Abstract: Whether the image matching was good or not was an important factor affecting the quality of image stitching, and the traditional image matching method was sensitive to radiation distortion. In order to solve this problem, a robust image matching method was proposed. The phase consistency model which can be used to detect image features was obtained through the calculation of log-gabor filter and image convolution, and then the standardized maximum index map with rotation invariance was constructed by considering the results of log-gabor convolution in all directions. Finally, the distributed histogram method was used to construct feature descriptors on the maximum index map to achieve image matching. Matching experiments are carried out on remote sensing images with radiation distortions. When the image to be matched is rotated, the number of correct matches of the proposed method is more than 50, and the Root Mean Square Error (RMSE) is 1.6 pixel - 2.1 pixel, which is better than the traditional Radiation Invariant Feature Transformation (RIFT) algorithm. The experimental results shows that the proposed method is an effective feature matching algorithm.