Swin-VETnet: A multi-source blur removal method for mobile video terminals

Factors such as motion blur, camera shake, and environmental scattering contribute to multiple sources of blur in the video, which degrade video quality. To enhance the performance of mobile video systems across a variety of environments, it is essential to address these multiple blur sources. Existing deblurring methods are typically tailored to static scenes or fail to handle the complex, dynamic nature of mobile video. This paper introduces the Swin-VETnet model, a novel approach to deblurring for mobile video that accounts for multiple blur sources, including atmospheric scattering, motion blur, and sensor noise. The model is built on high-resolution, real-world mobile video datasets with varying blur types. By leveraging the Swin Transformer’s capabilities in both spatial and temporal domains, the model achieves time-space consistent deblurring, restoring clarity across dynamic video content. Experimental results demonstrate that Swin-VETnet significantly improves deblurring performance by effectively handling the challenges posed by diverse blur sources in mobile video.