The technology of image matching is very important in the field of image analysis. Studying on image correlation is the key to improve the accuracy of image matching, while the fractal self-similarity has the unique potential in making full use of image correlation. This paper put forwards a matching algorithm of the sub-pixel image matching based on fractal dimension. Firstly, by searching template subarea based on traverse in being matched image, and the sub-subareas are classified. Then, the fractal dimension of subareas and sub-subareas is calculated by using the optimal box counting method. The similarity maximum position can be found according to the statistical correlation. Finally, fractal interpolation method is used to improve the precision of image matching. Both theoretical analysis and experimental results show that this approach has good matching effect.