Abstract: In this paper, a new method based on multiple regressions is proposed to determine the weight distribution of the digital ray image. Firstly, the factors that affect the quality of digital ray images are analyzed. Twelve factors which affect the quality of images are selected in this paper. The variables that are not significant in t inspection are eliminated and improved by model diagnosis. Finally, a multivariate regression model, including mean times, gray standard deviation, noise, distortion, structural similarity, geometric magnification being as variables and subjective evaluation results, was constructed. The experimental results show that the composite index of the image obtained by this method is highly fitting with subjective evaluation results. The problems of subjective judgment being inconsistent with objective one were solved. The disadvantages of artificial subjective image quality are effectively avoided.