Image matching is an important technique in image processing and recognition. In order to enhance the matching effectiveness and to obtain accurate matches

an iterative graph transformation matching algorithm is proposed in this paper to remove the error matches

which usually exist in the initial matches obtained by image features such as scale invariant feature transform (SIFT) and speeded up robust features (SURF). Generally

the proposed algorithm carries out the following three steps: first

the algorithm generates the accurate correspondences from the initial one-to-one correspondence set by using the traditional graph transformation matching algorithm

whose process is similar to that in graph transformation matching (GTM) algorithm; then

it further revises the initial correspondences by using the geometric relationship between the obtained correct matches and the initial matches; finally

based on the revised initial matches

the algorithm further searches the correct matches from the revised initializations by using graph transformation matching algorithm. Compared with GTM

the proposed algorithm further explores the geometric relationship in the matching process and thus returns more accurate matches. Experimental results on real-world image matching shows that under various image transformation scenes

the proposed algorithm can significantly outperform GTM on the matching recall while retain the similar high matching precision. The proposed iterative graph transformation matching algorithm can overcome the main drawback of GTM that usually returns small number of precise matches.