基于相位一致性的兴趣点检测方法

针对当前大多数兴趣点检测方法中没有统一阈值的问题,提出了一种基于对数-G abor小波计算兴趣点的相位一致性的方法。具有最大相位一致性的点对应于图像中各种类型的二维特征点,包括灰度角点、直线端点以及各种类型的交点。由于相位一致性是1个无量纲的度量,它对图像亮度和对比度变化具有不变性,因此可以实现兴趣点检测中统一阈值的设定。实验表明,该方法能够实现兴趣点的准确检测和精确定位。

Improving Precision of Location Determination of Interest Point in Image Detection with Phase Congruency Based Approach

Purpose.Most existing methods for determining location of interest point in image detection are,in our opinion,still deficient in precision.We present a phase congruency approach that uses log-Gabor wavelets and can improve such precision considerably because phase congruency is a dimensionless measure that remains invariant when changes occur in brightness and contrast.In the full paper,we explain our phase congruency approach in much detail;here we just list the three topics discussed in the detailed explanation:(1) log-Gabor filter; (2) phase congruency,whose subtopics are phase congruency of 1D signals(2.1),phase congruency of 2D signals(2.2);the subsubtopics of subtopic 2.2 are design of 2D filters(2.2.1),phase congruency of 1D features(2.2.2) and phase congruency of 2D features(interest points)(2.2.3);(3) flow chart of phase congruency.Experimental data for synthetic and real images are presented respectively in Fig. 3 and Fig.4,each with 4 subfigures;the 8 subfigures show preliminarily that different kinds of interest points can be detected and located with good precision;thus the proposed method can be applied over wide classes of images.