An image matching algorithm based on KAZE and Perceptual Hash is proposed to solve the mismatching problem in image feature matching. Firstly, it uses KAZE algorithm to extract and match the image feature points. Then the perceptual Hash algorithm is used to filter and purify the initial matching pairs. Finally, random sampling consensus (RANSAC) algorithm is adopted to remove the false matching points, and thus the final matching results are obtained. Experiments performed on Mikolajczyk standard dataset indicate that the matching accuracy of the proposed algorithm is high with image changes in rotation, scale, illumination, blurring and JPEG compression, and has good reliability and robustness.