Quaternion Phase Congruency model for edge saliency map extraction and image blur measurement

Zihao Feng, Yi Xu, Xiaokang Yang

Abstract: Image edge detection is one of the most important foundations of image analysis and comprehension. Traditional edge detectors are not robust enough to illumination variations. In last decade, detector based on Phase Congruency (PC) model was proposed and invariant to illumination variations. In this paper, we introduce the quaternion Wavelet Transform to PC model and propose a new edge metric of quaternion Phase Congruency (QPC). QPC model overcomes one of the disadvantages of PC model. It is reliable to distinguish the significant skeleton structures from the relatively trivial texture structures. Experimental results show that our new model provides a more consistent edge saliency metric with human perception.