An Image Matching Algorithm Based on Bipartite Graph

Inhalt:
In view of the low accuracy and a lot of false matches of traditional image matching algorithm. In order to obtain more accurate results and estimate the transformation between images accurately, this paper proposes an feature points’ matching algorithm based on bipartite graph optimal matching. Firstly, extracting feature points and descriptors from query images and training images and building bipartite graph based on feature points and Euclidean distance of descriptors. The points in the graph are the feature points in two images, one side corresponds to the query image, and the other side corresponds to the training image. The edges are the distances of two points. Further more, setting a threshold where edges with distances less than this threshold are preserved, so a sparse bipartite graph can be obtained. The Time Complexity of this algorithm will decrease because of sparsity. Finally, using bipartite graph optimal matching algorithm Kuhn-munkres to get image matches. The results of experiments show that the quality of image matching is improved.