Abstract: In order to solve the uncertainty of the scaling effect of the traditional content-aware image scaling technique, and the method is not applicable to the multi-spectral remote sensing image, an image scaling method combining triangular mesh and fish-eye transformation is proposed. Firstly, the sensitive target area is marked according to the feature identification of the sensitive target. Then, the non-uniform scaled transformation based on triangular mesh is performed inside the sensitive area, and the remaining part is transformed according to the zooming change of the sensitive area. The experimental results show that the image-sensitive region of the internal distortion is very small and the characteristics of sensitive target are well maintained. The scaling effect is significantly improved compared to the traditional content-aware image scaling method.