基于二维希尔伯特变换的相位一致模型图像特征检测方法_王珂.pdfVIP

39 　6 　　　 Vol.39, No.6 　2010 12 Acta Geodaetica et Cartographica Sinica Dec., 2010 　　 :1001-1595 (2010)06-0605-06 王　珂, 肖鹏峰 , 210093 The Algorithm of Image Features Detection from Phase Congruency Model Based on 2-D ilbert Transform WANG Ke, XIAO Pengfeng Department of Geographical Information Science of Nanj ing University, Nanjing 210093, China Abstract:The algorithmof phase congruency developed from phase information of image in frequency domain, is employed for imagefeature detection.Thetheory of phase congruency is that image features, such as step edge, roof, Mach band and delta, always occur at points where the phases of harmonic components come to the maximum congruency.This algorithm is realized to extract the image features by constructing the local energy model being normalized to get the value of phase congruency of every point in image.This paper introduces the 2-D ilbert transform instead of 1-D ilbert transform to propose the algorithm of phase congruency for detecting the image features.The modified algorithm can take account of the full directions of the image features.Meanwhile, the proposed method simplifies the calculation of the numerator of local energy by convoluting the original image with thewindow operator to remove DC (Direct Current)component of currentwindow and 2-D ilbert transform respec- tively.Moreover, this algorithmmakes the denominatorof the model of phase congruency adding the DC component to suppress the noise of image.The modified algorithmof phase congruency is implemented into the natural image and remotely sensed imagery, and the results show that the modified algorithm iseffectiveto detect image features. Key words:phase congruency;2-D ilbert transform;local energy;image feature detection 　:相位一致方法是从频域中的相位信息理论中延伸出的一种图像特征检测方法。 原理是指图像特征如跃级边 缘、线形、屋脊形和马赫带等, 总发生在相位的最大叠合处。该原理通过构造局部能量模型, 并经标准化后, 度量 图像 各 个位置的相位一致值。在前人的基础上对