A new self-adaptive multi-sensor image fusion algorithm based on Curvelet transform is proposed. Full-color and multispectral images are decomposed and transformed using the algorithm,and different fusion rules are selected according to different frequency characteristics. Self-adpative regional energy coefficients fusion rule and regional feature self-adaptive fusion rule are selected for low- and high-frequency coefficients. The final fusion image is obtained by reconstruction. Compared with other fusion algorithms this algorithm is a feasible and effective image fusion algorithm.