1: \begin{abstract}

2: The inverse renormalization group is studied based on the image 

3: super-resolution using the deep convolutional neural networks. 

4: We consider the improved correlation configuration instead of 

5: spin configuration for the spin models, such as the two-dimensional 

6: Ising and three-state Potts models. 

7: We propose a block-cluster transformation as an alternative to 

8: the block-spin transformation in dealing with the improved estimators. 

9: In the framework of the dual Monte Carlo algorithm, 

10: the block-cluster transformation is regarded as a transformation in the 

11: graph degrees of freedom, whereas the block-spin transformation 

12: is that in the spin degrees of freedom. 

13: We demonstrate that the renormalized improved correlation configuration 

14: successfully reproduces the original configuration at all the temperatures 

15: by the super-resolution scheme. Using the rule of enlargement, 

16: we repeatedly make inverse renormalization procedure 

17: to generate larger correlation configurations.  

18: To connect thermodynamics, an approximate temperature rescaling is discussed. 

19: The enlarged systems generated using the super-resolution 

20: satisfy the finite-size scaling.

21: \end{abstract}