,将多光谱图像像元看成是具有辐射能力的质点,利用数据场理论计算不同像元之间的相互作用,从而将多光谱图像从特征空间映射到数据场的势空间.多光谱图像的结构能够通过势场图解释,从而可以通过势场图获得多光谱图像的边缘图像.实验结果表明,基于数据场的边缘检测算法对多光谱遥感图像的检测具有显著的效果,能够正确识别地物边缘,有利于遥感图像地学分析的进一步处理.

The data field model can simulate the forces in the physical world and has the advantages of describing the data distribution and reflecting the data relationship.On the basis of these advantages,this paper presents a novel multispectral image edge detection method.In the feature space,each multispectral image pixel is assumed to be a particle w ith emissivity.The interactions betw een the pixels are calculated by the data field theory,and then the multispectral image can be mapped from the feature space to an appropriate potential space.The structure of the multispectral image can be illustrated by the potential maps and the edge maps of the multispectral image are obtained by the potential map.Experimental results show that the new method is efficient and effective for multispectral images.The edges of landmark can be correctly discerned,w hich is available for the next processing of geoanalysis for remote sensing.