The invention discloses a super-resolution image quality assessment method based on a deep residual network. According to the method, during assessment of the quality of a super-resolution image, an original image does not need to be regarded as the reference, and an image quality score in accordance with subjective assessment of the human vision can be directly given. Currently, most super-resolution images are assessed through a full reference index, and it is very difficult or even impossible to obtain a reference image of a to-be-assessed image in many practical applications. Besides, mostrelated algorithms still need to extract image characteristics through a manual design. To solve the problems, the invention provides a non-reference assessment method in accordance with the qualityof the super-resolution image in accordance with the perception of the human vision. The innovation of the method is automatic extraction of the characteristics by direct adoption of the designed deepresidual network instead of the manual design characterizes. The method also proposes a method for grade ranking based on a distorted image, through which the problem of insufficient super-resolutionimage data sets containing human subjective scores can be effectively solved.