【摘要】 Ｍｏｒｒｏｎｅ 和她的合作者提出一种对灰度图像提取边缘的模型。对一维信号，边缘对应于信号最大傅里叶相位叠合处，通过信号与两个相位差为９０°的滤波器卷积，然后计算其局部能量，边缘的位置定位于局部能量的峰值处，但是这一模型存在下面一些问题：（１） 如果傅里叶分量幅度的和很小，计算相位叠合的问题将变为一个病态问题；（２） 如果信号能量落在信号噪声能量区域，相位叠合将对噪声很敏感；（３） 如果只存在一个频率分量或者频带很窄，相位叠合的值总是１ 或１附近很小的范围内，很难得到好的相位叠合图。这种情况常常出现在经过平滑的信号（ 图像） 中；（４） 由于相位叠合函数随相位偏差的余弦变化，可能会带来由于余弦函数下降太慢造成定位不精确，为此，给出了一种新的相位叠合模型的定义更多还原

【Abstract】 Morrone and co\|workers have proposed a model for edge detection in grey\|level images. Assuming an one\|dimensional visual signal, edges corresponding to points of maximal Fourier phase congruency of the signals and localized at peaks of an energe function obtained as the quadratic combination of convolutional of the signal with two filters in Fourier quadrature, but there are some problems in this model: (1) the caculation of phase congruency becomes ill\|conditioned if the sum of the Fourier amplitudes is very small. (2) phase congruency becomes highly sensitive to noise in regions where the signal energy falls within the background local energy value of the signal noise. (3) if there is only (or nearly only one) frequency component in the signal, phase congruency is always one. (4) the definition of phase congruency does not provide good localization as this function only varies with the cosine of phase deviation。 In order to resolve these problems, a new defination is presented.更多还原