In the fusion of multi-source remote sensing image, for different ground region, usually the characteristics of spectral features and spatial detail information are different. In view of the problem, a new salient analysis model called IR (adaptive-radius-search) salient analysis model was proposed, which realized the identification and classification of the salient regions and non salient regions. Combined with HIS and wavelet transform fusion algorithm, an improved adaptive HIS fusion method was put forward based on average gradient weighting, which achieved the fusion of the road, farmland, residential area and other salient regions, and the method had better keeping the spatial details information of its abundant space. Meanwhile, the fusion method of HIS+WT transform was proposed, which realized the fusion of mountain, forest land, and other areas are not salient, retaining more spectral information. The method completed the adaptive remote sensing image fusion of different regions. The experimental results show that the proposed method of the paper could guarantee the fusion image of both to maintain high spatial detail performance ability, but also to retain good spectral information. The proposed method of the paper provides certain theoretical basis and application value for the research on agricultural science, forest planning and future forest remote sensing image classification, identification and so on.