Abstract: This paper presented a new image matching algorithm based on improved sift and shape-context,and aimed to solve the disadvantages of conventional sift and shape-context algorithm.We took color invariant into consideration and constructed color sift descriptors.In shape context algorithm,shape context histogram based on central points is used instead of the traditional shape context histogram based on contour points.Then the new joint descriptors combining with sift and shape context are applied to lead the feature points matching according to the new given joint distance and achieve the initial matching pairs.Finally,the partial least squares method is used to eliminate mismatching points.The experimental results show that the proposed algorithm can improve image matching accuracy effectively.

Key words: SIFT,Color invariance,Shape context,Descriptor,Image matching,PLS,Mismatching