Abstract: SIFT has been widely used for its good performance on feature extraction and matching．However,effects under conditions of insufficient illumination and blur are not so satisfactory．We proposed an adaptive threshold selection method based on global and local information．First,initial threshold can be obtained according to image contrast,thus it’s adapted to insufficient illumination and image blur．Second,in order to control the number and distribution of feature points,the threshold is adjusted secondly according to feature points distribution．Finally,the mismatch removing method has also been improved．Experiment results show that the improved SIFT algorithm is not only well adapted to low light and image blur,but also can adjust feature point numbers and reduce clustering effects.

Key words: SIFT,Threshold,Image matching