Detection of Line Features and Types Based on Phase Information

Line features of image are perceived at points where the Fourier components are maximally in phase. Phase congruency is a normalized dimensionless value, it is invariant to changes of image brightness and contrast, and easy to set threshold. But line feature types are no further distinguished by former methods based on phase congruency, the types of line features are considered in the same case. Actually, types of line features are different, and usually divided into three kinds of edges: the rising edge; the falling edge and the roof edge, phase of each edge is different. Correctly distinguished line types will improve the reliability of image matching and registration, and it is important to extract line features automatically. In this paper, detecting line features and distinguishing their types can be achieved simultaneously by the information of phase and phase congruency which is used by log-Gabor wavelets. Experiments show that line features are accurate detected and the types of different edges (such as the rising edge, the falling edge and the roof edge) are distinguished by this approach.