Fast image registration based on SURF-DAISY algorithm and randomized kd trees
Baigen He · Journal of Optoelectronics·laser · 2012
A new method for fast image registration based on SURF-DAISY algorithm and randomized kd trees is proposed.Firstly,feature points are extracted using the classic SURF feature detector in the reference image and sensed image,respectively.Then,the DAISY descriptor is utilized to substitute for the customary SURF description algorithm to characterize those feature points to speed up the process.After that,the matching process is carried out via multiple randomized kd trees and the RANSAC is used to eliminate wrong matches.At last,the best transform parameters are estimated by the least square method and the image registration process is accomplished.The experimental results indicate that the proposed method reduces the cost time of the registration process up to 45.6% compared with the classic SURF algorithm while nearly retaining the same performance.