【摘要】 由于小波变换方向分辨率较低,而高分影像中地物几何结构丰富,边缘存在于各个方向,导致了图像小波变换的频率系数仅被分解到水平、垂直和对角3个方向子带上,使得小波变换对具有复杂几何结构的地物边缘提取存在不足。为此提出一种Directionlet变换理论结合模极大值法的高分影像边缘提取方法。首先,对原始影像进行基于格的分解获得任意方向一维线集合,再进行小波变换并恢复图像格式得到高频方向子带;然后,利用改进模极大值法和双阈值法对系数进行处理得到边缘结果;最后,采用数学形态学方法对边缘提取结果进行细化连接等后处理,从而实现对高分影像边缘的提取。实验结果表明,该方法与传统边缘检测方法和小波变换相比,边缘提取更加完整,定位精度更高。更多还原

【Abstract】 In view of the low resolution in direction of wavelet transform,the rich geometric structure of high spatial resolution remote sensing imagery(HSR) and the existence of edge in various directions which causes shortcomings in edge extraction of wavelet transform on objects with complex geometric structure,the authors proposed a HRS image edge extraction method of multi-direction wavelet transform based on Directionlet theory and modulus maximum method.The method first decomposes original image based lattice to obtain one-dimensional line set,then carries out wavelet transform and obtains the high frequency directional sub-band by restore image format.The edge result is obtained by using improved module maximum and the dual threshold method.Finally,the mathematical morphology is used to refine and connect edge results.Experiment result shows that the proposed method can get more complete edge compared with traditional method and standard two-dimensional wavelet transform.更多还原