In order to effectively find the appropriate combination of data augmentation (CutBlur, CutOut, etc.) to control/improve the quality of super-resolution, we propose a method to visualize the relationship of evaluation indices (PSNR, Affinity, and Diversity) and control valuables as data augmentation based on the Self-Organizing MAP (SOM). Using the proposed method, it is possible to visualize the results of a small number of trials, to grasp the characteristics of the combination of each data augmentation and their effectiveness for the quality control of super-resolution. Furthermore, in the case of super-resolution considered as a pre-processing step for various image processing (e.g., the image classification and 3D reconstruction), the proposed method can also provide an effective combination of data augmentation. To investigate the effectiveness of the proposed method, the enhanced deep super-resolution network (EDSR) model is employed by using 16 combinations of data augmentation, CutBlur, RGB, Cutout, and CutMix, and the evaluation indices, PSNR, Affinity, and Diversity are measured. In the case of the experiment using data sets, Manga109, Urban100, the quality of super-resolution images obtained by the EDSR optimized by the proposed method, is improved the PSNR by 0.08 (dB) over the original super-resolution. Furthermore, we applied the proposed super-resolution as pre-processing of the image classification and 3D reconstruction, and it is confirmed that the accuracy and quality of the image classification (CIFAR-10) and 3D-reconstruction (buckwheat real field images) based on the proposed method are improved compared with the original ones. Comparing the number of cases without super-resolution and with super-resolution based on the proposed method, the number of keypoints is approximately 4.9 times and the number of matching points is approximately 2.2 times, indicating that, from a quantitative perspective.