Multi-scale Harris-corner Detection Algorithm Based on Region Detection

WU Peng,XU Hongling,LI Wenlin,SONG Wenlong,ZHANG Jiawei

DOI: https://doi.org/10.11990/jheu.201507028

2016-01-01

Abstract:In this paper, we address the problems associated with the Harris?corner detection algorithm, in which corners are extracted from the entire image and there is sensitivity to scale change. We combine region detection and the multi?scale Harris?corner detection algorithm, such that fewer corners are detected and more efficient subse?quent image matching and reconstruction is achieved. First, we use the region detection algorithm to construct an image saliency map, and use the erosion and dilation operations to extract the target region as a candidate detection region. Next, multi?scale method combines with non?maxima suppression method is used to improve Harris algo?rithm, to detect the corners and mark. Simulation results show that the method achieves better accuracy and speed in corner detection. Furthermore, without changing any parameters, this algorithm reduces the differences due to the extracted corners and enhances the multi?scale of the algorithm for image rotation.