CNB Net: A Two-Stage Approach for Effective Image Deblurring

Xiu Zhang,Fengbo Zheng,Lifen Jiang,Haoyu Guo

DOI: https://doi.org/10.3390/electronics13020404

IF: 2.9

2024-01-19

Electronics

Abstract:Image blur, often caused by camera shake and object movement, poses a significant challenge in computer vision. Image deblurring strives to restore clarity to these images. Traditional single-stage methods, while effective in detail enhancement, often neglect global context in favor of local information. Yet, both aspects are crucial, especially in real-life scenarios where images are typically large and subject to various blurs. Addressing this, we introduce CNB Net, an innovative deblurring network adept at integrating global and local insights for enhanced image restoration. The network operates in two stages, utilizing our specially designed Convolution and Normalization-Based Block (CNB Block) and Convolution and Normalization-Based Plus Block (CNBP Block) for multi-scale information extraction. A progressive learning approach is adopted with a Feature Active Selection (FAS) module at the end of each stage that captures spatial detail information under the guidance of real images. The Two-Stage Feature Fusion (TSFF) module reduces information loss caused by downsampling operations while enriching features across stages for increased robustness. We conduct experiments on the GoPro dataset and the HIDE dataset. On the GoPro dataset, our Peak Signal-to-Noise Ratio (PSNR) result is 32.21 and the Structural Similarity (SSIM) result is 0.950; and on the HIDE dataset, our PSNR result is 30.38 and the SSIM result is 0.932. Our results exceed other similar algorithms. By comparing the generated feature maps, we find that our model takes into account both global and local information well.

engineering, electrical & electronic,computer science, information systems,physics, applied

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to effectively integrate global and local information in image deblurring techniques to improve the quality of image restoration. Traditional methods often ignore the global context while enhancing details, resulting in poor performance when dealing with large - sized images in real - world scenarios. To overcome this limitation, the author proposes a new model named CNB Net, which extracts multi - scale information through a two - stage method and combines global and local features to achieve higher - quality image deblurring.

Specifically, the main contributions of CNB Net include:
- Proposing the CNB Net model composed of CNB Block, CNBP Block, TSFF module and FAS module. These modules are respectively used to extract multi - scale features and learn global and local information.
- The experimental results on the GoPro and HIDE datasets show that CNB Net outperforms existing methods in evaluation metrics such as Peak Signal - to - Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM).
- By analyzing the generated feature maps, it is proved that the model can well balance global and local information.

Through this method, the paper aims to solve the shortcomings of existing single - stage deblurring methods in dealing with the global context, thereby improving the overall performance of image deblurring.