文中给出一种适应于图像多种特征(阶跃边缘,线条,屋脊边缘以及马赫带等)的特征提取方法。这种方法利用图像傅里叶分量在特征点叠合次数最多的概念(相位叠合),利用相位叠合来标记特征点比用梯度方法有明显的优点:相位叠合是一种无量纲的量,而且对图像的亮度或对比度是不变的,因此它能提供对特征点的绝对度量。实验证明,相位叠合提取特征对图像的照度(或对比度)以及可能出现的噪声是不变的 In this paper, a feature extraction method that adapts to various features of image (step edge, line, ridge edge and Mach band) is presented. This method utilizes the concept that the Fourier component of the image has the largest number of coincidences at the feature points (phase congruency). Using the phase congruency to tag feature points has the obvious advantage over the gradient method that phase congruency is a dimensionless Quantity, but also the brightness or contrast of the image is constant, so it can provide an absolute measure of feature points. Experiments show that the phase overlap extraction feature does not change the illumination (or contrast) of the image and the possible noise