Fig. 1. Basic architectures.

(a) The architecture of our proposed self-similarity enhancement network;

(b) The cross-level feature enhancement module; (c) The pooling attention dense blocks

Fig. 2. The proposed feature enhancement module

Fig. 3. Receptive field block

Fig. 4. The proposed Cross-Level Co-Attention architec-ture. "F_gp" denotes the global average pooling

Fig. 5. Schematic illustration of the pooling attention

Fig. 6. Super-resolution results of " Img048" in Urban100 dataset for 4× magnification

Fig. 7. Super-resolution results of " Img092" in Urban100 dataset for 4× magnification

Fig. 8. Super-resolution results of " 223061" in BSD100 dataset for 4× magnification

Fig. 9. Super-resolution results of " 253027" in BSD100 dataset for 4× magnification

Fig. 10. Convergence analysis on CLFE and PADB. The curves for each combination are based on the PSNR on Set5 with scaling factor 4× in 800 epochs.

Fig. 11. Results of each module in the network.

(a) The result of first layer convolution; (b) The results of cross-level feature enhancement module;

(c) The results of Stacked pooling attention dense blocks