Image matching algorithm for tire impression based on SIFT-Gabor transform

An improved Scale Invariant Feature Transform (SIFT) feature matching algorithm based on image Gabor transform was introduced to solve the problems of large calculating scale and high complexity for traditional feature matching algorithms and to balance the accurate key-point detection and location for Gabor wavelet transform. The feature points in a tire impression image were detected firstly with the SIFT transform and multi-scale and multidirectional values were given by the tire impression image. Then, Gabor transform on key-point were adopted as SIFT feature vector descriptors to reduce the dimensions of SIFT feature vector and to improve the efficiency of feature matching. Experimental results show that the threshold can be set as 0.7 by comparing the matching result of different thresholds for the tire impression image. The false non-match rate is 0.2% when corresponding point pairs are 15. The matching time of the Gabor-SIFT algorithm is less than that of standard SIFT algorithm for reducing dimensions of feature vector. The improved algorithm has higher matching accuracy and less consuming time, so it is suitable for matching the tire impressions image.