When an image has multiple similar regions

SIFT descriptors are similar

which results in much mismatches. Aiming at the problem presented above

a improved SIFT features matching algorithm with global context vector is presented. Firstly

detect feature points in scale space . Then descriptors which consist of a SIFT descriptor representing local properties and a global context vector to disambiguate locally similar features are computed. Finally

according to BBF searching strategy

match the feature vectors using Euclidean distance. The experimental results show that the improved algorithm which can describe feature points in a larger region reduces mismatch probability of experimental images from 19% to 11%

which improves matching results greatly.