Search

Image registration is a crucial step in the field of computer vision. However, the traditional scale invariant feature transform (SIFT) based method often suffers from many mismatches when being utilized to register unmanned aerial vehicle (UAV) images with obvious rotation, viewpoint change and similar textures. In order to tackle this problem, we formulate the image matching problem as a Markov Random Field (MRF) energy minimization problem and propose an accurate and fast image registration framework. First, a SIFT algorithm is utilized to extract keypoints and an algorithm called “Center of Mass” (CoM) is exploited to assign a single orientation for every point instead of SIFT since SIFT may cause ambiguity in subsequent process; second, based on the local adjacent spatial relationship between a feature point and its neighboring points, a new concept of local spatial constraints is proposed to characterize the geometric consistency between them and incorporated into a belief propagation algorithm to obtain initial matching results; finally the residual mismatches are eliminated by the Random Sample Consensus (RANSAC) algorithm, meanwhile the transformation parameters are estimated. Experiment results demonstrate that the proposed framework can significantly enhance the registration performance in UAV image registration. [ABSTRACT FROM AUTHOR]

Accurate image registration is a vital step in many computer vision processes. However, traditional SIFT based methods are not able to obtain satisfactory results in some cases. In this paper, we turn the matching problem into a Markov Random Field (MRF) optimization problem and propose a fast and accurate image registration method based on message passing to screen out mismatches caused by the conventional SIFT-based algorithm. The proposed method comprises three steps. First, a SIFT detector is used to detect key points and then these points are described by Daisy instead of the SIFT detector (to improve efficiency); second, a new geometric constraint with distance and direction terms is designed and incorporated into sequential tree-reweighted message passing (TRW-S) – this obtains the coarse matching result; finally, the RANSAC algorithm is adopted to filter out the mismatches and obtain an accurate transformation. The experimental results demonstrate that the proposed method significantly improves the efficiency and the accuracy of the registration process and secures a desirable result. [ABSTRACT FROM AUTHOR]