In order to solve the problem of difficulty and low accuracy of registration between cross-scale data, we propose a cross-scale data registration method based on fractal dimension characterization. The discrete wavelet transform algorithm is used to decompose the raw data to obtain its approximation at each scale. Then the fractal dimension is adapted to characterize the scale parameter of the decomposed data. According to the fractal dimension, we pick out the scale-approximated data from different raw data, using iterative closest point algorithm for fine registration as a usual case. The calculation result of fractal dimension at different scales shows that the fractal dimension can represent scale parameters effectively. And the registration results indicate that the proposed registration method can achieve the registration of cross-scale effectively.