Spatial Variation Motion Blur Bridge Image Quality Improvement

China is a leading nation in bridge construction, with bridge infrastructure not only forming a critical part of the transportation network but also driving regional economic development. To ensure the safety and reliability of these structures, regular inspection and maintenance are required, with crack detection being particularly crucial. With technological advancements, machine vision has started to be applied in bridge health monitoring. However, due to issues such as low light, noise, and camera shake, images often exhibit motion blur, leading to the loss of key details and structural deformation, which affects the accuracy of crack detection. To address this, this paper proposes an image restoration algorithm that incorporates a strip attention mechanism, leveraging intra- and inter-strip attention blocks within the Transformer framework to handle blurs of different directions and magnitudes. Additionally, an operator is used for sharpening to reduce artifacts and noise, thereby improving the deblurring effect. Experimental results demonstrate that this algorithm effectively enhances image quality, improving the reliability of data used for bridge health management.