A hierarchical approach for image registration using line features
Chuntao Wei, Zuxun Zhang, Jianqing Zhang · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2008
Image registration is a fundamental process in many remote sensing applications, such as image fusion, temporal change detection. Generally, Algorithms for image registration can be classifeed into two categories: featurebased and area-based methods. feature-based methods are relatively fast, robust and reliable, and area-based methods can get high accuracy with high computational cost. Because the result produced by traditional feature detector may vary with image contrast, it is dificult to set appropriate thresholds automatically for the reference image and sensed image. To solved these problem, an automatic approach for image registration is presented in this paper. It use a feature-based approach to get a coarse registration at first. Then, area-based method used to improve the accuracy of the result. In feature detection stage, it employs a feature detector implemented in frequency domain to obtain features with normalized measure. Constant thresholds can be applied for different images. Due to feature matching is time-consuming and computation expensive, line features detected from the images with approximate direction are mapped into Hough space to estimate the transformation parameters with Modified Iterative Hough Transform method. Furthermore, it use a hierarchical framework to speed up the registration process. The experiments show that the approach mentioned above is feasible and efficient.