Resolution Effect of Training Sets in Several Super-resolution Algorithms

In this paper, we mainly focus on the effect of training data set on the image reconstruction result of super-resolution algorithm based on deep learning. The main work is divided into two parts. ID the first part, we compare three representative algorithms of super resolution image reconstruction based on interpolation with three representative algorithms of super resolution image reconstruction based on deep learning under the same evaluation index. In the second part, we obtained four sets of training data sets with different resolutions by sampling the same data set with different multipliers, and used the three super-resolution algorithms based on deep learning to train the four sets of training data. The final experimental results not only show that the super-resolution algorithm based on deep learning is significantly better than the traditional super-resolution algorithm based on interpolation in the quality of image reconstruction, but also that without considering the influence of objective factors, the greater the image resolution of the training data set, the better the final image reconstruction effect.