Dynamic Structure-Aware Modulation Network for Underwater Image Super-Resolution
Author Information
Author(s): Wang Li, Li Ke, Dong Chengang, Shen Keyong, Mu Yang
Primary Institution: Nanjing Vocational University of Industry Technology
Hypothesis
Can a dynamic structure-aware modulation network improve the quality of underwater images?
Conclusion
The proposed DSMN outperforms existing super-resolution methods while being computationally efficient.
Supporting Evidence
- The DSMN effectively captures both local and global image features.
- The Mixed Transformer enhances the recovery of degraded image details.
- The DIMM modulates the Mixed Transformer output based on input statistics.
- The HAFM aggregates features to improve image reconstruction quality.
Takeaway
This study created a smart network that helps underwater pictures look clearer and more detailed, just like how our eyes adjust to see better underwater.
Methodology
The study used a Mixed Transformer architecture with a dynamic information modulation module and a hybrid-attention fusion module to enhance underwater image quality.
Limitations
The model's reconstruction of structure and detail needs improvement, and the up-sampling strategy does not adapt to image content.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website