The invention relates to a three-step hyperspectral image fusion method based on spectrum reconstruction, i.e., principal component analysis image fusion is perfected through processing methods of wave band selection and spectrum reconstruction, so that a better fusion effect is obtained. The method is as follows: 1) hyperspectral image data and high spatial resolution image data to be fused are obtained through a human-computer interaction interface module, and related parameters are initialized; 2) through a hyperspectral image wave band selection module, a wave band is designed and a matrix is selected, and wave band selection is performed on an initial hyperspectral image; 3) through a hyperspectral image-high spatial resolution image data fusion module, principle component analysis fusion is performed on an input image; 4) through an spectrum reconstruction module, high-precision spectrum reconstruction is performed on an initial fusion image; and 5) through a fusion result output module, a final fusion result is output. The three-step hyperspectral image fusion method based on spectrum reconstruction has the advantages that the data volume to be fused is reduced, the spectrum distortion is reduced, the spectrum retentivity is improved, the image quality is enhanced, and the fusion quality is good, and the method has a wide range of application.