基于相位叠合的不变特征提取

Invariant feature detection based on phase congruency

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)

In this paper, a kind of feature extraction method suitable for a variety of features (step edge, line, roof edge, Mach band etc) is presented. This method makes use of the concept which Fourier components have maximum congruency at feature points (phase congruency). And using phase congruency for marking features has significant advantages over gradient based methods. It is a dimentionless quality that is invariant to changes in image brightness or contrast, hence it provides an absolute measure of the significance of feature point. Experimental results prove that extracted features using phase congruency is invariant to image illuminance (or contrast) and noise. \;