A method for evaluating fusion quality of remote sensing images is provided. Considering the influence of spectral domain distortion and spatial domain distortion on the image quality of remote sensing fusion, the method extracts red-band images, green-band images, blue-band images, near infrared-band images, normalized vegetation index images, normalized difference water index images, and road and building index images, and the original multivariate Gaussian model of training sample set is constructed by extracting statistical eigenvectors of all sub-blocks. In the test phase, by acquiring the original multivariate Gaussian model of test sample set composed of test remote sensing fusion images, according to the original multivariate Gaussian model constructed in the training stage, the objective quality evaluation prediction value of the test remote sensing fusion image is predicted. Because the statistical feature vector information obtained can reflect the quality change of the remote sensing fusion image, the correlation between the objective evaluation result and the subjective perception is effectively improved.