Abstract: The traditional algorithm of image matching exist the problems of little feature information
and high rate false match. An image matching algorithm is presented based on SURF feature
extraction and FLANN search. Firstly, the extremum value of local image is gotten using the Hessian
matrix. Secondly, the feature vector is simultaneously processed in multilayer image scale space by
using of different size feature description. Finally, the FLANN algorithm is used for feature matching.
The experiments show that this algorithm is better than the traditional algorithm of image matching in
the aspect of effectiveness and efficiency.

Key words: speed up robust features, Hessian, FLANN, image matching