A multi-sensor image fusion algorithm based on SCDPT transform is proposed in this paper. SCDPT transform is used to decompose source images in each scale and direction to get low-pass sub-band coefficients and band-pass directional sub-band coefficients. The principle for low-pass sub-band coefficients is based on structural similarity (SSIM), regional energy and regional average gradient, while the principle for directional band-pass sub-band coefficients is based on SSIM and regional variance. Finally, fused image is obtained by SCDPT inverse transform. The proposed method is compared to the wavelet transform, Laplacian pyramid transform and gradient pyramid transform. Our algorithm not only has more flexible directional and shift invariance, but also is able to accurately capture the image information of the contour feature and texture details.