The work aims to improve the visual quality of fusion image. The focus image fusion algorithm with regional multi-features and improved DS evidence theory rule is proposed. First of all, the 2nd generation Curvelet transform was introduced to decompose source image and obtain the coarse scale coefficients and fine scale coefficients. Then, according to the absolute values of the coarse scale coefficients in the selected region, the maximum fusion rule was constructed to fuse the coarse scale coefficients. In combination with such features as regional variance, information entropy and regional energy, the regional features of the fine scale layers were extracted. Moreover, by defining the probability constraints, the fusion rule of DS evidence theory was improved, the credibility of DS fusion rule was enhanced, and the fine scale coefficients of the image were effectively fused, so that the fuse image could preserve more details. Finally, the image fusion was completed by inverse Curvelet transform. The simulation results showed that, compared with the current image fusion algorithm, the fuse image of the proposed algorithm can preserve more abundant details and have better visual effect and its fusion is less time-consuming. It is concluded that the proposed algorithm has taken into account the cross correlation between pixels and further optimized the image fuse quality, and it can be applied to such fields as remote sensing and package printing detection.